Speeech

Alexa style's speech recognition with hotword offline support detection for nodejs. Hotword recognition it's offline then you can use an online service streaming like Google Speech or Wit.ai.

Important: This is an Work in Progress

Installation

npm install --save speeech

Requirements

You need to have SoX installed.

Debian and derivatives:

sudo apt-get install sox libsox-fmt-all

Arch Linux:

yaourt sox cblas

MacOS:

brew install sox

Once SoX it's installed you can try with:

rec new-file.wav

Usage

Inside of examples folder you can find templates for each service. Check out the comments for install others dependencies.

Google Speech

const speeech = require("speeech");

const serviceConfig = {
  auth: {
    projectId: "streaming-speech-sample",
    keyFilename: "./keyfile.json"
  },
  streaming: {
    config: {
      encoding: "LINEAR16",
      sampleRateHertz: 16000,
      languageCode: "en-US"
    },
    singleUtterance: true,
    interimResults: true
  }
};

speeech.emit("start", speeech.googleService(serviceConfig));
speeech.on("result", result => console.log("result", result));

You need to download a keyfile from google console.

Wit.ai

const speeech = require("speeech");

const serviceConfig = require("./witkeyfile.json");

speeech.emit("start", speeech.witService(serviceConfig));
speeech.on("result", result => console.log("result", result));

witkeyfile.json:

{
  "witToken": "XXXXXXXXXXXXXXXXXXX"
}

Api.api

Api.ai use a webhook to comunicate with external services. Inside of example you can use apiaiWebhook.js as template for service. For the speech logic you can use googleApiAiExample.js

accukeyfile.json looks like:

{
  "apikey": "XXXXXXXXXXXXXXX",
  "language": "es-cl"
}

apiAikeyfile.json:

{
  "apiAiToken": "XXXXXXXXXXXXXXX"
}