#144 · Primary category: Speech & Audio

Automatic_Speech_Recognition

audio automatic-speech-recognition chinese-speech-recognition cnn data-preprocessing deep-learning end-to-end evaluation feature-vector layer-normalization lstm paper phonemes rnn rnn-encoder-decoder speech-recognition tensorflow timit-dataset

End-to-end Automatic Speech Recognition for Madarian and English in Tensorflow

Project last updated:03/24/23

GitHub Stars

2.8K

Forks

535

Contributors

11

License

MIT

Why we included this project

This repo is a compact end-to-end speech recognition implementation in TensorFlow, written so researchers and students can study how a complete ASR system is assembled. It spans the whole pipeline, from feature extraction for Timit, LibriSpeech, WSJ, and Mandarin corpora to sequence-to-sequence RNN encoders with attention, plus working DeepSpeech2 and layer-normalized RNN variants. Because it handles both English and Mandarin and ships data-preprocessing utilities, it is a solid starting point for learning to train a recognizer from raw audio instead of gluing together black-box cloud APIs. The code is older and tied to a specific TensorFlow version, so treat it as an educational blueprint and benchmark baseline rather than a production service.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category