Research Article
Application of Generative Adversarial Nets (GANs) in Active Sound Production System of Electric Automobiles
Table 2
Discriminator structure of the raw audio GAN model.
| | Layer | Type | Length and width of convolution kernels | Number of convolution kernels |
| | Input layer G(z) | Input layer | — | — | | Conv1D (stride = 4) | Convolutional layer | 1∗25 | 64 | | LReLU (α = 0.2) | Activation layer | — | — | | Phase conversion | Phase conversion | — | — | | Conv1D (stride = 4) | Convolutional layer | 1∗25 | 128 | | LReLU (α = 0.2) | Activation layer | — | — | | Phase conversion (n = 2) | Phase conversion | — | — | | Conv1D (stride = 4) | Convolutional layer | 1∗25 | 256 | | LReLU (α = 0.2) | Activation layer | — | — | | Phase conversion (n = 2) | Phase conversion | — | — | | Conv1D (stride = 4) | Convolutional layer | 1∗25 | 512 | | LReLU (α = 0.2) | Activation layer | — | — | | Phase conversion (n = 2) | Phase conversion | — | — | | Conv1D (stride = 4) | Convolutional layer | 1∗25 | 1024 | | LReLU (α = 0.2) | Activation layer | — | — | | Reshape | Reconstruction layer | — | — | | Fully connected layer | Fully connected layer | 16384∗1 | 64 |
|
|
