ReadMe.md

Simulation set synthesis

1. The simulation rirs

Parameter and path in configuration challenge_rirgenerator.py file

export LD_LIBRARY_PATH=/path/to/ConferencingSpeech2021/simulation/:$LD_LIBRARY_PATH
python ./challenge_rirgenerator.py

Please make sure that your Python version is higher than 3.6. If you use the supplied RIR, this step is not required.

2. Prepare speech and noise sets

2.1 Download the data

speech data:

aishell-1

aishell-3

librispeech-360

VCTK

noise data:

MUSAN

Audioset

2.2 Generate the list file and partition the training and checksum sets

Configure the path to several datasets in ./prepare.sh

bash ./prepare.sh

3. Generate the data

Configure using the supplied parameters

Attention to the data/[dev | train]_[linear|circle]_simu_mix.config . In the config file path should be replaced with the corresponding path. Generate these config file as follow.

# training set of linear mic array
python mix_wav.py --clean_wav_list=./data/train_clean.lst --noise_wav_list=./data/train_noise.lst --rir_wav_list=./data/train_linear_rir.lst --mix_config_path=./data/train_linear_simu_mix.config --save_dir=./data/wavs/train/simu_linear --chunk_len=4 --generate_config=True

# dev set of linear mic array 
python mix_wav.py --clean_wav_list=./data/dev_clean.lst --noise_wav_list=./data/dev_noise.lst --rir_wav_list=./data/dev_linear_rir.lst --mix_config_path=./data/dev_linear_simu_mix.config --save_dir=./data/wavs/dev/simu_linear --chunk_len=6 --generate_config=True

the output of save_dir will include three folder:

mix\ # speech with reverberation and noise
noreverb_ref\ # clean speech with  early reverberation (default 50ms)
reverb_ref\ # clean speech with reverberation