Table 11 Sensitivity and specificity of the sconn model for each emotion class using the mel Spectrogram, MFCC, and mel Spectrogram + MFCC features of the RAVDESS dataset.
From: Stacked convolutional neural network for emotion recognition using multi feature speech analysis
Emotion Classes | Mel Spectrogram | MFCC | Mel Spectrogram + MFCC | |||
|---|---|---|---|---|---|---|
Sensitivity | Specificity | Sensitivity | Specificity | Sensitivity | Specificity | |
angry | 92.89% | 99.13% | 95.43% | 98.61% | 97.97% | 99.04% |
calm | 93.33% | 98.20% | 98.33% | 98.20% | 96.67% | 99.66% |
disgust | 91.01% | 97.92% | 89.42% | 98.70% | 92.06% | 97.75% |
fearful | 87.21% | 98.38% | 84.88% | 98.81% | 89.53% | 99.40% |
happy | 89.40% | 99.29% | 90.32% | 97.96% | 90.32% | 98.94% |
sad | 84.29% | 97.92% | 89.01% | 98.61% | 90.58% | 99.05% |
surprised | 95.96% | 98.25% | 92.93% | 99.21% | 95.96% | 98.34% |
