humix-sense-speech

speech-to-text and text-to-speech enabling project for Humix

dependencies:

• nodejs >= 4.2

• gcc >= 4.8
  instructions could be found here https://somewideopenspace.wordpress.com/2014/02/28/gcc-4-8-on-raspberry-pi-wheezy/

• packages

  • sudo apt-get install bison
  • sudo apt-get install libasound2-dev
  • sudo apt-get install swig
  • sudo apt-get install python-dev
  • sudo apt-get install mplayer
  • sudo apt-get install flac
  • sudo apt-get install libsndfile1-dev
  • sudo apt-get install libflac++-dev

• nodes modules:

  • npm install nats soap

• natsd

Run the application

Make sure gnatsd server is running and you are on the project's root folder. Then issue the following command:

    node .

Configuration

module.exports = {

  //settings inside options here are optional
  //you may need these settings to fine tune the underlying poketphinx
  options : {
    vad_threshold: '3.5',
    upperf: '1000',
    hmm: './deps/pocketsphinx-5prealpha/model/en-us/en-us',
    //http://www.speech.cs.cmu.edu/tools/lmtool.html use this website
    //to generate lm and dict file based on your keyword-name
    lm: './lm/humix.lm',
    dict: './lm/humix.dic',
    'keyword-name': 'HUMIX', //all capital characters
    samprate: '16000',
    maxwpf: '5',
    topn: '2',
    maxhmmpf: '3000',
    pl_window: '7',
    ds: '2',
    'wav-proc': './voice/interlude/beep2.wav',
    'wav-bye': './voice/interlude/beep1.wav',
  },

  //language setting for stt/tts engine
  lang: 'en', // 'en', 'cht' or 'chs'
  //specify the stt-engine, then use the corresponding credential bellow
  'stt-engine': 'watson', // 'watson' or 'google',
  //specify the tts-engine, then use the corresponding credential bellow
  'tts-engine': 'watson', // 'watson' or 'itri' or 'iflytek'

  //credential of different stt-engines
  stt: {
    watson: {
      username: '<your_username>',
      passwd: '<your_password>'
    },
    google: {
      username: 'xxxxx',
      passwd: '<replace as your appid>'
    }
  },
  
  //credential of different tts-engines
  tts: {
    watson: {
      username: '<your_username>',
      passwd: '<your_password>'
    },
    iflytek: {
      appid: '<app_id>'
    },
    itri: {
      username: '<your_username>',
      passwd: '<your_password>',
      speaker: 'Bruce',
    }
  }
};

STT & TTS credentials

At least, you have to specify which tts-engine and stt-engine you want to use and the corresponding credential for the stt/tts engine that you use. For watson tts/stt service, please see detail information here

Keyword

By default, 'humix' is the keyword to trigger the dialog module to process the speech data. If you want to change the 'humix' keyword, go to lm online tool and upload a text file which contains your own keyword. Then download the *.lm and *.dic. And modify the following settings in config.js to point to these 2 files for your own keyword.

    lm: './lm/humix.lm',
    dict: './lm/humix.dic',

And change the following setting as well:

    'keyword-name': 'HUMIX', //all capital characters