Recognition for Human Gestures Based on Convolutional Neural Network Using the Off-the-Shelf Wi-Fi Routers

Table 1

Hierarchical structure of proposed 1D-CNN.

Number	Layer name	Function	Parameter setting
1	Convolutional layer	Extracting local features	Number of convolutional kernels: 16; Size of convolutional kernel: 3
2	Pooling layer	Spatially dimension reduction	Number of pooling kernels: 16; Size of pooling kernel: 3
3	Convolutional layer	Extracting local features	Number of convolutional kernels: 32; Size of convolutional kernel: 3
4	Pooling layer	Spatially dimension reduction	Number of pooling kernels: 16; Size of pooling kernel: 3
5~7	Convolutional layer	Extracting local features	Number of convolutional kernels: 64; Size of convolutional kernel: 3
8	Pooling layer	Spatially dimension reduction	Number of pooling kernels: 16; Size of pooling kernel: 3
9	Flat layer	Realizing the transition from convolution layer to full connection layer
10	Fully connected layer		Output size: 64
11	Fully connected layer	Classification and recognition using SVM classifier	Output size: 3