ODESSA

HMM based Automatic Speech Recognition System

The ODESSA system, developed as part of the EE516 - Computer Speech Processing course at the University of Washington, leverages Hidden Markov Models (HMMs) to provide efficient and accurate automatic speech recognition (ASR).

AI generated image of ODESSA

Key Features:

Efficient ASR Technology: Utilizes Mel-Frequency Cepstral Coefficients (MFCCs) for feature extraction and HMM training to ensure high performance with low computational overhead.

Real-Time Processing: Capable of real-time speech detection and recognition, continuously monitoring audio input and providing immediate responses.

Robust Performance: Demonstrates low Word Error Rates (WER) across various utterances and acoustic conditions.

Resources:

This project showcases the practical application of HMMs in speech recognition, offering a reliable solution for resource-constrained environments.