/speech_ae

Speech reconstruction and compression with autoencoders

Primary LanguagePython

speech reconstruction with Autoencoders

Speech compression and reconstruction with autoencoders

Steps

0. Get data

Get a lott of speech data and put them in the data/audio_ds folder (It can be in any folder structure)

1. Generate data

--time_per_sample [time per sample]\
--dataset ['mfcc', 'waveform', 'mel_spec']\
--params [params dict]\
--ignore_warning [true / false default- false]

2. Train test split

python train_test_split.py\
--dataset ['mfcc', 'waveform', 'mel_spec']\
--train_pct [train_pct]

3. Calculate dataset stats

python calculate_stats.py --dataset [dataset]

4. Calculate train set stats

python calculate_train_stats.py --dataset [dataset]

5. Train model