Table 9 Classification accuracy of the proposed sconn deep learning models on the RAVDESS, SAVEE and TESS datasets.
From: Stacked convolutional neural network for emotion recognition using multi feature speech analysis
Feature Sets | RAVDESS | SAVEE | TESS |
|---|---|---|---|
MFCC (40) | 91.51% | 91.43% | 99.93% |
Mel Spectrogram (128) | 90.63% | 94.76% | 99.68% |
MFCC (40) + Mel Spectrogram (128) | 93.30% | 95.00% | 99.93% |
