Emotion Recognition using BiLSTM Classifier

Abstract

Human emotions are play a vital role in work locations and clinical studies. Speech emotion recognition (SER) is easy for humans and it is bit difficult for machines like computers and robots. These systems are unable to identify emotion of a user and are unable to react to their feelings. To improve human-machine interaction, a deep learning (DL) based approach is proposed. This paper presents a SER system using Bidirectional Long Short-Term Memory (Bi-LSTM). Features such as Gamma tone & Mel-frequency cepstral coefficients (GTCC & MFCC) are extracted for SER. Ability of proposed DL approach is observed at different noisy conditions, testing rates and with EMO-DB data set.