Speaker-Independent Speech Recognition Using Hidden Markov Models

Resource Overview

Implementation of speaker-independent speech recognition through Hidden Markov Models (HMM) using MATLAB 7.5 [2008b], featuring feature extraction, model training, and pattern matching algorithms

Detailed Documentation

This project focuses on developing a speaker-independent speech recognition system using Hidden Markov Models (HMM) implemented in MATLAB 7.5 (2008b release). The system employs MFCC (Mel-frequency cepstral coefficients) for feature extraction, which captures the spectral characteristics of speech signals. The implementation includes Baum-Welch algorithm for HMM parameter training and Viterbi algorithm for decoding the most likely sequence of phonemes or words. The code structure involves separate modules for feature extraction, HMM training, and recognition phases, ensuring robust performance across different speakers by utilizing universal acoustic models rather than speaker-dependent adaptations.