/Double-talk-Detection-aided-Residual-Echo-Suppression-via-Spectrogram-Masking-and-Refinement

Double-Talk Detection-Aided Residual Echo Suppression via Spectrogram Masking and Refinement

Primary LanguagePythonMIT LicenseMIT

Double-talk Detection-aided Residual Echo Suppression via Spectrogram Masking and Refinement

This repository contains a PyTorch implementation of the deep-learning model architectures described in the paper "Double-talk Detection-aided Residual Echo Suppression via Spectrogram Masking and Refinement" (yet to be published). In the paper we describe a two-stage approach to residual echo suppression in the low singal-to-echo ratio (SER) scenario. We also provide a few .WAV audio samples to demonestrate the proposed system's performance.