Baidu Speech to Text

Latest Version: 2.0.2 CHANGELOG

Hi, this project is aim to speech to text by baidu ai service, now, it's good for chinese you can also use it for English.

you can send me email [email protected]

BIG CHANGE 💣

Because I use snowboy to support hot word, so it is only running on Nodejs 6.0 now.

Before you install the module, you must install sox on Mac OS X or Windows, or arecord on Linux and alsa on Linux

Prepare

The project depends mic, so you should install sox(Mac/Windows) OR arecord(Linux) before install this module.

you can install arecord under linux with command, arecord is apart of alsa-utils

$ sudo apt-get update
$ sudo apt-get upgrade
$ sudo apt-get install alsa-base alsa-utils

you can install sox on Mac with brew

brew install sox

If you want have continual ablitilty, you should USE Node.js 6.0 OR LTS Version product, and you can find more detial information with Snowboy

Install

you can use npm

npm install baidu-stt

or you can also use yarn

yarn add baidu-stt

After install the module, you should make a resources folder and copy resources file to the folder

mkdir ./resources
cp ./node_modules/baidu-stt/resources/Alexa.pmdl ./resources

How to use

const Baidu_STT = require('baidu-stt');

const bdstt = new Baidu_STT({
  apiKey: '......',
  secertKey: '......',
  language:'zh',
  recordVoice:false,
  debug:false,
  continual:false,
  modelFile:'./resources/Alexa.pmdl',
  sensitivity:'0.5',
  gain:2.0,
  silenceDelay:2000,
  hotwords:"Alexa",
  voicePath:'./',
  voiceRate:'16000',
  voiceType:'wav'
})

bdstt.init();
bdstt.on('ready', () => {
  console.log('Start listen...');
  bdstt.listen();
});

bdstt.on('start', () => console.log('Please speak something, I\'m listening...'));
bdstt.on('listening', () => console.log('I\'m listen'));
bdstt.on('stop', () => console.log('Copy that'));
bdstt.on('upload', () => console.log('Uploading voice data...'));

bdstt.on('success', data => {
  console.log(data);
});

bdstt.on('fail', err => console.log(err));

HOT WORD

The default hot word is Alexa, for it is the most popular keywords on snowboy, you can train your personal hot word with snowboy website

Becareful, personal model file extesion is .pdml, it is avalibe file type, you can directly use it.

API

Baidu_STT(options)

Return an Baidu_STT object instance

• options JSON containing command line options. Following are valid options:
  • apiKey {string} App api key
  • secretKey {string} App secret key
  • language {stirng} Choose your language, default is 'zh', you can choose 'en' for English
  • continual {string} Continue to recoding after uploading voice data, if true it will be contunal
  • modelFile {string} The hot word detected model file, default is ./resources/Alexa.pmdl, you should copy the model file to ./resources folder
  • hotwords {string} The keyword for wakeing up, default is Alexa
  • sensitivity {string}
  • debug {string} Display capture data
  • recordVoice {boolean} Whether save the voice on your disk, default is false
  • voiceRate {string} The rate of voice record, default is '16000', you can set '8000' or any vaild vaule in sox OR arecord
  • voicePath {string} The path of saved voice, default is ./
  • voiceType {string} The type of recod voice file, default is wav

instance.init()

Request token then emit ready event;

instance.listen()

Set instance listen for your voice;

instance.pause()

When you set contiunal is true, you can use this fucntion pause record sound;

instance.resume()

When you set continual is true and paused, you can use it to resume;

instance.stop()

Stop recording.

Event

start

Start record voice after you call listen()

wakeup

Be waked up by hot word

sleep

After upload voice data, it is going to sleep status

listening

Recoding voice

stop

Stop recoding voice

upload

Upload voice date to service

success

Success for recogniting your voice content, and return the result it is array type, because there may be many result. Normally the first result is better than other. I see

fail

When baidu can not recognting, it will return error message, it's string type.