feat(example): add speech-to-text-to-redis example

This example fills a redis list with the text picked up by speech to
text. It also demonstrates how to use the on.('data', ... pattern for
speech to text.

Co-Authored-By: John Son <[email protected]>
Co-Authored-By: Sam Couch <[email protected]>

Author: 3 people

examples/speech_to_text_microphone_input/package.json

@@ -12,9 +12,10 @@
   "license": "Apache-2.0",
   "dependencies": {
     "dotenv": "^4.0.0",
+    "ibm-watson": "latest",
+    "ioredis": "latest",
     "line-in": "^0.1.2",
     "mic": "^2.1.1",
-    "ibm-watson": "latest",
     "wav": "^1.0.1"
   }
 }

examples/speech_to_text_microphone_input/transcribe-mic-to-redis.js

@@ -0,0 +1,87 @@
+// Adapted from other examples
+// Copyright (c) 2019 IBM
+//
+// This creates a redis list with the latest words picked up by watson speech to text
+// This is a great starting out point for microservices reacting to live speech data
+// Note: This deletes and populates the redis list multiple times a second, you may wish
+// to optimize that out depending on your use case.
+//
+// IBM Cloud Redis: https://www.ibm.com/cloud/databases-for-redis
+//
+// set environment variables: SPEECH_TO_TEXT_IAM_APIKEY and REDIS_CONN_URL
+
+'use strict';
+var BUFFER_LENGTH = 100;
+
+var REDIS_CONN_URL = process.env.REDIS_CONN_URL;
+var Redis = require('ioredis');
+
+require('dotenv').config({ silent: true }); // optional, handy for local development
+var SpeechToText = require('ibm-watson/speech-to-text/v1');
+// var LineIn = require('line-in'); // the `mic` package also works - it's more flexible but requires a bit more setup
+var mic = require('mic');
+var wav = require('wav');
+
+var speechToText = new SpeechToText({
+  // pick up SPEECH_TO_TEXT_IAM_APIKEY env variable here
+});
+
+// connect to "DB"
+var redis = new Redis(REDIS_CONN_URL);
+
+// init buffer
+redis.del('speech-to-text-buffer');
+for (var i = 0; i < BUFFER_LENGTH; i++) {
+  redis.lpush('speech-to-text-buffer', 'init');
+}
+
+// var lineIn = new LineIn(); // 2-channel 16-bit little-endian signed integer pcm encoded audio @ 44100 Hz
+var micInstance = mic({
+  rate: '48000',
+  channels: '1',
+  debug: false,
+});
+
+var micInputStream = micInstance.getAudioStream();
+
+var wavStream = new wav.Writer({
+  sampleRate: 44100,
+  channels: 2,
+});
+
+var recognizeStream = speechToText.recognizeUsingWebSocket({
+  content_type: 'audio/wav',
+  interim_results: true,
+  inactivity_timeout: -1,
+  objectMode: true,
+  max_alternatives: 1,
+  word_alternatives_threshold: 0.9,
+  speaker_labels: true,
+});
+
+// lineIn.pipe(wavStream);
+micInputStream.pipe(wavStream);
+
+wavStream.pipe(recognizeStream);
+
+// recognizeStream.pipe(process.stdout);
+recognizeStream.on('data', function(message) {
+  console.log(message);
+
+  try {
+    var transcript = message.results[0].alternatives[0].transcript;
+    console.log(transcript);
+    redis.del('speech-to-text-buffer');
+    transcript.split(' ').forEach(function(word) {
+      // console.log(word);
+      // redis.rpop('speech-to-text-buffer');
+      redis.rpush('speech-to-text-buffer', word);
+    });
+  } catch (error) {
+    console.log(error);
+  }
+});
+
+console.log('Recording, press any key to exit');
+
+micInstance.start();