Research on Speech Emotion Recognition Method Based A-CapsNet

Speech emotion recognition is a crucial work direction in speech recognition. To increase the performance of detection, researchers have worked relentlessly to improve data augmentation, feature extraction, and pattern formation. address concerns limited resources model training overfitting, A-CapsNet, neural network based on augmentation methodologies, proposed this research. In order solve issue scarcity achieve goal noise from Noisex-92 database first combined with four different division methods (emotion-independent random-division, emotion-dependent emotion-independent cross-validation methods, abbreviated as EIRD, EDRD, EICV EDCV, respectively). The EMODB then used analyze compare paper under signal-to-noise ratios, results show that are effective.