TABLE 3.

Separable convolutional neural network (SepCNN) architecture for within-participant recognition.

Layer	Kernel	Size	Options	Output
Input		200 × 30
BatchNorm1D				200 × 30
DepthConv1D	30	10	Stride = 6, Padding = 4	34 × 30
BatchNorm1D				34 × 30
PointConv1D	4	1	Stride = 1, Padding = 0	34 × 4
Activation			Tanh	34 × 4
Flatten				136
Activation			Sigmoid	1