/aps

A personal toolkit for single/multi-channel speech recognition & enhancement & separation.

Primary LanguagePythonApache License 2.0Apache-2.0

APS

License Python-Version codecov pre-commit CI

My open-source toolkit for speech related tasks, e.g., single/multi-channel speech enhancement & separation & recognition. The goal is to simplify the training and evaluation procedure and make it easy and flexible for me to do experiments and verify neural network based methods.

Setup

git clone https://github.com/funcwj/aps
# set up the python environments
# 1) using "pip install -r requirements.txt" or
# 2) create conda enviroments based on requirements.txt (recommended, see docker/Dockerfile)
cd aps && pip install -r requirements.txt # the optional packages are not shown in requirements.txt

For developers (who want to make commits or PRs), continue to run

# set up the git hook scripts
pip install -r requirements-dev.txt && pre-commit install

to setup the development environments. To build C++ sources and demo commands, running

mkdir build && cd build
cmake .. && make -j

Introduction

Acknowledge

The project was started at early 2019 when the author was a master student of the Audio, Speech and Language Processing Group (ASLP) in Northwestern Polytechnical University (NWPU), Xi'an, China. Originally it was used to collect the source code of the experiments that the author did in the past.