Amplitude independent muscle activity detection algorithm of soft robotic glove system for hemiparesis stroke patients using single sEMG channel

Hand impairment is a consequence of many neurological diseases such as stroke, where the stroke affects about 15 million people worldwide annually and it is one of the main causes of hand disability. Therefore, hand robotic devices can be used to help stroke patients to perform activities of daily living and at home rehabilitation. Control of hand robotic devices by using Surface Electromyography (sEMG) signal is the most preferred control technique due to the advantages of this method like naturalness. However, robust controlling by using such method is still a challenging process because the amplitude of these signals is not constant over the recording time due to the variations in the electrode-skin interface characteristics; these involuntary amplitude variations deteriorate the detection performance of the amplitude-dependent methods and produce false alarms. Many algorithms have been developed in the literature to detect muscle activities; however, most of these algorithms depend on amplitude features in the detection process. The performance of the amplitude-dependent methods is highly deteriorated when the signal to noise ratio (SNR) is low, such as for signals obtained from the paretic muscles. To simplify soft robotic glove systems and make them more practical for use in daily basis, they should have minimum number of sEMG channels. In spite of some algorithms that have been developed in the literature to classify some hand motions by using single channel, the current implementation of soft robotic glove systems are still employing two channels for detecting the closing and opening movements of the hand, due to the intensive calculations required by these algorithms which impose difficulties on real time implementation. This thesis addresses the aforementioned problems, by innovating an amplitude-independent and computationally efficient muscle activity detection algorithm to control a soft robotic glove intended for hemiparesis stroke patients by using single channel. The algorithm employs the First Lag Autocorrelation and the Modified Sample Entropy methods to detect and classify weak hand closing and opening muscle activities by using signal obtained from the Flexor Carpi Ulnaris forearm muscle. The detection performance of the proposed algorithm compared to three amplitude-dependent algorithms was verified on seven healthy subjects and on six hemiparesis stroke patients. The performance of the proposed algorithm has outperformed that of the amplitude-dependent algorithms regarding the detection of weak muscle activities and robustness against false alarms. High classification accuracies have been achieved for the seven healthy subjects (92%-100%) which are comparable to that obtained by applying sophisticated single channel classification algorithms in previous studies; moreover, good accuracies (70%-85%) have been obtained for the stroke patients. The computation efficiency of the proposed algorithm has enabled the implementation of the soft robotic glove system prototype by using simple hardware.

Text FK 2020 61 - ir.pdf Download (1MB)

Actions (login required)

View Item