Timit data process for dnn ai-aec (acoustic echo cancellation) experiments

This repo is following the data setup from Deep Learning for Acoustic Echo Cancellation in Noisy and Double-TalkScenarios.

It' a draft script, I will modify it and put all changeable configurations into a json so that it can be used more friendly.

By the way, if you want to do some work in deep learning aec, I recommend using farend data from AEC-challenge and mix with other clean open source datasets.

Notification

References:

Paper: Deep Learning for Acoustic Echo Cancellation in Noisy and Double-TalkScenarios

DNS-CHALLENGE: INTERSPEECH 2021 Deep Noise Suppression Challenge
DNS-CHALLENGE CODE: INTERSPEECH 2021 Deep Noise Suppression Challenge

AEC-CHALLENGE:ICASSP 2021 ACOUSTIC ECHO CANCELLATION CHALLENGE: DATASETS, TESTINGFRAMEWORK, AND RESULTS
AEC-CHALLENGE CODE:ICASSP 2021 ACOUSTIC ECHO CANCELLATION CHALLENGE: DATASETS, TESTINGFRAMEWORK, AND RESULTS

How to use

1. change dataPath, noisePath, outPath and rirPath according to your setups, p.s. rirPath is provided from DNS-CHALLENGE where you can review above

2. python timit_pre_process.py

Last Modification

1. add json
2. randomly pad signal to certain length
3. add non-linear