Table 1 Classification accuracy of various methods.
Method | Data augmentation | Features extraction | Sample rate (kHz) | Accuracy (%) |
|---|---|---|---|---|
CNNs(5/9/13)23 | \ | Mel energies | 32 | 69.2 |
ResNet43 | Mix-up | HPSS | 48 | 71.9 |
CNNs_Averaging844 | \ | HRTF, NNF | 44.1 | 64.0 |
CDNN46 | Mix-up | Single frequency cepstral coefficients (SFCC), log-Mel energies | 48 | 70.4 |
3D-SEResNet | Mix-up | Mel Spectrogram | 16 | 80.1 |
3D-SEResNet | Mix-up | Log-Mel Spectrogram | 16 | 81.3 |
3D-CNN | Mix-up | STFT | 16 | 83.5 |
Our Method | Mix-up | Mel Spectrogram, Chromagram, STFT | 16 | 86.4 |
