Paper Key : IRJ************844
Author: Vivek Kripashankar Paswan
Date Published: 31 Oct 2023
Abstract
Speech Emotion Recognition (SER) plays a pivotal role in enhancing human-computer interaction by enabling machines to understand and respond to the emotional content embedded in spoken language. This study presents an approach to SER utilizing Hidden Markov Models (HMMs). The proposed framework leverages the temporal dynamics of speech signals, employing Mel Frequency Cepstral Coefficients (MFCCs) as feature vectors. The model is trained in a supervised manner, associating emotional states with HMM states and estimating transition probabilities. Gaussian Mixture Models (GMMs) capture the emission probabilities of HMM states, enhancing the model's ability to discriminate between emotional states. Recognition is achieved through the Viterbi algorithm, which identifies the most likely sequence of emotional states given observed features. The application of HMMs in SER offers a nuanced understanding of emotional expression in speech, contributing to the development of emotionally intelligent systems. Experimental results demonstrate the effectiveness of the proposed approach in capturing the intricate patterns of emotional dynamics within speech signals.