speech-denoising | Speech Denoising project for the Deep Learning course

speech-denoising is a Jupyter Notebook library. speech-denoising has no bugs, it has no vulnerabilities, it has a Permissive License and it has low support. You can download it from GitHub.

Speech Denoising project for the Deep Learning course at Tsinghua University, Spring semester 2021. This code uses a Source Separation approach to recover clean speech signals from a noisy acoustic environment. The high diversity of noises in the dataset motivated to perform the optimization on 2 sources, namely the clean speech signal and the background noise. We study 4 different model architectures (2 for time domain and 2 for frequency domain), and compare their performance using Source Separation metrics (SDR, SIR, SAR) and Speech Quality metrics (PESQ, STOI).