LexiconCode / Benjamin

Bijan ASR Engine Based-on KalD

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Benjamin

Bijan ASR Engine Based-on KalD Model.

YouTube Demo

Installation

  1. If this is your first time using Benjamin, You need to train a model from your voice. Please follow Train_Linux.md or Train_Windows.md to create one.
  2. Build BaTool. Follow INSTALL.md guide
  • Only use release source code or commit that tagged with stable version. the main branch is in development and may not work.

Features

  • Zero delay output
  • Fast and lightweight
  • Portable and easy to install

Project Strurcture

  • Digits: Example project implementing zero to nine recognition.
  • ENN: Ehsan Neural Network, Used to calculate confidence scores.
  • Nato: NATO task used to create a low vocabulary ASR Model based on CMVN and Delta-Delta Tri-Phone Algorithm.
  • Tools: BaTool project do the realtime detection.

Aknolegdment

This project wouldn't be possible without the help of:

  • M.P. Zanoosi
  • S. Dadashi
  • Dr. E. Akhavan
  • M. Abdollah Zadeh
  • Tiny-DNN
  • OpenFST
  • HTKBook
  • Kaldi
  • rxi-ini

  • To control the PC this project connect with Rebound and uses Linux uInput module.

About

Bijan ASR Engine Based-on KalD


Languages

Language:C++ 55.6%Language:Shell 15.6%Language:Perl 12.7%Language:QML 7.6%Language:Python 4.6%Language:C 1.7%Language:Lua 1.1%Language:QMake 1.0%