Decoding of Walking Imagery and Idle State Using Sparse Representation Based on fNIRS

Table 1

Average classification accuracy (%) obtained by cdSRC, SVM, KNN, LDA, and LR under different features and their combinations of HbO signals during walking imagery and idle state (binary classification) of 15 subjects.

Classifier	Statistics type	Different features of HbO and their combinations
		Mean	Peak	RMS	M&P	M&R	P&R	M&P&R
cdSRC	Average accuracy	86.98	84.05	82.63	90.32	88.21	87.51	91.55
		2.63	3.52	3.76	3.58	3.54	3.69	3.30
SVM	Average accuracy	78.92	74.82	72.52	84.62	77.11	76.63	86.37
		6.21	5.43	7.03	5.23	3.98	3.74	4.42
KNN	Average accuracy	74.96	73.72	72.99	83.76	76.72	76.24	85.65
		4.61	4.16	5.15	5.01	4.95	3.48	5.01
LDA	Average accuracy	77.32	76.03	73.98	85.24	79.60	75.33	86.43
		4.93	4.61	5.01	4.35	5.21	5.81	4.41
LR	Average accuracy	71.44	70.29	69.52	74.31	73.60	73.05	76.14
		5.33	5.83	5.61	4.02	4.54	5.26	5.32