Flutter语音识别插件google_speech的使用

Google Speech插件允许在Flutter应用中使用Google Speech API进行语音识别。该插件通过gRPC实现，并支持流式转录功能。以下是详细的使用步骤和示例代码。

准备工作

在开始使用之前，您需要创建一个Google Cloud账户并启用Speech API。具体步骤可以参考官方文档。

1. 创建服务账号并下载JSON格式的凭证文件。
2. 确保您的项目中已添加了google_speech插件。

dependencies:
  google_speech: ^5.0.0

认证方式

通过服务账号认证

有两种方式可以通过服务账号进行认证：

1. 直接传入JSON文件：

   import 'package:google_speech/speech_client_authenticator.dart';
   import 'dart:io';
   
   final serviceAccount = ServiceAccount.fromFile(File('PATH_TO_FILE'));
   
   final speechToText = SpeechToText.viaServiceAccount(serviceAccount);
   

2. 传入JSON字符串：

   final serviceAccount = ServiceAccount.fromString(r'''{YOUR_JSON_STRING}''');
   
   /// 或者从assets加载数据
   final serviceAccount = ServiceAccount.fromString(
       '${(await rootBundle.loadString('assets/test_service_account.json'))}');
   
   final speechToText = SpeechToText.viaServiceAccount(serviceAccount);
   

通过Token认证（版本 >= 3.0.0）

final speechToText = SpeechToText.viaToken(
    'Bearer',
    '<token-here>',
);

通过第三方认证器认证（版本 >= 3.0.0）

final speechToText = SpeechToText.viaThirdPartyAuthenticator(
    ThirdPartyAuthenticator(
        obtainCredentialsFromThirdParty: () async {
            // 请求API获取token
            final json = await requestCredentialFromMyApi();
            return AccessCredentials.fromJson(json);
        },
    ),
);

转录音频文件

使用recognize方法

1. 定义RecognitionConfig：

   final config = RecognitionConfig(
       encoding: AudioEncoding.LINEAR16,
       model: RecognitionModel.basic,
       enableAutomaticPunctuation: true,
       sampleRateHertz: 16000,
       languageCode: 'en-US');
   

2. 获取音频文件内容：

   Future<List<int>> _getAudioContent(String name) async {
       final directory = await getApplicationDocumentsDirectory();
       final path = directory.path + '/$name';
       return File(path).readAsBytesSync().toList();
   }
   
   final audio = await _getAudioContent('test.wav');
   

3. 发送请求：

   final response = await speechToText.recognize(config, audio);
   

使用streamRecognize方法

1. 定义StreamingRecognitionConfig：

   final streamingConfig = StreamingRecognitionConfig(config: config, interimResults: true);
   

2. 获取音频文件流：

   Future<Stream<List<int>>> _getAudioStream(String name) async {
       final directory = await getApplicationDocumentsDirectory();
       final path = directory.path + '/$name';
       return File(path).openRead();
   }
   
   final audio = await _getAudioStream('test.wav');
   

3. 发送请求：

   final responseStream = speechToText.streamingRecognize(streamingConfig, audio);
   responseStream.listen((data) {
       // 监听响应
   });
   

示例应用

以下是一个完整的Flutter示例应用，演示如何使用google_speech插件进行语音识别：

import 'dart:io';
import 'package:flutter/material.dart';
import 'package:flutter/services.dart';
import 'package:google_speech/google_speech.dart';
import 'package:path_provider/path_provider.dart';

void main() {
  runApp(MyApp());
}

class MyApp extends StatelessWidget {
  [@override](/user/override)
  Widget build(BuildContext context) {
    return MaterialApp(
      title: 'Audio File Example',
      theme: ThemeData(
        primarySwatch: Colors.blue,
        visualDensity: VisualDensity.adaptivePlatformDensity,
      ),
      home: AudioRecognize(),
    );
  }
}

class AudioRecognize extends StatefulWidget {
  [@override](/user/override)
  _AudioRecognizeState createState() => _AudioRecognizeState();
}

class _AudioRecognizeState extends State<AudioRecognize> {
  bool recognizing = false;
  bool recognizeFinished = false;
  String text = '';

  void recognize() async {
    setState(() {
      recognizing = true;
    });

    final serviceAccount = ServiceAccount.fromString(
        '${(await rootBundle.loadString('assets/test_service_account.json'))}');
    final speechToText = SpeechToText.viaServiceAccount(serviceAccount);
    final config = _getConfig();
    final audio = await _getAudioContent('test.wav');

    await speechToText.recognize(config, audio).then((value) {
      setState(() {
        text = value.results.map((e) => e.alternatives.first.transcript).join('\n');
      });
    }).whenComplete(() => setState(() {
          recognizeFinished = true;
          recognizing = false;
        }));
  }

  void streamingRecognize() async {
    setState(() {
      recognizing = true;
    });

    final serviceAccount = ServiceAccount.fromString(
        '${(await rootBundle.loadString('assets/test_service_account.json'))}');
    final speechToText = SpeechToText.viaServiceAccount(serviceAccount);
    final config = _getConfig();

    final responseStream = speechToText.streamingRecognize(
        StreamingRecognitionConfig(config: config, interimResults: true),
        await _getAudioStream('test.wav'));

    responseStream.listen((data) {
      setState(() {
        text = data.results.map((e) => e.alternatives.first.transcript).join('\n');
        recognizeFinished = true;
      });
    }, onDone: () {
      setState(() {
        recognizing = false;
      });
    });
  }

  RecognitionConfig _getConfig() => RecognitionConfig(
        encoding: AudioEncoding.LINEAR16,
        model: RecognitionModel.basic,
        enableAutomaticPunctuation: true,
        sampleRateHertz: 16000,
        languageCode: 'en-US',
      );

  Future<List<int>> _getAudioContent(String name) async {
    final directory = await getApplicationDocumentsDirectory();
    final path = directory.path + '/$name';
    return File(path).readAsBytesSync().toList();
  }

  Future<Stream<List<int>>> _getAudioStream(String name) async {
    final directory = await getApplicationDocumentsDirectory();
    final path = directory.path + '/$name';
    return File(path).openRead();
  }

  [@override](/user/override)
  Widget build(BuildContext context) {
    return Scaffold(
      appBar: AppBar(
        title: Text('Audio File Example'),
      ),
      body: Center(
        child: Column(
          mainAxisAlignment: MainAxisAlignment.spaceAround,
          children: [
            ElevatedButton(
              onPressed: recognize,
              child: Text('Recognize'),
            ),
            ElevatedButton(
              onPressed: streamingRecognize,
              child: Text('Streaming Recognize'),
            ),
            _RecognizeContent(text: text),
          ],
        ),
      ),
    );
  }
}

class _RecognizeContent extends StatelessWidget {
  final String text;

  const _RecognizeContent({Key? key, required this.text}) : super(key: key);

  [@override](/user/override)
  Widget build(BuildContext context) {
    return Padding(
      padding: const EdgeInsets.all(16.0),
      child: Column(
        children: [
          Text('Recognized Text:'),
          Text(text),
        ],
      ),
    );
  }
}