An Efficient sEMG Signal Acquisition System for Decoding of Hand Movements

A pattern recognition system based on human movements or electromyography signals comprises various stages, from signal acquisition to control. The primary signal acquisition stage is the most important since the further stages largely depend on the quality of recorded signals. This work proposes an NI ELVIS-based signal acquisition system for recording human muscle movement signals. To capture muscle signals, NI ELVIS board is integrated with a BioAmp-EMG biopotential amplifier module. Six distinct hand movements for five subjects are recorded to evaluate the proposed hardware setup. The obtained signal-to-noise ratio (SNR) for different measurements is above 10 dB, and the average of the SNR values is approximately 16.44 dB, which signifies the high noise immunity of the proposed hardware. The SNR is also compared with the state of the art available sEMG acquisition modules. Further, eight time-domain features, including a novel feature, have been extracted to train Linear discriminant Analysis (LDA), Decision Tree (DT), and Multi-layer Perceptron (MLP) classifiers for a pattern recognition-based movement decoding application. The testing results show an average accuracy of 91.01 % for LDA, 83.82 % for DT, and 80.29 % for MLP classifiers.