Wrist EMG Monitoring Using Neural Networks Techniques

Table 6

Comparison of our system with relevant literature.

System	Classification method	Employed features	Channels number	(%) Accuracy
Chu et al. [27]	PCA + self-organizing feature map	Wave package transform	4	97.4
Ahsan et al. [28]	BPNN	MAV, RMS, VAR, standard deviation, zero crossing, and waveform length.	1	86.76
Tohi et al. [29]	GANN	Fast fourier transform	4	77.5
Khezri and Jahed [30]	PCA + NN + FIS	MAV, slope sign changes, autoregressive model coefficients (time domain); discrete wavelet transform (frequency domain)	2	82
Liu et al. [15]	CKLM + SVM	coefficients autoregressive model and histogram of EMG	3	93.54
George et al. [31]	LDA + NN	mean, variance, skewness, and absolute value	1	90
Matsumura et al. [32]	PCA + NN	Fast Fourier transform	4	72.86
This work	MLP + BP	—	2	95.69

Abbreviations, principal components analysis (PCA), back-propagation neural network (BPNN), linear discriminant analysis (LDA), genetic algorithm combined with neural network (GANN), cascaded kernel learning machine (CKML), and support vector machine (SVM).