Speech Emotion Recognition Using LSTM

Abstract

Emotion recognition from speech is a crucial task in many applications, such as affective computing, psychological research, and human-computer interaction. LSTM networks shows promise in modeling sequential data, such as voice signals, because of its capacity to capture long-term dependencies. The main objective of this research work is to assign emotions to one of the four categories: fear, rage, sadness, and happiness. The samples are used to determine energy, pitch, MFCC and LPCC coefficients, speaker rate, and other important characteristics. Speech is a widely utilized signal for human-to-human communication, which implies that speech is also used for communication between people and the machines. This interactive system's aim is to improve the speech emotion recognition (SER) technology by using LSTM Model. Speech Emotion Recognition (SER) is critically important for applications that assess spoken emotions in real-time. Dataset used in this Research work are taken from kaggle which is named as Toronto emotional speech set (TESS). The dataset is arranged according to the emotion expressed by the two female actors, which is organized within a separate folder, And the audio clips are arranged with 200 targeted words. The audio file is saved in WAV format. Additionally, a system utilizing an LSTM algorithm and an MFCC features are presented in this paper's preliminary results.