suwoncjh / improved-speech-separation

Code for the paper: "Improved Speech Separation with Time-and-Frequency Cross-domain Joint Embedding and Clustering".

Improved Speech Separation with Time-and-Frequency Cross-domain Joint Embedding and Clustering

This is a Tensorflow implementation of speaker indepentent source separation describe here.

Environments

1. Tenforflow 1.13
2. mir_eval 0.5
3. scipy 1.2.1

Usage

Training:

python main.py -m train -c json/config.json

Testing:

python main.py -m test -c models/name/config.json

Configuration

A detailed description of all configurable parameters can be found in json/config.json

Optional command-line arguments:

Argument	Valid Inputs	Default	Description
mode	train/test	training
config	string	config.json	Path to JSON-formatted config file
ckpt	string	None	Path to model's checkpoint. If not specfied, will automatically load the latest checkpoint.

Dataset Preprocess

From SPHERE to wav : bash convert_wsj0.sh

Generate WSJ0-2mix (Wall Street Journal with 2-speaker mixture) or WSJ0-3mix

1. Download official code or use my modified version in create_wav_2speakers.m and create_wav_3speakers.m

2. Download voicebox

3. Steps to run octave on linux:

   (1) run octave-cli

   (2) load package pkg load <pkg-name>

   (3) run create_wav_2speakers.m or create_wav_3speakers.m

Citation

If you find this repo helpful, please kindly cite my paper.

@article{yang2019improved,
  title={Improved Speech Separation with Time-and-Frequency Cross-domain Joint Embedding and Clustering},
  author={Yang, Gene-Ping and Tuan, Chao-I and Lee, Hung-Yi and Lee, Lin-shan},
  journal={arXiv preprint arXiv:1904.07845},
  year={2019}
}