EchoEval is an open-source toolkit for objectively assessing Residual Echo Suppression (RES) performance, especially in double-talk conditions.
This project provides two key evaluation metrics:
- DSML (Desired-Speech Maintained Level) – measures how well the target speech is preserved.
- RESL (Residual-Echo Suppression Level) – quantifies the effectiveness of echo suppression.
✅ Standardized DSML & RESL evaluation metrics
✅ Compatible with deep learning, traditional signal processing, and hybrid models
✅ Applicable to speech enhancement, speech recognition, and echo cancellation
✅ Open-source, reproducible, and easy to integrate into Python-based workflows
This repository is designed for researchers, audio engineers, and developers looking to enhance speech quality evaluation.