Skip to content

lingualogic/my-speech-listen-en

Repository files navigation

Speech-Angular: Getting Started with ListenService

In this Getting Started we want to create an Angluar app and extend it with the ListenService. The service enables (Automatic) Speech Recognition - ASR within the app.

Installation

To follow this tutorial you will need the Angular CLI. Please install it if you don't have it yet. Afterwards the project dependencies can be installed.

$ npm install -g @angular/cli

Hint: Angular requires Node.js version > 10.x. To check your version, run node -v in a terminal. To get Node.js, go to nodejs.org.

Start

To start the example use this command:

$ ng serve

Open localhost on port 4200 in Chrome to test the speech recognition.

Build

To build a version for production, use:

$ ng build

The build can be found in the dist directory.

Step-by-Step Instruction

Create a new Angular app project and change directory to the workspace.

$ ng new my-speech-listen
$ cd my-speech-listen

To be able to use the ListenService, we install the speech-angular:

$ npm install speech-angular

Let's continue in the code. For programming we can recommend VScode.

$ code .

Provide the ListenService

With the service we will enable (Automatic) Speech Recognition within the app.

In app.module.ts we import the service and make it available as a provider.

import { BrowserModule } from '@angular/platform-browser';
import { FormsModule } from '@angular/forms';
import { NgModule } from '@angular/core';

import { AppComponent } from './app.component';

import { ListenService } from 'speech-angular';    // <== Import ListenService

@NgModule({
    declarations: [
        AppComponent
    ],
    imports: [
        BrowserModule,
        FormsModule
    ],
    providers: [
        ListenService                              // <== Add Provider
    ],
    bootstrap: [AppComponent]
})
export class AppModule { }

Now the service is available in our components.

Start the ListenService

We open app.component.ts and import the ListenService. In the constructor of the AppComponent class we create a ListenService object via Dependency Injection. The functionality of the service can be lookup here.

In the start() function we set the language and start the speech recognition.

import { Component } from '@angular/core';
import { ListenService } from 'speech-angular';               // <== Import ListenService

@Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
})

export class AppComponent {
    title = 'my-speech-listen';

    constructor( private listenService: ListenService ) { }   // <== Dependency Injection

    start(): void {
        this.listenService.language = 'en';                   // <== Set Language
        this.listenService.start();                           // <== Start ListenService
    }
}

Get the result of speech recognition

In order to get the result of the speech recognition, need to subscribe the resultEvent of the ListenService. Therefore we implement the OnInit interface in the AppComponent class, initialize the listenResultEvent and set the callback function in the ngOnInit() function. In the callback the result of the speech recognition gets log in the console.

import { Component, OnInit } from '@angular/core';             // <== Add OnInit
import { ListenService } from 'speech-angular';

@Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
})

export class AppComponent implements OnInit {                  // <== Implement OnInit
    title = 'my-speech-listen';

    listenResultEvent: any;                                    // <== Init Event

    constructor( private listenService: ListenService ) { }

    ngOnInit(): void {
        this.listenResultEvent = this.listenService.resultEvent.subscribe(aText => {
            console.log(aText);                               // <== Here is the Result
        });
    }

    start(): void {
        this.listenService.language = 'en';
        this.listenService.start();
    }
}

Show the result of the speech recognition

To show the result of the speech recognition in the view, we instantiate a string listenResult. In the callbak function of the listenResultEvent we set its value and update the view.

import { Component, OnInit, ChangeDetectorRef } from '@angular/core'; // <== More Imports
import { ListenService } from 'speech-angular';

@Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
})

export class AppComponent implements OnInit {
    title = 'my-speech-listen';

    listenResult: string;                                             // <== Init String

    listenResultEvent: any;

    constructor( private listenService: ListenService,
                 private ref: ChangeDetectorRef ) { }                 // <==

    ngOnInit(): void {
        this.listenResultEvent = this.listenService.resultEvent.subscribe(aText => {
            this.listenResult = aText;                                // <== Set Result
            this.ref.detectChanges();                                 // <== Update View
        });
    }

    start(): void {
        this.listenService.language = 'en';
        this.listenService.start();
    }
}

Edit the view

Subsequently we edit the template file app.component.html. We add an input - to show the listenResult and a button - to call the start() fucntion.

<div style="text-align:center">

<h1>Welcome to {{ title }}!</h1>

<img width="300" alt="Angular Logo" src="data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNTAgMjUwIj4KICAgIDxwYXRoIGZpbGw9IiNERDAwMzEiIGQ9Ik0xMjUgMzBMMzEuOSA2My4ybDE0LjIgMTIzLjFMMTI1IDIzMGw3OC45LTQzLjcgMTQuMi0xMjMuMXoiIC8+CiAgICA8cGF0aCBmaWxsPSIjQzMwMDJGIiBkPSJNMTI1IDMwdjIyLjItLjFWMjMwbDc4LjktNDMuNyAxNC4yLTEyMy4xTDEyNSAzMHoiIC8+CiAgICA8cGF0aCAgZmlsbD0iI0ZGRkZGRiIgZD0iTTEyNSA1Mi4xTDY2LjggMTgyLjZoMjEuN2wxMS43LTI5LjJoNDkuNGwxMS43IDI5LjJIMTgzTDEyNSA1Mi4xem0xNyA4My4zaC0zNGwxNy00MC45IDE3IDQwLjl6IiAvPgogIDwvc3ZnPg==">

<h2>This App can recognize spoken language:</h2>

<!-- Shows the ListenResult -->
<input type="text" [(ngModel)]="listenResult" style="padding: 16px 30px; font-size: 16px;" />

<!-- Start the ListenService -->
<button (click)="start()" style="padding: 16px 32px; font-size: 16px;">Press me!</button>

Start

Now the Angular app can be started.

$ ng serve 

Open localhost on port 4200 in Chrome to test the speech recognition.

Does the app understands you? Send your feedback to info@lingualogic.de.

Extension

Speech recognition using HTML5 and the Web-Audio-API is only supported in Chrome, to enable speech recognition on all browsers and plattforms, you may use a cloud service, like Google Cloud Speech-to-Text.

If you want to analyse the users intention, you need natural language understanding, which is provided in the IntentService.

Enable speech recognition for all browsers

In order to enable speech recognition for all browsers, we add the Google Cloud Speech-to-Text API to our project. Therefore we follow the steps:

  1. Go to the Google Cloud Console, setup a project and activate the Google Cloud Speech-to-Text API.

  2. Set up a server (like the speech-tokenserver) or a FaaS to create client access token. This Example uses the gcf in functions/tokenserver. Follow the instruction and enter the GOOGLE_SERVER_URL in the credentials/google-credentials.ts.

  3. In environment.ts and environment.prod.ts add a google parameter:

     export const environment = {
     	production: false,
     	google: false,
     };
    
  4. In main.ts import google module, set the google-credentials and init the cloud-service:

     import { GoogleModule } from 'speech-angular';
    
     ...
    
     import { GOOGLE_APP_KEY, GOOGLE_SERVER_URL } from './../credentials/google-credentials-my';
     const googleOption = {
     	googleDynamicCredentialsFlag: false,
     	googleAppKey: GOOGLE_APP_KEY,
     	googleServerUrl: GOOGLE_SERVER_URL,
     	errorOutputFlag: true
     };
    
     ...
    
     GoogleModule.init( googleOption, (aGoogleFlag: boolean) => {
     if ( googleOption && googleOption.errorOutputFlag ) {
     	console.log( '===> Google:', aGoogleFlag);
     }
     environment.google = aGoogleFlag;
     });
    
  5. In the app.component.ts we can now shwich the ASR used by the ListenService.

     ngOnInit(): void {
     	this.listenService.asr = 'ASRGoogle';
     	... }
    

Test your app on another browser (than Chrome), it should work!