Publications

Preprints

1. Resting-state Functional Connectivity Predicts Cochlear-Implant Speech Outcomes

   Jamal Esmaelpoor, Tommy Peng, Beth Jelfs, Darren Mao, Maureen Shader, and Colette McKay

   medRxiv:10.1101/2024.01.30.24301908

   abstract doi pdf

   Background: Cochlear implants (CIs) have significantly improved hearing restoration for individuals with severe or profound hearing loss. Nonetheless, a substantial variability in CI outcomes remains unexplained, even when considering subject-specific factors such as age and duration of deafness. This study investigates the use of resting-state functional near-infrared spectroscopy (fNIRS) recordings to predict speech understanding outcomes after cochlear implantation. Our hypothesis posits that resting-state functional connectivity reflects brain plasticity following hearing loss and implantation, and that the average clustering coefficient in brain functional connectivity networks can capture this variation among CI users. Methods: Twenty-two cochlear implant candidates participated in this study. Resting-state fNIRS data were collected pre- implantation and at one month, three months, and one year post-implantation. Speech understanding performance was assessed using CNC words in quiet and BKB sentences in noise one year post-implantation. Resting-state functional connectivity networks were constructed using regularized partial correlation, and the average clustering coefficient was measured in the signed weighted networks as a predictive measure for implantation outcomes. Results: Our findings demonstrate a significant correlation between the average clustering coefficient in resting-state functional networks and speech understanding outcomes. Importantly, our analysis reveals that this measure provides unique information not accounted for by subject-specific factors such as age and duration of deafness. Conclusion: This approach utilizes an easily deployable resting-state functional brain imaging metric to predict speech under-standing outcomes in implant recipients. The results indicate that the average clustering coefficient, both pre and post implantation, correlates with speech understanding outcomes.

2. Mapping Extended Landmarks for Radar SLAM

   Shuai Sun, Christopher Gilliam, Kamran Ghorbani, Glenn Matthews, and Beth Jelfs

   arXiv:2210.17207

   abstract arXiv pdf

   Simultaneous localization and mapping (SLAM) using automotive radar sensors can provide enhanced sensing capabilities for autonomous systems. In SLAM applications, with a greater requirement for the environment map, information on the extent of landmarks is vital for precise navigation and path planning. Although object extent estimation has been successfully applied in target tracking, its adaption to SLAM remains unaddressed due to the additional uncertainty of the sensor platform, bias in the odometer reading, as well as the measurement non-linearity. In this paper, we propose to incorporate the Bayesian random matrix approach to estimate the extent of landmarks in radar SLAM. We describe the details for implementation of landmark extent initialization, prediction and update. To validate the performance of our proposed approach we compare with the model-free ellipse fitting algorithm with results showing more consistent extent estimation. We also demonstrate that exploiting the landmark extent in the state update can improve localization accuracy.

3. fMRI-based Static and Dynamic Functional Connectivity Analysis for Post-stroke Motor Dysfunction Patient: A Review

   Kaichao Wu, Beth Jelfs, Katrina Neville, and John Q. Fang

   arXiv:2301.07171

   abstract arXiv pdf

   Functional magnetic resonance imaging (fMRI) has been widely utilized to study the motor deficits and rehabilitation following stroke. In particular, functional connectivity(FC) analyses with fMRI at rest can be employed to reveal the neural connectivity rationale behind this post-stroke motor function impairment and recovery. However, the methods and findings have not been summarized in a review focusing on post-stroke functional connectivity analysis. In this context, we broadly review the static functional connectivity network analysis (SFC) and dynamic functional connectivity network analysis (DFC) for post-stroke motor dysfunction patients, aiming to provide method guides and the latest findings regarding post-stroke motor function recovery. Specifically, a brief overview of the SFC and DFC methods for fMRI analysis is provided, along with the preprocessing and denoising procedures that go into these methods. Following that, the current status of research in functional connectivity networks for post-stoke patients under these two views was synthesized individually. Results show that SFC is the most frequent post-stroke functional connectivity analysis method. The SFC findings demonstrate that the stroke lesion reduces FC between motor areas, and the FC increase positively correlates with functional recovery. Meanwhile, the current DFC analysis in post-stroke has just been uncovered as the tip of the iceberg of its prospect, and its exceptionally rapidly progressing development can be expected.

Journal Articles

1. Dynamic Reconfiguration of Brain Functional Network in Stroke

   Kaichao Wu, Beth Jelfs, Katrina Neville, Seedahmed S. Mahmoud, Wenzhen He, and Qiang Fang

   IEEE Journal of Biomedical and Health Informatics, 2024, pp. 1–11.

   abstract doi arXiv pdf

   The brain continually reorganizes its functional network to adapt to post-stroke functional impairments. Previous studies using static modularity analysis have presented global-level behavior patterns of this network reorganization. However, it is far from understood how the brain reconfigures its functional network dynamically following a stroke. This study collected resting-state functional MRI data from 15 stroke patients, with mild (n = 6) and severe (n = 9) two subgroups based on their clinical symptoms. Additionally, 15 age-matched healthy subjects were considered as controls. By applying a multilayer temporal network method, a dynamic modular structure was recognized based on a time-resolved function network. The dynamic network measurements (recruitment, integration, and flexibility) were calculated to characterize the dynamic reconfiguration of post-stroke brain functional networks, hence, revealing the neural functional rebuilding process. It was found from this investigation that severe patients tended to have reduced recruitment and increased between-network integration, while mild patients exhibited low network flexibility and less network integration. It’s also noted that previous studies using static methods could not reveal this severity-dependent alteration in network interaction. Clinically, the obtained knowledge of the diverse patterns of dynamic adjustment in brain functional networks observed from the brain neuronal images could help understand the underlying mechanism of the motor, speech, and cognitive functional impairments caused by stroke attacks. The present method not only could be used to evaluate patients’ current brain status but also has the potential to provide insights into prognosis analysis and prediction.

Journal Articles

1. Virtual Reality and Motor Imagery for Early Post-Stroke Rehabilitation

   Chi S. Choy, Qiang Fang, Katrina Neville, Bingrui Ding, Akshay Kumar, Seedahmed S. Mahmoud, Xudong Gu, Jianming Fu, and Beth Jelfs

   BioMedical Engineering OnLine, 2023, vol. 22, no. 66. pp. 1–11.

   abstract doi pdf

   Background: Motor impairment is a common consequence of stroke causing difficulty in independent movement. The first month of post-stroke rehabilitation is the most effective period for recovery. Movement imagination, known as motor imagery, in combination with virtual reality may provide a way for stroke patients with severe motor disabilities to begin rehabilitation. Methods: The aim of this study is to verify whether motor imagery and virtual reality help to activate stroke patients’ motor cortex. 16 acute/subacute (< 6 months) stroke patients participated in this study. All participants performed motor imagery of basketball shooting which involved the following tasks: listening to audio instruction only, watching a basketball shooting animation in 3D with audio, and also performing motor imagery afterwards. Electroencephalogram (EEG) was recorded for analysis of motor-related features of the brain such as power spectral analysis in the α and β frequency bands and spectral entropy. 18 EEG channels over the motor cortex were used for all stroke patients. Results: All results are normalised relative to all tasks for each participant. The power spectral densities peak near the α band for all participants and also the β band for some participants. Tasks with instructions during motor imagery generally show greater power spectral peaks. The p-values of the Wilcoxon signed-rank test for band power comparison from the 18 EEG channels between different pairs of tasks show a 0.01 significance of rejecting the band powers being the same for most tasks done by stroke subjects. The motor cortex of most stroke patients is more active when virtual reality is involved during motor imagery as indicated by their respective scalp maps of band power and spectral entropy. Conclusion: The resulting activation of stroke patient’s motor cortices in this study reveals evidence that it is induced by imagination of movement and virtual reality supports motor imagery. The framework of the current study also provides an efficient way to investigate motor imagery and virtual reality during post-stroke rehabilitation.

2. Tracking Functional Network Connectivity Dynamics in the Elderly

   Kaichao Wu, Beth Jelfs, Seedahmed S. Mahmoud, Katrina Neville, and John Q. Fang

   Frontiers in Neuroscience, 2023, vol. 17, no. 1146264. pp. 1–11.

   abstract doi pdf supp code

   Introduction: Functional magnetic resonance imaging (fMRI) has shown that aging disturbs healthy brain organization and functional connectivity. However, how this age-induced alteration impacts dynamic brain function interaction has not yet been fully investigated. Dynamic function network connectivity (DFNC) analysis can produce a brain representation based on the time-varying network connectivity changes, which can be further used to study the brain aging mechanism for people at different age stages. Method: This presented investigation examined the dynamic functional connectivity representation and its relationship with brain age for people at an elderly stage as well as in early adulthood. Specifically, the resting-state fMRI data from the University of North Carolina cohort of 34 young adults and 28 elderly participants were fed into a DFNC analysis pipeline. This DFNC pipeline forms an integrated dynamic functional connectivity (FC) analysis framework, which consists of brain functional network parcellation, dynamic FC feature extraction, and FC dynamics examination. Results: The statistical analysis demonstrates that extensive dynamic connection changes in the elderly concerning the transient brain state and the method of functional interaction in the brain. In addition, various machine learning algorithms have been developed to verify the ability of dynamic FC features to distinguish the age stage. The fraction time of DFNC states has the highest performance, which can achieve a classification accuracy of over 88% by a decision tree. Discussion: The results proved there are dynamic FC alterations in the elderly, and the alteration was found to be correlated with mnemonic discrimination ability and could have an impact on the balance of functional integration and segregation.

Conference Articles

1. Motor Imagery Observed by fNIRS

   Chi Sang Choy, Zixin Ye, Ziyang Huang, Qifeng Zheng, Qiang Fang, Seedahmed S. Mahmoud, Katrina Neville, and Beth Jelfs

   In Proc. IEEE International Conference on Body Sensor Networks, 2023,

   abstract doi pdf

   Motor imagery (MI) is expected to activate brain areas related to motor functions, yet, the brain is constantly active to some extent leading to difficulties in differentiating MI from supposedly non-motor tasks. In this study, we use functional near-infrared spectroscopy (fNIRS) to offer an objective and spatially precise measurement of brain activity during MI. The fNIRS findings of this study using our MI-based exercise framework with 15 healthy subjects indicate that casual imagination and especially relaxation do not induce an intense motor brain activation compared to active MI performance. Furthermore, dynamic visual cues for MI appear to enhance brain activation around the motor brain areas of the majority of subjects and MI training helps some subjects to activate their motor cortex. Future research may refer to our framework to validate the competence of stroke patients in MI-based motor rehabilitation.

2. Brain Functional Connectivity Networks do not Return to Resting-state During Control Trials in Block Design Experiments

   J. Esmaelpoor, T. Peng, B. Jelfs, M. J. Shader, C. M. McKay, and D. Mao

   In Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2023,

   abstract doi pdf

   Many studies on morphology analysis show that if short inter-stimulus intervals separate tasks, the hemodynamic response amplitude will return to the resting-state baseline before the subsequent stimulation onset; hence, responses to successive tasks do not overlap. Accordingly, popular brain imaging analysis techniques assume changes in hemodynamic response amplitude subside after a short time (around 15 seconds). However, whether this assumption holds when studying brain functional connectivity has yet to be investigated. This paper assesses whether or not the functional connectivity network in control trials returns to the resting-state functional connectivity network. Traditionally, control trials in block-design experiments are used to evaluate response morphology to no stimulus. We analyzed data from an event-related experiment with audio and visual stimuli and resting state. Our results showed that functional connectivity networks during control trials were more similar to that of tasks than resting-state networks. In other words, contrary to task-related changes in the hemodynamic amplitude, where responses settle after a short time, the brain’s functional connectivity networks do not return to their intrinsic resting-state network in such short intervals.

3. Evaluation of Module Dynamics in Functional Brain Networks After Stroke

   K. Wu, Q. Fang, K. Neville, and B. Jelfs

   In Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2023,

   abstract doi pdf

   The brain’s functional network can be analyzed as a set of distributed functional modules. Previous studies using the static method suggested the modularity of the brain function network decreased due to stroke; however, how the modular network changes after stroke, particularly over time, is far from understood. This study collected resting-state functional MRI data from 15 stroke patients and 15 age-matched healthy controls. The patients exhibit distinct clinical symptoms, presenting in mild (n = 6) and severe (n = 9) subgroups. By using a multilayer network model, a dynamic modular structure was detected and corresponding interaction measurements were calculated. The results demonstrated that the module structure and interaction had changed following the stroke. Importantly, the significant differences in dynamic interaction measures demonstrated that the module interaction alterations were not independent of the initial degree of clinical severity. Mild patients were observed to have a significantly lower between-module interaction than severe patients as well as healthy controls. In contrast, severe patients showed remarkably lower within-module interaction and had a reduced overall interaction compared to healthy controls. These findings contributed to the development of post-stroke dynamics analysis and shed new light on brain network interaction for stroke patients.Clinical relevance— Dynamic module interaction analysis underpins the post-stroke functional plasticity and reorganization, and may enable new insight into rehabilitation strategies to promote recovery of function.

Conference Articles

1. Landmark Management in the Application of Radar SLAM

   Shuai Sun, Beth Jelfs, Kamran Ghorbani, Glenn Matthews, and Christopher Gilliam

   In Proc. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, 2022, pp. 903–910.

   abstract doi arXiv pdf slides code

   This paper focuses on efficient landmark management in radar based simultaneous localization and mapping (SLAM). Landmark management is necessary in order to maintain a consistent map of the estimated landmarks relative to the estimate of the platform’s pose. This task is particularly important when faced with multiple detections from the same landmark and/or dynamic environments where the location of a landmark can change. A further challenge with radar data is the presence of false detections. Accordingly, we propose a simple yet efficient rule based solution for radar SLAM landmark management. Assuming a low-dynamic environment, there are several steps in our solution: new landmarks need to be detected and included, false landmarks need to be identified and removed, and the consistency of the landmarks registered in the map needs to be maintained. To illustrate our solution, we run an extended Kalman filter SLAM algorithm in an environment containing both stationary and temporally stationary landmarks. Our simulation results demonstrate that the proposed solution is capable of reliably managing landmarks even when faced with false detections and multiple detections from the same landmark.

Journal Articles

1. An Adaptive All-Pass Filter for Time-Varying Delay Estimation

   B. Jelfs, S. Sun, K. Ghorbani, and C. Gilliam

   IEEE Signal Processing Letters, 2021, vol. 28, no. 1146264. pp. 628–632.

   abstract doi arXiv pdf code

   The focus of this paper is the estimation of a delay between two signals. Such a problem is common in signal processing and particularly challenging when the delay is non-stationary in nature. Our proposed solution is based on an all-pass filter framework comprising of two elements: a time delay is equivalent to all-pass filtering and an all-pass filter can be represented in terms of a ratio of a finite impulse response (FIR) filter and its time reversal. Using these elements, we propose an adaptive filtering algorithm with an LMS style update that estimates the FIR filter coefficients and the time delay. Specifically, at each time step, the algorithm updates the filter coefficients based on a gradient descent update and then extracts an estimate of the time delay from the filter. We validate our algorithm on synthetic data demonstrating that it is both accurate and capable of tracking time-varying delays.

2. Cooperative Localization Using Distance Measurements for Mobile Nodes

   Wenchao Li, Beth Jelfs, Allison Kealy, Xuezhi Wang, and Bill Moran

   Sensors, 2021, vol. 21, no. 4:1507. pp. 628–632.

   abstract doi pdf

   This paper considers the two-dimensional (2D) anchorless localization problem for sensor networks in global positioning system (GPS)-denied environments. We present an efficient method, based on the multidimensional scaling (MDS) algorithm, in order to estimate the positions of the nodes in the network using measurements of the inter-node distances. The proposed method takes advantage of the mobility of the nodes to address the location ambiguity problem, i.e., rotation and flip ambiguity, which arises in the anchorless MDS algorithm. Knowledge of the displacement of the moving node is used to produce an analytical solution for the noise-free case. Subsequently, a least squares estimator is presented for the noisy scenario and the associated closed-form solution derived. The simulations show that the proposed algorithm accurately and efficiently estimates the locations of nodes, outperforming alternative methods.

3. An Efficient and Flexible Spike Train Model via Empirical Bayes

   Q. She, Xiaoli Wu, B. Jelfs, A. S. Charles, and R. H. M. Chan

   IEEE Transactions on Signal Processing, 2021, vol. 69, no. 4:1507. pp. 3236–3251.

   abstract doi arXiv pdf code

   Accurate statistical models of neural spike responses can characterize the information carried by neural populations. But the limited samples of spike counts during recording usually result in model overfitting. Besides, current models assume spike counts to be Poisson-distributed, which ignores the fact that many neurons demonstrate over-dispersed spiking behaviour. Although the Negative Binomial Generalized Linear Model (NB-GLM) provides a powerful tool for modeling over-dispersed spike counts, the maximum likelihood-based standard NB-GLM leads to highly variable and inaccurate parameter estimates. Thus, we propose a hierarchical parametric empirical Bayes method to estimate the neural spike responses among neuronal population. Our method integrates both Generalized Linear Models (GLMs) and empirical Bayes theory, which aims to (1) improve the accuracy and reliability of parameter estimation, compared to the maximum likelihood-based method for NB-GLM and Poisson-GLM; (2) effectively capture the over-dispersion nature of spike counts from both simulated data and experimental data; and (3) provide insight into both neural interactions and spiking behaviours of the neuronal populations. We apply our approach to study both simulated data and experimental neural data. The estimation of simulation data indicates that the new framework can accurately predict mean spike counts simulated from different models and recover the connectivity weights among neural populations. The estimation based on retinal neurons demonstrate the proposed method outperforms both NB-GLM and Poisson-GLM in terms of the predictive log-likelihood of held-out data.

4. Toward Autonomous UAV Localization via Aerial Image Registration

   Xuezhi Wang, Allison Kealy, Wenchao Li, Beth Jelfs, Christopher Gilliam, Samantha Le May, and Bill Moran

   Electronics, 2021, vol. 10, no. 4:435. pp. 3236–3251.

   abstract doi pdf

   Absolute localization of a flying UAV on its own in a global-navigation-satellite-system (GNSS)-denied environment is always a challenge. In this paper, we present a landmark-based approach where a UAV is automatically locked into the landmark scene shown in a georeferenced image via a feedback control loop, which is driven by the output of an aerial image registration. To pursue a real-time application, we design and implement a speeded-up-robust-features (SURF)-based image registration algorithm that focuses efficiency and robustness under a 2D geometric transformation. A linear UAV controller with signals of four degrees of freedom is derived from the estimated transformation matrix. The approach is validated in a virtual simulation environment, with experimental results demonstrating the effectiveness and robustness of the proposed UAV self-localization system.

5. Weakly-Supervised Lesion Analysis With a CNN-Based Framework for COVID-19

   Kaichao Wu, Beth Jelfs, Xiangyuan Ma, Ruitian Ke, Xuerui Tan, and Qiang Fang

   Physics in Medicine & Biology, 2021, vol. 10, no. 4:435. pp. 3236–3251.

   abstract doi pdf code

   Objective. Lesions of COVID-19 can be clearly visualized using chest CT images, and hence provide valuable evidence for clinicians when making a diagnosis. However, due to the variety of COVID-19 lesions and the complexity of the manual delineation procedure, automatic analysis of lesions with unknown and diverse types from a CT image remains a challenging task. In this paper we propose a weakly-supervised framework for this task requiring only a series of normal and abnormal CT images without the need for annotations of the specific locations and types of lesions. Approach. A deep learning-based diagnosis branch is employed for classification of the CT image and then a lesion identification branch is leveraged to capture multiple types of lesions. Main Results. Our framework is verified on publicly available datasets and CT data collected from 13 patients of the First Affiliated Hospital of Shantou University Medical College, China. The results show that the proposed framework can achieve state-of-the-art diagnosis prediction, and the extracted lesion features are capable of distinguishing between lesions showing ground glass opacity and consolidation. Significance. The proposed approach integrates COVID-19 positive diagnosis and lesion analysis into a unified framework without extra pixel-wise supervision. Further exploration also demonstrates that this framework has the potential to discover lesion types that have not been reported and can potentially be generalized to lesion detection of other chest-based diseases.

Journal Articles

1. Complexity Measures of Voice Recordings as a Discriminative Tool for Parkinson’s Disease

   R. Viswanathan, S. P. Arjunan, A. Bingham, B. Jelfs, P. Kempster, S. Raghav, and D. K. Kumar

   Biosensors, 2020, vol. 10, no. 1:1. pp. 3236–3251.

   abstract doi pdf

   In this paper, we have investigated the differences in the voices of Parkinson’s disease (PD) and age-matched control (CO) subjects when uttering three phonemes using two complexity measures: fractal dimension (FD) and normalised mutual information (NMI). Three sustained phonetic voice recordings, /a/, /u/ and /m/, from 22 CO (mean age = 66.91) and 24 PD (mean age = 71.83) participants were analysed. FD was first computed for PD and CO voice recordings, followed by the computation of NMI between the test groups: PD-CO, PD-PD and CO-CO. Four features reported in the literature–normalised pitch period entropy (Norm. PPE), glottal-to-noise excitation ratio (GNE), detrended fluctuation analysis (DFA) and glottal closing quotient (ClQ)–were also computed for comparison with the proposed complexity measures. The statistical significance of the features was tested using a one-way ANOVA test. Support vector machine (SVM) with a linear kernel was used to classify the test groups, using a leave-one-out validation method. The results showed that PD voice recordings had lower FD compared to CO (p < 0.008). It was also observed that the average NMI between CO voice recordings was significantly lower compared with the CO-PD and PD-PD groups (p < 0.036) for the three phonetic sounds. The average NMI and FD demonstrated higher accuracy (>80%) in differentiating the test groups compared with other speech feature-based classifications. This study has demonstrated that the voices of PD patients has reduced FD, and NMI between voice recordings of PD-CO and PD-PD is higher compared with CO-CO. This suggests that the use of NMI obtained from the sample voice, when paired with known groups of CO and PD, can be used to identify PD voices. These findings could have applications for population screening.

Conference Articles

1. Application of Image Processing and Circular Statistics to 3D Cellular Alignment

   B. Jelfs, and C. Gilliam

   In Proc. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, 2020, pp. 992–1000.

   abstract html pdf code

   Alignment and orientation of cells play an important part in the function of biological tissue. Recent developments in bioengineering using 3D scaffolds have created an increased need for computational techniques to measure orientation which extend beyond 2D measures to produce 3D measures of orientation. Initial studies of 3D alignment have focused on determining individual orientations, however, to truly understand the impact these structures have on the cellular alignment we need to understand the overall distribution of the orientations and their statistics. Hence, in this paper we develop an approach for determining 3D cellular alignment based on image gradients and directional statistics. The intensity gradients of the volumetric image are used to construct a 3D vector field and the local dominant orientations of this vector field then determined.

Journal Articles

1. Prosthetic Hand Control: A Multidisciplinary Review to Identify Strengths, Shortcomings, and the Future

   D. K. Kumar, B. Jelfs, X. Sui, and S. P. Arjunan

   Biomedical Signal Processing and Control, 2019, vol. 53, no. 101588. pp. 3236–3251.

   abstract doi pdf

   Prosthetic hand control has fired the imagination of many researchers and thousands of papers have been published in this field, but the user acceptance has not been strong and there appears to be a significant gap between the published research and its translation. One observation of the literature is that while this requires multidisciplinary research, most articles appear to be topic focused, with lack of literature that connect across the different disciplines. This paper reports a multidisciplinary, candid review which has evaluated literature of four major associated topics: (i) User requirements, (ii) Signal recording, (iii) Signal analysis and (iv) User feedback, with the aim to identify the potential directions for research that will improve the translation of this technology. Special effort was made to collate diverse views and authors. This review has found that more research for the analysis and evaluation of the user requirements is necessary to ensure that the amputees use these devices extensively. Further research is also required into the development of both, the paradigm and the technology to give feedback to the user from the prosthetic hand device. There is also the need to improve the electrodes and recording techniques to ensure uninterrupted user-control over extended periods of time. One important outcome of this paper is that it has uncovered the differences of performance measures used by different authors because of which it is difficult to compare the results reported in their papers.

2. Which Gait Parameters and Walking Patterns Show the Significant Differences Between Parkinson’s Disease and Healthy Participants?

   S. M. Keloth, R. Viswanathan, B. Jelfs, S. Arjunan, S. Raghav, and D. Kumar

   Biosensors, 2019, vol. 9, no. 2:59. pp. 3236–3251.

   abstract doi pdf

   This study investigated the difference in the gait of patients with Parkinson’s disease (PD), age-matched controls and young controls during three walking patterns. Experiments were conducted with 24 PD, 24 age-matched controls and 24 young controls, and four gait intervals were measured using inertial measurement units (IMU). Group differences between the mean and variance of the gait parameters (stride interval, stance interval, swing interval and double support interval) for the three groups were calculated and statistical significance was tested. The results showed that the variance in each of the four gait parameters of PD patients was significantly higher compared with the controls, irrespective of the three walking patterns. This study showed that the variance of any of the gait interval parameters obtained using IMU during any of the walking patterns could be used to differentiate between the gait of PD and control people.

Conference Articles

1. Fast & Efficient Delay Estimation Using Local All-Pass & Kalman Filters

   B. Jelfs, and C. Gilliam

   In Proc. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, 2019, pp. 1533–1539.

   abstract doi pdf slides code

   Delay estimation is a common problem in signal processing which becomes particularly challenging when the delay is time-varying and the recorded signals are non-stationary. While methods for time-varying delay (TVD) estimation exist many of these are based on maximum likelihood estimation and thus are not well suited to real-time implementation. In this paper we present a method for TVD estimation which is suitable for real-time non-stationary applications. The proposed method combines local all-pass (LAP) filters with a Kalman filter. By using measurement fusion to combine the outputs of several LAP filters in the Kalman filter we can accurately track TVDs whilst allowing for fast and efficient parallel computation. Illustrative simulations demonstrate the effectiveness of the proposed approach.

2. Normalized Mutual Information of Phonetic Sound to Distinguish the Speech of Parkinson’s Disease

   R. Viswanathan, A. Bingham, S. Raghav, S. P. Arjunan, B. Jelfs, P. Kempster, and D. K. Kumar

   In Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2019, pp. 3525–3526.

   abstract doi pdf

   This study has investigated the use of inter-personnel mutual information computed from the phonetic sound recordings to differentiate between Parkinson’s disease (PD) and control subjects. The normalized mutual information (NMI) denotes the amount of information shared between the voice recordings of people within the same group: PD and Control. The hypothesis of this study was that within group NMI will be significantly different when compared with inter- group NMI. For each phonetic sound, the NMI was computed for every pairing of recordings for both the PD and control groups. Pearson correlation coefficient analysis was used to determine the association of NMI with clinical parameters including Unified Parkinson’s Disease Rating Scale (UPDRS), Montreal cognitive assessment (MoCA) and disease duration. ANOVA test for the three phonetic sounds of control and PD subjects showed that there is significant difference between the intra-group mean NMI for the two groups (p < 0.003) and also showed significant association with the UPDRS motor examination score, MoCA and disease duration.

Conference Articles

1. Identifying Noisy Electrodes in High Density Surface Electromyography Recordings Through Analysis of Spatial Similarities

   A. Bingham, B. Jelfs, S. P. Arjunan, and D. K. Kumar

   In Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2018, pp. 2325–2328.

   abstract doi pdf

   In this study we developed a technique for identifying noisy electrodes in high density surface electromyography (HD-sEMG). The technique finds the spatial similarity of each electrode in the electrode array by counting the number of interactions the electrode has. Using this information the technique identifies noisy electrodes by finding electrodes that are significantly dissimilar to the other electrodes. The HD-sEMG recordings used in this study were taken from three participants who performed two isometric contractions of their biceps at 40% and 80% of their maximum voluntary contraction (MVC) force. White Gaussian noisy was added to a varying number of recorded signals before being digital filtering to generate a variety of recordings to test the technique with. In the recordings, groups of 2, 4, 8, and 16 electrodes had noise added such that the signal to noise ratios (SNR) were 0, 5, 10, 15, and 20dB. The results show that the technique can reliably identify groups of 2, 4, and 8 noisy electrodes with SNRs of 0, 5, and 10dB.

2. Estimating Muscle Fibre Conduction Velocity in the Presence of Array Misalignment

   C. Gilliam, and B. Jelfs

   In Proc. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, 2018, pp. 853–860.

   abstract doi pdf slides code

   Surface electromyography (sEMG) has the potential to provide valuable information regarding the status and health of a muscle. In particular, recent developments in high density sEMG (HD-sEMG), which allow simultaneous recordings from a greater number of electrodes, enable the calculation of muscle attributes such as the conduction velocity of motor unit action potentials. However, as with standard recording montages, HD-sEMG requires careful placement of the electrodes to align with the direction of the muscle fibres, thus limiting practical applications. In this paper we demonstrate an algorithm for calculating muscle fibre conduction velocity which is independent of the alignment of the array. The algorithm automatically corrects for the misalignment of the array whilst estimating the conduction velocity using common local all-pass (CLAP) filters. Specifically, the misalignment is modelled as a rotation of the array relative to the fibre and this rotation is estimated by iteratively fitting the model to the output of the CLAP filters. We validate the proposed algorithm on simulated HD-sEMG data generated from a realistic biological model, demonstrating that the algorithm obtains an accurate estimate of the conduction velocity even when the array is misaligned.

3. Time-Varying Delay Estimation Using Common Local All-Pass Filters with Application to Surface Electromyography

   C. Gilliam, A. Bingham, T. Blu, and B. Jelfs

   In Proc. IEEE International Conference on Acoustics, Speech and Signal Processing, 2018, pp. 841–845.

   abstract doi pdf slides code

   Estimation of conduction velocity (CV) is an important task in the analysis of surface electromyography (sEMG). The problem can be framed as estimation of a time-varying delay (TVD) between electrode recordings. In this paper we present an algorithm which incorporates information from multiple electrodes into a single TVD estimation. The algorithm uses a common all-pass filter to relate two groups of signals at a local level. We also address a current limitation of CV estimators by providing an automated way of identifying the innervation zone from a set of electrode recordings, thus allowing incorporation of the entire array into the estimation. We validate the algorithm on both synthetic and real sEMG data with results showing the proposed algorithm is both robust and accurate.

Journal Articles

1. Normalised Mutual Information of High-Density Surface Electromyography during Muscle Fatigue

   A. Bingham, S. P. Arjunan, B. Jelfs, and D. K. Kumar

   Entropy, 2017, vol. 19, no. 12, pp. 697.

   abstract doi pdf

   This study has developed a technique for identifying the presence of muscle fatigue based on the spatial changes of the normalised mutual information (NMI) between multiple high density surface electromyography (HD-sEMG) channels. Muscle fatigue in the tibialis anterior (TA) during isometric contractions at 40% and 80% maximum voluntary contraction levels was investigated in ten healthy participants (Age range: 21 to 35 years; Mean age = 26 years; Male = 4, Female = 6). HD-sEMG was used to record 64 channels of sEMG using a 16 by 4 electrode array placed over the TA. The NMI of each electrode with every other electrode was calculated to form an NMI distribution for each electrode. The total NMI for each electrode (the summation of the electrode’s NMI distribution) highlighted regions of high dependence in the electrode array and was observed to increase as the muscle fatigued. To summarise this increase, a function, M(k), was defined and was found to be significantly affected by fatigue and not by contraction force. The technique discussed in this study has overcome issues regarding electrode placement and was used to investigate how the dependences between sEMG signals within the same muscle change spatially during fatigue.

2. Directionality Indices: Testing Information Transfer with Surrogate Correction

   B. Jelfs, and R. H. M. Chan

   Phys. Rev. E, 2017, vol. 96, no. 5:052220. pp. 697.

   abstract doi pdf

   Directionality indices can be used as an indicator of the asymmetry in coupling between systems and have found particular application in relation to neurological systems. The directionality index between two systems is a function of measures of information transfer in both directions. Here we illustrate that before inferring the directionality of coupling it is first necessary to consider the use of appropriate tests of significance. We propose a surrogate corrected directionality index which incorporates such testing. We also highlight the differences between testing the significance of the directionality index itself versus testing the individual measures of information transfer in each direction. To validate the approach we compared two different methods of estimating coupling, both of which have previously been used to estimate directionality indices. These were the modeling-based evolution map approach and a conditional mutual information (CMI) method for calculating dynamic information rates. For the CMI-based approach we also compared two different methods for estimating the CMI, an equiquantization-based estimator and a k-nearest neighbors estimator.

3. Self-Recalibrating Surface EMG Pattern Recognition for Neuroprosthesis Control Based on Convolutional Neural Network

   X. Zhai, B. Jelfs, R. H. M. Chan, and C. Tin

   Frontiers in Neuroscience, 2017, vol. 11, no. 379. pp. 697.

   abstract doi pdf supp

   Hand movement classification based on surface electromyography (sEMG) pattern recognition is a promising approach for upper limb neuroprosthetic control. However, maintaining day-to-day performance is challenged by the non-stationary nature of sEMG in real-life operation. In this study, we propose a self-recalibrating classifier that can be automatically updated to maintain a stable performance over time without the need for user retraining. Our classifier is based on convolutional neural network (CNN) using short latency dimension-reduced sEMG spectrograms as inputs. The pretrained classifier is recalibrated routinely using a corrected version of the prediction results from recent testing sessions. Our proposed system was evaluated with the NinaPro database comprising of hand movement data of 40 intact and 11 amputee subjects. Our system was able to achieve  10.18% (intact, 50 movement types) and  2.99% (amputee, 10 movement types) increase in classification accuracy averaged over five testing sessions with respect to the unrecalibrated classifier. When compared with a support vector machine (SVM) classifier, our CNN-based system consistently showed higher absolute performance and larger improvement as well as more efficient training. These results suggest that the proposed system can be a useful tool to facilitate long-term adoption of prosthetics for amputees in real-life applications.

Conference Articles

1. Outlier Removal in Facial Surface Electromyography Through Hampel Filtering Technique

   S. Bhowmik, B. Jelfs, S. P. Arjunan, and D. K. Kumar

   In Proc. IEEE Life Sciences Conference, 2017, pp. 258–261.

   abstract doi pdf

   The aim of the present investigation was to filter outliers in facial surface electromyography (fSEMG) originating from eye blinks, through a decision based filtering technique. Since, these outliers lie within the frequency range of electromyographic activity (30-300 Hz), conventional filtering methods fail to remove them. Hence, an application of an outlier filtering technique, Hampel filtering, has been introduced which is proficient at removing high frequency impulsive spikes (100-150 Hz) from facial sEMG. The Hampel filter removes the outliers without distorting the original data sequence and improves the quality of the signal as observed in time-frequency analysis.

2. Entropy of Surface EMG Reflects Object Weight in Grasp-and-Lift Task

   Y. Li, B. Jelfs, and R. H.M. Chan

   In Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2017, pp. 2530–2533.

   abstract doi pdf

   Fingertip force coordination is crucial to the success of grasp-and-lift tasks. In the development of motor prosthesis for daily applications, the ability to accurately classify the desired grasp-and-lift from multi-channel surface electromyography (sEMG) is essential. In order to extract reliable indicators for fingertip force coordination, we searched an extensive set of sEMG features for the optimal subset of relevant features. Using mutual information based feature selection we found that a subset of not more than 10 sEMG features selected from over seven thousand, could effectively classify object weights in grasp-and-lift tasks. Average classification accuracies of 82.53% in the acceleration phase and 88.61% in the isometric contraction phase were achieved. Furthermore, sEMG features associated with object weights and common across individuals were identified. These time-domain features (entropy, mean/median absolute deviation, pNNx) can be calculated efficiently, providing possible new indicators.

Journal Articles

1. Vagus Nerve Stimulation Alters Phase Synchrony of the Anterior Cingulate Cortex and Facilitates Decision Making in Rats

   B. Cao, J. Wang, M. Shahed, B. Jelfs, R. H. M. Chan, and Y. Li

   Scientific Reports, 2016, vol. 6, no. 35135. pp. 697.

   abstract doi pdf supp code

   Vagus nerve stimulation (VNS) can enhance memory and cognitive functions in both rats and humans. Studies have shown that VNS influenced decision-making in epileptic patients. However, the sites of action involved in the cognitive-enhancement are poorly understood. By employing a conscious rat model equipped with vagus nerve cuff electrode, we assess the role of chronic VNS on decision-making in rat gambling task (RGT). Simultaneous multichannel-recordings offer an ideal setup to test the hypothesis that VNS may induce alterations of in both spike-field-coherence and synchronization of theta oscillations across brain areas in the anterior cingulate cortex (ACC) and basolateral amygdala (BLA). Daily VNS, administered immediately following training sessions of RGT, caused an increase in ‘good decision-maker’ rats. Neural spikes in the ACC became synchronized with the ongoing theta oscillations of local field potential (LFP) in BLA following VNS. Moreover, cross-correlation analysis revealed synchronization between the ACC and BLA. Our results provide specific evidence that VNS facilitates decision-making and unveils several important roles for VNS in regulating LFP and spike phases, as well as enhancing spike-phase coherence between key brain areas involved in cognitive performance. These data may serve to provide fundamental notions regarding neurophysiological biomarkers for therapeutic VNS in cognitive impairment.

2. Impairment of Decision Making and Disruption of Synchrony Between Basolateral Amygdala and Anterior Cingulate Cortex in the Maternally Separated Rat

   B. Cao, J. Wang, X. Zhang, X. Yang, D. C.-H. Poon, B. Jelfs, R. H. M. Chan, J. C.-Y. Wu, and Y. Li

   Neurobiology of Learning and Memory, 2016, vol. 136, no. 35135. pp. 74–85.

   abstract doi pdf code

   There is considerable evidence to suggest early life experiences, such as maternal separation (MS), play a role in the prevalence of emotional dysregulation and cognitive impairment. At the same time, optimal decision making requires functional integrity between the amygdala and anterior cingulate cortex (ACC), and any dysfunction of this system is believed to induce decision-making deficits. However, the impact of MS on decision-making behavior and the underlying neurophysiological mechanisms have not been thoroughly studied. As such, we consider the impact of MS on the emotional and cognitive functions of rats by employing the open-field test, elevated plus-maze test, and rat gambling task (RGT). Using multi-channel recordings from freely behaving rats, we assessed the effects of MS on the large scale synchrony between the basolateral amygdala (BLA) and the ACC; while also characterizing the relationship between neural spiking activity and the ongoing oscillations in theta frequency band across the BLA and ACC. The results indicated that the MS rats demonstrated anxiety-like behavior. While the RGT showed a decrease in the percentage of good decision-makers, and an increase in the percentage of poor decision-makers. Electrophysiological data revealed an increase in the total power in the theta band of the LFP in the BLA and a decrease in theta power in the ACC in MS rats. MS was also found to disrupt the spike-field coherence of the ACC single unit spiking activity to the ongoing theta oscillations in the BLA and interrupt the synchrony in the BLA-ACC pathway. We provide specific evidence that MS leads to decision-making deficits that are accompanied by alteration of the theta band LFP in the BLA-ACC circuitries and disruption of the neural network integrity. These observations may help revise fundamental notions regarding neurophysiological biomarkers to treat cognitive impairment induced by early life stress.

3. Computational Classification of Different Wild-Type Zebrafish Strains Based on Their Variation in Light-Induced Locomotor Response

   Y. Gao, G. Zhang, B. Jelfs, R. Carmer, P. Venkatraman, M. Ghadami, S. A. Brown, C. P. Pang, Y. F. Leung, R. H. M. Chan, and M. Zhang

   Computers in Biology and Medicine, 2016, vol. 69, no. 35135. pp. 1–9.

   abstract doi pdf supp

   Zebrafish larvae display a rapid and characteristic swimming behaviour after abrupt light onset or offset. This light-induced locomotor response (LLR) has been widely used for behavioural research and drug screening. However, the locomotor responses have long been shown to be different between different wild-type (WT) strains. Thus, it is critical to define the differences in the WT LLR to facilitate accurate interpretation of behavioural data. In this investigation, we used support vector machine (SVM) models to classify LLR data collected from three WT strains: AB, TL and TLAB (a hybrid of AB and TL), during early embryogenesis, from 3 to 9 days post-fertilisation (dpf). We analysed both the complete dataset and a subset of the data during the first 30 after light change. This initial period of activity is substantially driven by vision, and is also known as the visual motor response (VMR). The analyses have resulted in three major conclusions: First, the LLR is different between the three WT strains, and at different developmental stages. Second, the distinguishable information in the VMR is comparable to, if not better than, the full dataset for classification purposes. Third, the distinguishable information of WT strains in the light-onset response differs from that in the light-offset response. While the classification accuracies were higher for the light-offset than light-onset response when using the complete LLR dataset, a reverse trend was observed when using a shorter VMR dataset. Together, our results indicate that one should use caution when extrapolating interpretations of LLR/VMR obtained from one WT strain to another.

Conference Articles

1. Fuzzy Entropy Based Nonnegative Matrix Factorization for Muscle Synergy Extraction

   B. Jelfs, L. Li, C. Tin, and R. H. M. Chan

   In Proc. IEEE International Conference on Acoustics, Speech and Signal Processing, 2016, pp. 739–743.

   abstract doi pdf poster

   The concept of muscle synergies has proven to be an effective method for representing patterns of muscle activation. The number of degrees of freedom to be controlled are reduced while also providing a flexible platform for producing detailed movements using synergies as building blocks. It has previously been shown that small components of movement are crucial to producing precise and coordinated movement. Methods which focus on the variance of the data make it possible to overlook these small components in the synergy extraction process. However, algorithms which address the inherent complexity in the neuromuscular system are lacking. To that end we propose a new nonnegative matrix factorization algorithm which employs a cross fuzzy entropy similarity measure, thus, extracting muscle synergies which preserve the complexity of the recorded muscular data. The performance of the proposed algorithm is illustrated on representative EMG data.

2. Cross-frequency Information Transfer From EEG to EMG in Grasping

   W. K. Y. So, L. Yang, B. Jelfs, Q. She, S. W. H. Wong, J. N. Mak, and R. H. M. Chan

   In Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2016, pp. 4531–4534.

   abstract doi pdf

   This paper presents an investigation into the cortico-muscular relationship during a grasping task by evaluating the information transfer between EEG and EMG signals. Information transfer was computed via a non-linear model-free measure, transfer entropy (TE). To examine the cross-frequency interaction, TEs were computed after the times series were decomposed into various frequency ranges via wavelet transform. Our results demonstrate the capability of TE to capture the direct interaction between EEG and EMG. In addition, the cross-frequency analysis revealed instantaneous decrease in information transfer from EEG to the high frequency component of EMG (100-200Hz) during the onset of movement.

3. Short Latency Hand Movement Classification Based on Surface EMG Spectrogram with PCA

   X. Zhai, B. Jelfs, R. H. M. Chan, and C. Tin

   In Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2016, pp. 327–330.

   abstract doi pdf

   Hand gesture recognition from forearm surface electromyography (sEMG) is an active research field in the development of motor prosthesis. Studies have shown that classification accuracy and efficiency is highly dependent on the features extracted from the EMG. In this paper, we show that EMG spectrograms are a particularly effective feature for discriminating multiple classes of hand gesture when subjected to principal component analysis for dimensionality reduction. We tested our method on the Ninapro database which includes sEMG data (12 channels) of 40 subjects performing 50 different hand movements. Our results demonstrate improved classification accuracy (by  10%) over purely time domain features for 50 different hand movements, including small finger movements and different levels of force exertion. Our method has also reduced the error rate (by  12%) at the transition phase of gestures which could improve robustness of gesture recognition when continuous classification from sEMG is required.

Journal Articles

1. Impairment of Cognitive Function by Chemotherapy: Association with the Disruption of Phase-Locking and Synchronization in Anterior Cingulate Cortex

   L. Mu, J. Wang, B. Cao, B. Jelfs, R. H. M. Chan, X. Xu, M. Hasan, X. Zhang, and Y. Li

   Molecular Brain, 2015, vol. 8, no. 32. pp. 1–9.

   abstract doi pdf code

   Background Patients following prolonged cancer chemotherapy are at high risk of emotional and cognitive deficits. Research indicates that the brain neuronal temporal coding and synaptic long-term potentiation (LTP) are critical in memory and perception. We studied the effects of cisplatin on induction of LTP in the basolateral amygdala (BLA)-anterior cingulate cortex (ACC) pathway, characterized the coordination of spike timing with local theta oscillation, and identified synchrony in the BLA-ACC network integrity. Results In the study presented, the impacts of cisplatin on emotional and cognitive functions were investigated by elevated plus-maze test, Morris water maze test, and rat Iowa gambling task (RGT). Electrophysiological recordings were conducted to study long-term potentiation. Simultaneous recordings from multi-electrodes were performed to characterize the neural spike firing and ongoing theta oscillation of local field potential (LFP), and to clarify the synchronization of large scale of theta oscillation in the BLA-ACC pathway. Cisplatin-treated rats demonstrated anxiety- like behavior, exhibited impaired spatial reference memory. RGT showed decrease of the percentage of good decision-makers, and increase in the percentage of maladaptive behavior (delay-good decision-makers plus poor decision-makers). Cisplatin suppressed the LTP, and disrupted the phase-locking of ACC single neural firings to the ongoing theta oscillation; further, cisplatin interrupted the synchrony in the BLA-ACC pathway. Conclusions We provide the first direct evidence that the cisplatin interrupts theta-frequency phase-locking of ACC neurons. The block of LTP and disruption of synchronized theta oscillations in the BLA-ACC pathway are associated with emotional and cognitive deficits in rats, following cancer chemotherapy.

2. Theta-Frequency Phase-Locking of Single Anterior Cingulate Cortex Neurons and Synchronization with the Medial Thalamus are Modulated by Visceral Noxious Stimulation in Rats

   J. Wang, B. Cao, T. R. Yu, B. Jelfs, J. Yan, R. H. M. Chan, and Y. Li

   Neuroscience, 2015, vol. 298, no. 32. pp. 200–210.

   abstract doi pdf code

   The rodent anterior cingulate cortex (ACC) is critical for visceral pain and pain-related aversive response in chronic visceral hypersensitive (VH) state. Long-term potentiation (LTP), induced by theta burst stimulation (TBS) in the medial thalamus (MT)-ACC pathway, is blocked in VH rats. However, the neuronal intrinsic firing characteristics and the MT-ACC connectivity have not been investigated in visceral pain. Using repetitive distension of the colon and rectum (rCRD) as a sensitization paradigm, we have identified that the spontaneous firing rates of ACC neurons and the CRD-stimulated neuronal firings were increased after repetitive visceral noxious stimulation. This correlates with increases in visceral pain responses (visceromotor responses, VMRs). Two multichannel arrays of electrodes were implanted in the MT and ACC. Recordings were performed in free-moving rats before and after repeated CRD treatment. Power spectral density analysis showed that the local field potential (LFP) recorded in the ACC displayed increases in theta band power (4-10 Hz) that were modulated by rCRD. Neural spike activity in the ACC becomes synchronized with ongoing theta oscillations of LFP. Furthermore, cross correlation analysis showed augmented synchronization of thalamo-ACC theta band LFPs, which was consistent with an increase of neuronal communication between the two regions. In conclusion, these results reveal theta oscillations and theta-frequency phase-locking as prominent features of neural activity in the ACC and a candidate neural mechanism underlying acute visceral pain.

Conference Articles

1. Recruitment of Small Synergistic Movement Makes a Good Pianist

   B. Jelfs, S. Zhou, B. K. Y. Wong, C. Tin, and R. H. M. Chan

   In Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2015, pp. 242–245.

   abstract doi pdf slides

   Time-varying synergies from kinematic data can be used to discern fundamental patterns of movement. We show through simultaneous extraction of synergies from both novice and experienced pianists that movement common to both groups can be identified. The extracted synergies successfully allow for the majority of the variability of the data to be accounted for by a limited number of components. Furthermore, classification of the weightings representing the recruitment of each of the synergies accurately distinguishes between the piano playing of the two groups of subjects. However, the major differences between the two groups lie not in the synergies representing the majority of the variance of the data but in the recruitment of smaller synergies.

Journal Articles

1. A Unifying Framework for the Analysis of proportionate NLMS Algorithms

   B. Jelfs, and D. P. Mandic

   International Journal of Adaptive Control and Signal Processing, 2014, vol. 29, no. 9, pp. 1073–1085.

   abstract doi pdf

   Despite being a de facto standard in sparse adaptive filtering, the two most important members of the class of proportionate normalised least mean square (PNLMS) algorithms are introduced empirically. Our aim is to provide a unifying framework for the derivation of PNLMS algorithms and their variants with an adaptive step-size. These include algorithms with gradient adaptive learning rates and algorithms with adaptive regularisation parameters. Convergence analysis is provided for the proportionate least mean square (PLMS) algorithm in both the mean and mean square sense and bounds on its parameters are derived. An alternative, more insightful approach to the convergence analysis is also presented and is shown to provide an estimate of the optimal step-size of the PLMS. Incorporating the so obtained step-size into the PLMS gives the standard PNLMS together with a unified framework for introducing other adaptive learning rates. Simulations on benchmark sparse impulse responses support the approach.

Journal Articles

1. Modelling Noninvasively Measured Cerebral Signals during a Hypoxemia Challenge: Steps towards Individualised Modelling

   B. Jelfs, M. Banaji, I. Tachtsidis, C. E. Cooper, and C. E. Elwell

   PLoS ONE, 2012, vol. 7, no. 6:e38297. pp. 1073–1085.

   abstract doi pdf code

   Noninvasive approaches to measuring cerebral circulation and metabolism are crucial to furthering our understanding of brain function. These approaches also have considerable potential for clinical use “at the bedside”. However, a highly nontrivial task and precondition if such methods are to be used routinely is the robust physiological interpretation of the data. In this paper, we explore the ability of a previously developed model of brain circulation and metabolism to explain and predict quantitatively the responses of physiological signals. The five signals all noninvasively-measured during hypoxemia in healthy volunteers include four signals measured using near-infrared spectroscopy along with middle cerebral artery blood flow measured using transcranial Doppler flowmetry. We show that optimising the model using partial data from an individual can increase its predictive power thus aiding the interpretation of NIRS signals in individuals. At the same time such optimisation can also help refine model parametrisation and provide confidence intervals on model parameters. Discrepancies between model and data which persist despite model optimisation are used to flag up important questions concerning the underlying physiology, and the reliability and physiological meaning of the signals.

2. An Adaptive Approach for the Identification of Improper Complex Signals

   B. Jelfs, D. P. Mandic, and S. C. Douglas

   Signal Process., 2012, vol. 92, no. 2, pp. 335–344.

   abstract doi pdf code

   A real-time approach for the identification of second-order noncircularity (improperness) of complex valued signals is introduced. This is achieved based on a convex combination of a standard and widely linear complex adaptive filter, trained by the corresponding complex least mean square (CLMS) and augmented CLMS (ACLMS) algorithms. By providing a rigorous account of widely linear autoregressive modelling the analysis shows that the monitoring of the evolution of the adaptive convex mixing parameter within this structure makes it possible to both detect and track the complex improperness in real time, unlike current methods which are block based and static. The existence and uniqueness of the solution are illustrated through the analysis of the convergence of the convex mixing parameter. The analysis is supported by simulations on representative datasets, for a range of both proper and improper inputs.

3. Modelling of Brain Consciousness Based on Collaborative Adaptive Filters

   L. Li, Y. Xia, B. Jelfs, J. Cao, and D. P. Mandic

   Neurocomputing, 2012, vol. 76, no. 1, pp. 36–43.

   abstract doi pdf code

   A novel method for the discrimination between discrete states of brain consciousness is proposed, achieved through examination of nonlinear features within the electroencephalogram (EEG). To allow for real time modes of operation, a collaborative adaptive filtering architecture, using a convex combination of adaptive filters is implemented. The evolution of the mixing parameter within this structure is then used as an indication of the predominant nature of the EEG recordings. Simulations based upon a number of different filter combinations illustrate the suitability of this approach to differentiate between the coma and quasi-brain-death states based upon fundamental signal characteristics.

Conference Articles

1. Individualised Optimisation of Modelled Cerebral Oxygenation Near-Infrared Spectroscopy Signals

   B. Jelfs, J. Panovska-Griffiths, I. Tachtsidis, M. Banaji, and C. Elwell

   In Biomedical Optics and 3-D Imaging, 2012, no. JM3A.32. pp. 242–245.

   abstract doi pdf poster code

   Responses of NIRS signals in a healthy volunteer are predicted using a model of brain circulation. Optimisation using partial data is shown to increase the model’s predictive power which can aid the interpretation of NIRS signals in individuals.

Journal Articles

1. An Augmented Echo State Network for Nonlinear Adaptive Filtering of Complex Noncircular Signals

   Y. Xia, B. Jelfs, M. M. Van Hulle, J. C. Principe, and D. P. Mandic

   IEEE Transactions on Neural Networks, 2011, vol. 22, no. 1, pp. 74–83.

   abstract doi pdf

   A novel complex echo state network (ESN), utilizing full second-order statistical information in the complex domain, is introduced. This is achieved through the use of the so-called augmented complex statistics, thus making complex ESNs suitable for processing the generality of complex-valued signals, both second-order circular (proper) and noncircular (improper). Next, in order to deal with nonstationary processes with large nonlinear dynamics, a nonlinear readout layer is introduced and is further equipped with an adaptive amplitude of the nonlinearity. This combination of augmented complex statistics and enhanced adaptivity within ESNs also facilitates the processing of bivariate signals with strong component correlations. Simulations in the prediction setting on both circular and noncircular synthetic benchmark processes and real-world noncircular and nonstationary wind signals support the analysis.

Journal Articles

1. Characterisation of Signal Modality: Exploiting Signal Nonlinearity in Machine Learning and Signal Processing

   B. Jelfs, S. Javidi, P. Vayanos, and D. Mandic

   J. Signal Process. Syst., 2010, vol. 61, no. 1, pp. 105–115.

   abstract doi pdf code

   A novel method for online tracking of the changes in the nonlinearity within both real-domain and complex-valued signals is introduced. This is achieved by a collaborative adaptive signal processing approach based on a hybrid filter. By tracking the dynamics of the adaptive mixing parameter within the employed hybrid filtering architecture, we show that it is possible to quantify the degree of nonlinearity within both real- and complex-valued data. Implementations for tracking nonlinearity in general and then more specifically sparsity are illustrated on both benchmark and real world data. It is also shown that by combining the information obtained from hybrid filters of different natures it is possible to use this method to gain a more complete understanding of the nature of the nonlinearity within a signal. This also paves the way for building multidimensional feature spaces and their application in data/information fusion.

Conference Articles

1. Blind Extraction of Noncircular Complex Signals Using a Widely Linear Predictor

   S. Javidi, B. Jelfs, and D. P. Mandic

   In Proc. IEEE Workshop on Statistical Signal Processing, 2009, no. JM3A.32. pp. 501–504.

   abstract doi pdf

   Real valued blind source extraction based on a linear predictor is extended to the complex domain using recent advances in complex domain statistics. It is shown that, in general, the mean square prediction error of the algorithm depends both on the covariance matrix and the pseudo-covariance matrix of the source signals. To fully utilise the available information, it is thus natural to adopt a widely linear predictor to extract the latent sources from the observed mixture. This way, we derive a new algorithm for the extraction of general complex signals and provide simulation results using benchmark complex data.

Book Chapters

1. Collaborative Adaptive Filters for Online Knowledge Extraction and Information Fusion

   B. Jelfs, P. Vayanos, S. Javidi, V. S. L. Goh, and D. Mandic

   In Signal Processing Techniques for Knowledge Extraction and Information Fusion, D. Mandic et al. (eds), Springer, 2008, pp. 3–21.

   abstract doi pdf code

   We present a method for extracting information (or knowledge) about the nature of a signal. This is achieved by employing recent developments in signal characterisation for online analysis of the changes in signal modality. We show that it is possible to use the fusion of the outputs of adaptive filters to produce a single collaborative hybrid filter and that by tracking the dynamics of the mixing parameter of this filter rather than the actual filter performance, a clear indication as to the nature of the signal is given. Implementations of the proposed hybrid filter in both the real R and the complex C domains are analysed and the potential of such a scheme for tracking signal nonlinearity in both domains is highlighted. Simulations on linear and nonlinear signals in a prediction configuration support the analysis; real world applications of the approach have been illustrated on electroencephalogram (EEG), radar and wind data.

Conference Articles

1. Collaborative adaptive filtering in the complex domain

   B. Jelfs, Y. Xia, D. P. Mandic, and S. C. Douglas

   In Proc. IEEE Workshop on Machine Learning for Signal Processing, 2008, no. JM3A.32. pp. 421–425.

   abstract doi pdf slides code

   A novel hybrid filter combining the complex least mean square (CLMS) and augmented CLMS (ACLMS) algorithms for complex domain adaptive filtering is introduced. The ACLMS has been shown to have improved performance in terms of prediction of non-circular complex data compared to that of the CLMS. By taking advantage of this along with the faster convergence of the CLMS, the hybrid filter is shown to give improved performance over both algorithms for both circular and non-circular data. Simulations on complex-valued synthetic and real world data support the effectiveness of this approach.

2. Signal Modality Characterisation Using Collaborative Adaptive Filters

   B. Jelfs, and D. Mandic

   In Proc. IAPR Workshop on Cognitive Information Processing, 2008, no. JM3A.32. pp. 421–425.

   abstract html pdf slides code

   A method for extracting information (or knowledge) about the nature of a signal is presented, this is achieved by tracking the dynamics of the mixing parameter within a hybrid filter rather than the actual filter performance. Implementations of the hybrid filter for tracking the nonlinearity and the sparsity of a signal are illustrated and simulations on benchmark synthetic data in a prediction configuration support the analysis. It is then shown that by combining the information obtained from both hybrid filters it is possible to use this method to gain a more complete understanding of the nature of signals and changes in signal modality.

3. Online Tracking of the Degree of Nonlinearity Within Complex Signals

   D. P. Mandic, P. Vayanos, S. Javidi, B. Jelfs, and K. Aihara

   In Proc. IEEE International Conference on Acoustics, Speech and Signal Processing, 2008, no. JM3A.32. pp. 2061–2064.

   abstract doi pdf code

   A novel method for online tracking of the changes in the non- linearity within complex-valued signals is introduced. This is achieved by a collaborative adaptive signal processing approach by means of a hybrid filter. By tracking the dynamics of the adaptive mixing parameter within the employed hybrid filtering architecture, we show that it is possible to quantify the degree of nonlinearity within complex-valued data. Simulations on both benchmark and real world data support the approach.

Book Chapters

1. Exploiting Nonlinearity in Adaptive Signal Processing

   P. Vayanos, M. Chen, B. Jelfs, and D. P. Mandic

   In Advances in Nonlinear Speech Processing, M. Chetouani et al. (eds), Springer Berlin Heidelberg, 2007, pp. 57–77.

   abstract doi pdf code

   Quantitative performance criteria for the analysis of machine learning architectures and algorithms have been long established. However, the qualitative performance criteria, e.g., nonlinearity assessment, are still emerging. To that end, we employ some recent developments in signal characterisation and derive criteria for the assessment of the changes in the nature of the processed signal. In addition, we also propose a novel online method for tracking the system nonlinearity. A comprehensive set of simulations in both the linear and nonlinear settings and their combination supports the analysis.

Conference Articles

1. Assessment of Nonlinearity in Brain Electrical Activity: A DVV Approach

   M. Chen, T. Rutkowski, B. Jelfs, G. Souretis, J. Cao, and D. Mandic

   In Proc. RISP International Workshop on Nonlinear Circuits and Signal Processing, 2007, no. JM3A.32. pp. 461–464.

   abstract html pdf code

   In this paper, we provide a comprehensive analysis of the EEG data recorded in a steady-state visual evoked potentials (SSVEPs) experiment. This is based upon the recently introduced the ‘delay vector variance’ (DVV) method for signal characterisation, which examines the local predictability of the signal in the phase space and checks simultaneously the signal nonlinearity and the determinism. Simulation results support the analysis.

2. A Unifying Approach to the Derivation of the Class of PNLMS Algorithms

   B. Jelfs, D. P. Mandic, and A. Cichocki

   In Proc. International Conference on Digital Signal Processing, 2007, no. JM3A.32. pp. 35–38.

   abstract doi pdf slides

   A unifying approach to the derivation of the class of proportionate normalised least mean square (PNLMS) algorithms is provided. This is an important class of algorithms where the two most used algorithms are introduced empirically. It is shown that it is possible to derive PNLMS algorithms as a result of an optimisation procedure. This is achieved in a rigorous way, starting from the standard LMS through to the PNLMS with the “sparsification” factor in both the numerator and denominator of the weight update. The proposed approach is generic and also applies to other LMS types of adaptive algorithms. Simulations on benchmark sparse impulse responses support the approach.

3. A Class of Adaptively Regularised PNLMS Algorithms

   B. Jelfs, D. P. Mandic, and J. Benesty

   In Proc. International Conference on Digital Signal Processing, 2007, no. JM3A.32. pp. 19–22.

   abstract doi pdf slides

   A class of algorithms representing a robust variant of the proportionate normalised least-mean-square (PNLMS) algorithm is proposed. To achieve this, adaptive regularisation is introduced within the PNLMS update, with the analysis conducted for both individual and global regularisation factors. The update of the adaptive regularisation parameter is also made robust, to improve steady state performance and reduce computational complexity. The proposed algorithms are better suited not only for operation in nonstationary environments, but are also less sensitive to changes in the input dynamics and the choice of their parameters. Simulations in a sparse environment show the proposed class of algorithms offer enhanced performance and increased stability over the standard PNLMS.

4. Collaborative Adaptive Learning using Hybrid Filters

   D. Mandic, P. Vayanos, C. Boukis, B. Jelfs, S. L. Goh, T. Gautama, and T. Rutkowski

   In Proc. IEEE International Conference on Acoustics, Speech and Signal Processing, 2007, no. JM3A.32. pp. 921–924.

   abstract doi pdf code

   A novel stable and robust algorithm for training of finite impulse response adaptive filters is proposed. This is achieved based on a convex combination of the least mean square (LMS) and a recently proposed generalised normalised gradient descent (GNGD) algorithm. In this way, the desirable fast convergence and stability of GNGD is combined with the robustness and small steady state misadjustment of LMS. Simulations on linear and nonlinear signals in the prediction setting support the analysis.

Book Chapters

1. An Online Method for Detecting Nonlinearity Within a Signal

   B. Jelfs, P. Vayanos, M. Chen, S. L. Goh, C. Boukis, T. Gautama, T. Rutkowski, T. Kuh, and D. Mandic

   In Knowledge-Based Intelligent Information and Engineering Systems, B. Gabrys et al. (eds), Springer Berlin Heidelberg, 2006, pp. 1216–1223.

   abstract doi pdf code

   A novel method for online analysis of the changes in signal modality is proposed. This is achieved by tracking the dynamics of the mixing parameter within a hybrid filter rather than the actual filter performance. An implementation of the proposed hybrid filter using a combination of the Least Mean Square (LMS) and the Generalised Normalised Gradient Descent (GNGD) algorithms is analysed and the potential of such a scheme for tracking signal nonlinearity is highlighted. Simulations on linear and nonlinear signals in a prediction configuration support the analysis. Biological applications of the approach have been illustrated on EEG data of epileptic patients.

Conference Articles

1. Towards Online Monitorying of the Changes in Signal Modality: The Degree of Sparsity

   B. Jelfs, and D. Mandic

   In Proc. IMA International Conference on Mathematics in Signal Processing, 2006, no. JM3A.32. pp. 29–32.

   abstract html pdf slides code

   A novel method for online analysis of the changes in signal modality is proposed. This is achieved by tracking the dynamics of the mixing parameter within a hybrid filter rather than the actual filter performance. Implementations of the proposed hybrid filter using a combination of sparse and nonsparse algorithms are analysed highlighting the potential of such a scheme for tracking signal sparsity. Simulations in a prediction setting support the analysis and biological applications of the approach have been illustrated on EEG data of epileptic patients.

Conference Articles

1. Noninvasive Determination of Fetal Heart Rate and Short Term Heart Rate Variability Using Solely Doppler Ultrasound with Autocorrelation

   F. Schlindwein, A. Boardman, S. Vali, N. Wright, B. Jelfs, S. Mauger, A. Das, J. Waugh, R. Pannerai, and D. Evans

   In Proc. International Conference on Medical Signal & Information Processing, 2004, no. JM3A.32. pp. 29–32.

   abstract pdf

   Complications of pregnancy, labour and delivery represent the main risk for perinatal asphyxia. The Confidential Enquiry into Stillbirths and Deaths in Infancy reported a mortality rate of 0.62 per 1000 births in England, Wales and Northern Ireland in 1999. In this work we compare the heart rate information that can be derived from the Doppler audio signal with that derived from the ECG from a scalp electrode, on a beat-to-beat basis. Challenges associated with using the Doppler audio signal include elimination of the noise arising from the movement of the foetal heart and valves and correct peak blood velocity identification. From this beat-to-beat analysis, we can explore the use of a quantitative measure of heart rate variability for the detection of foetal distress. Previously it has been found, in rats, that this measure of heart rate variability increases considerably when asphyxia occurs. Data has been collected from five mothers in the first stage of labour. Currently, the comparison of heart rate signals from the Doppler audio and ECG signals shows an excellent agreement on a beat-to-beat basis (rms error of less than 10 ms). In this pilot study the fetal heart rate variability has remained at a relatively constant low level, which is expected, as only ‘normal’ labours have been studied so far. Our results are very encouraging and show that there is potential for the use of this method of intrapartum monitoring and the technique could also be applied at antepartum, but this is a pilot study and many more results are needed before firm conclusions and recommendations can be made.