6 April 2020 in IEEE Transactions on Signal Processing
Electromyographic (EMG) signals sensed at the skin surface on the forearm can be used to accurately infer wrist- hand poses. However, this is only possible when the EMG sensors are carefully placed over specific arm muscles. This cannot be guaranteed for wearable devices, which could acquire EMG from anywhere on the forearm. As a result, these devices detect fewer poses, less accurately. In addition, the complexity of the time-frequency analysis used in placed-sensor systems precludes real-time detection using the simple embedded processors on EMG wearables. This paper describes an approach which resolves both these shortcomings. It shows that, when random sensor placement is adopted, wrist-hand movement detection with performance equal the state-of-the-art can be achieved, with only 10% of the computational complexity. This latter property allows the first real-time wrist-hand movement detector using only simple embedded processors; specifically when using on ARM Cortex- A53 processor, execution time is lowered by 90% against the state-of-the-art, with no reduction in detection performance. It is shown how this can be further reduced by 30% by using fewer EMG channels or features, whilst maintaining good detection performance. To the best of the authors' knowledge, this is the first record of real-time high-performance wrist-hand movement detection for standalone, battery-powered EMG wearables.
11 January 2021 in IEEE Access
Electromyogram (EMG) acquisition and analysis is growing in importance with human attempts to interact with and control equipment such as robots, prostheses or virtual environments. In some cases, only approved users should be permitted these capabilities. For these applications, securing EMG-based control is a major open question - to the best of the authors’ knowledge, no prior art exists which can identify individuals from a wide range of wrist-hand gestures EMG readings within the wearable device. This paper addresses this problem. Techniques are presented which allow EMG to be used as a biometric, allowing users to verify themselves. An EMG-sensing armband attached to the lower forearm is used to anonymously authenticate users as a member of an approved group, or to identify themselves uniquely. For the development of extensive biometric system, three EMG datasets with similar EMG sensing in different sessions were exploited. For verification, accuracy of up to 93% is achieved, with 92% achieved for identification. The system is also shown to operate in real-time on an ARM Cortex A-53 embedded processor suitable for housing in an EMG wearable device, incurring latencies of 1.06 ms and 1.61 ms for verification and identification respectively. These metrics are comfortably sufficient for use in real-time, battery-powered EMG authentication devices.
19 March 2021 in Journal of Neural Engineering
Objective. To develop an automated system to classify the severity of hypoxic-ischaemic encephalopathy injury (HIE) in neonates from the background electroencephalogram (EEG). Approach. By combining a quadratic time–frequency distribution (TFD) with a convolutional neural network, we develop a system that classifies 4 EEG grades of HIE. The network learns directly from the two-dimensional TFD through 3 independent layers with convolution in the time, frequency, and time–frequency directions. Computationally efficient algorithms make it feasible to transform each 5 min epoch to the time–frequency domain by controlling for oversampling to reduce both computation and computer memory. The system is developed on EEG recordings from 54 neonates. Then the system is validated on a large unseen dataset of 338 h of EEG recordings from 91 neonates obtained across multiple international centres. Main results. The proposed EEG HIE-grading system achieves a leave-one-subject-out testing accuracy of 88.9% and kappa of 0.84 on the development dataset. Accuracy for the large unseen test dataset is 69.5% (95% confidence interval, CI: 65.3%–73.6%) and kappa of 0.54, which is a significant ( ) improvement over a state-of-the-art feature-based method with an accuracy of 56.8% (95% CI: 51.4%–61.7%) and kappa of 0.39. Performance of the proposed system was unaffected when the number of channels in testing was reduced from 8 to 2—accuracy for the large validation dataset remained at 69.5% (95% CI: 65.5%–74.0%). Significance. The proposed system outperforms the state-of-the-art machine learning algorithms for EEG grade classification on a large multi-centre unseen dataset, indicating the potential to assist clinical decision making for neonates with HIE.
10 March 2023 in Scientific Data
This report describes a set of neonatal electroencephalogram (EEG) recordings graded according to the severity of abnormalities in the background pattern. The dataset consists of 169 hours of multichannel EEG from 53 neonates recorded in a neonatal intensive care unit. All neonates received a diagnosis of hypoxic-ischaemic encephalopathy (HIE), the most common cause of brain injury in full term infants. For each neonate, multiple 1-hour epochs of good quality EEG were selected and then graded for background abnormalities. The grading system assesses EEG attributes such as amplitude, continuity, sleep–wake cycling, symmetry and synchrony, and abnormal waveforms. Background severity was then categorised into 4 grades: normal or mildly abnormal EEG, moderately abnormal EEG, majorly abnormal EEG, and inactive EEG. The data can be used as a reference set of multi-channel EEG for neonates with HIE, for EEG training purposes, or for developing and evaluating automated grading algorithms.