voxforge_ru_sphinx

ASR models for CMU Sphinx for Russian language, trained on voxforge.org

Overview

The following repository includes:

• model_parameters/* - trained HMM acoustic models
• voxforge_ru.lm.bin - compressed language model trained on ruscorpora ngram counts
• voxforge_ru.dic - dictionary, derived from the language model (generated by russian_g2p and manually verified)

Requirements

To use the following models you should have installed pocketsphinx on your machine.

Usage

Transcribe from file pocketsphinx_continuous -lm voxforge_ru.lm.bin -hmm model_parameters/voxforge_ru.cd_cont_200 -dict voxforge_ru.dic -infile your_wav_file.wav Or transcribe from microphone pocketsphinx_continuous -lm voxforge_ru.lm.bin -hmm model_parameters/voxforge_ru.cd_cont_200 -dict voxforge_ru.dic -inmic yes