Table 4.
Average classification accuracy (in %) of resized time-frequency image representations.
Signal Representation | Sound Event | Speech Command | ||
---|---|---|---|---|
Validation | Test | Validation | Test | |
Resized spectrogram (nearest-neighbour) | 93.51 | 94.19 | 93.20 | 93.10 |
Resized spectrogram (bilinear) | 95.71 | 96.31 | 94.10 | 93.81 |
Resized spectrogram (bicubic) | 96.02 | 96.59 | 94.03 | 93.97 |
Resized spectrogram (Lanczos-2) | 95.75 | 96.42 | 93.75 | 93.77 |
Resized spectrogram (Lanczos-3) | 97.01 | 97.13 | 94.02 | 93.75 |