Research Article
Through-Wall Radar Classification of Human Posture Using Convolutional Neural Networks
Table 1
CNN structure definitions of the proposed approach.
| Layer | Parameters | Value |
| Input layer | Input size | |
| Convolutional layer | Kernel size | | # of filters | 16 | Transfer function | ReLU |
| Batch normalization layer | Transfer function | ReLU |
| Pooling layer | Kernel size | | # of filters | 16 | Transfer function | ReLU |
| Convolutional layer | Kernel size | | # of filters | 32 | Transfer function | ReLU |
| Batch normalization layer | Transfer function | ReLU |
| Pooling layer | Kernel size | | # of filters | 32 | Transfer function | ReLU |
| Fully connected layer | # of neurons | 100 | Transfer function | ReLU | Dropout | 0.5 |
| Fully connected layer | # of neurons | 3 | Classifier | Softmax |
| Output layer | # of outputs | 3 | Classes | {empty, standing, sitting} |
|
|