Code to simulate a reverberated, noisy version of the WSJ0-2MIX dataset. Microphones are placed on a linear array with spacing between the devices resembling that of Microsoft Kinect ™, the device used to record the CHiME-5 dataset. This was done so that we could use the real ambient noise captured as part of CHiME-5 dataset. The room impulse responses (RIR) were simulated for a sampling rate of 16,000 Hz.
Run ./create_corrupted_speech.sh --stage 0 --wsj_data_path wsj_path --chime5_wav_base chime_path --dihard_sad_label_path dihard_path --dest save_path
- wsj_path : Path to precomputed wsj-2mix dataset. Should contain the folder 2speakers/wav16k/
- chime_path : Path to chime-5 dataset. Should contain the folders train, dev and eval
- dihard_path : Path to dihard labels. Should contain
*.labfiles for the train and dev set
- stage 0: Download RIR files (~ 7 GB) to save_path, deflate the compressed files and create a softlink to the RIRs in the current folder
- stage 1: Extract chime 5 noise using dihard labels
- stage 2: Create the mixture. See the code to allow parallelization
Creates the following sub-folders in each of tr, tt and cv folders:
- s1 : spatial image of s1 (Reverberated version of s1 speaker)
- s2 : spatial image of s2 (Reverberated version of s2 speaker)
- s1_direct : direct component of s1 at each of the microphones
- s2_direct : direct component of s2 at each of the microphones
- s1_early : Contains only the early reflections (first 50 ms of RIR. See config.py to change the value) of s1 at each of the microphones
- s2_early : Contains only the early reflections (first 50 ms of RIR. See config.py to change the value) of s2 at each of the microphones
- noise : Contains the noise imposed for each mixture
- mix : s1 + s2 + noise
- list.yaml : A yaml file containing the positions and direction of arrival (DOA) for each utterance and speaker
| Dataset type | Per sub-folder | Total * |
|---|---|---|
| Train (tr) | 21G | 168G |
| Validation (cv) | 5.2G | 41.6G |
| Test (tt) | 3.2G | 25.6G |
[*] Combination of mix, s1_early, s2_early, s1_direct, s2_direct and noise.
If you are using this code please cite the following paper:
@inproceedings{sivasankaran2019analyzing,
booktitle = {2020 28th {{European Signal Processing Conference}} ({{EUSIPCO}})},
title={Analyzing the impact of speaker localization errors on speech separation for automatic speech recognition},
author={Sunit Sivasankaran and Emmanuel Vincent and Dominique Fohr},
year={2021},
month = Jan,
}