Flutter语音识别插件vosk_flutter的使用

Flutter语音识别插件vosk_flutter的使用

Flutter插件用于Vosk语音识别。

平台支持

使用

配置

请参考安装页面的说明：

• 安装页面

Android

在android/app/proguard-rules.pro文件中添加以下规则（如果该文件不存在，则创建它）：

-keep class com.sun.jna.* { *; }
-keepclassmembers class * extends com.sun.jna.* { public *; }

如果要使用麦克风输入，请在AndroidManifest.xml中添加麦克风权限：

<uses-permission android:name="android.permission.RECORD_AUDIO" />

加载模型

在pubspec.yaml中添加以下配置：

flutter:
  assets:
    - assets/models/

加载模型：

final vosk = VoskFlutterPlugin.instance();
final enSmallModelPath = await ModelLoader()
    .loadFromAssets('assets/models/vosk-model-small-en-us-0.15.zip');

创建识别器

创建一个识别器实例：

final recognizer = await vosk.createRecognizer(
    model: model,
    sampleRate: sampleRate,
);
final recognizerWithGrammar = await vosk.createRecognizer(
    model: model,
    sampleRate: sampleRate,
    grammar: ['one', 'two', 'three'],
);

识别音频数据

处理音频数据并获取结果：

Uint8List audioBytes = ...; // 音频数据为PCM 16位单声道格式
List<String> results = [];
int chunkSize = 8192;
int pos = 0;

while (pos + chunkSize < audioBytes.length) {
    final resultReady = await recognizer.acceptWaveformBytes(
      Uint8List.fromList(audioBytes.getRange(pos, pos + chunkSize).toList()));
    pos += chunkSize;
    
    if (resultReady) {
      print(await recognizer.getResult());
    } else {
      print(await recognizer.getPartialResult());
    }
}
await recognizer.acceptWaveformBytes(
  Uint8List.fromList(audioBytes.getRange(pos, audioBytes.length).toList()));
print(await recognizer.getFinalResult());

识别麦克风数据

Android

初始化并启动语音服务：

final speechService = await vosk.initSpeechService(recognizer);
speechService.onPartial().forEach((partial) => print(partial));
speechService.onResult().forEach((result) => print(result));
await speechService.start();

Linux & Windows

使用适合的插件获取麦克风输入，并传递给识别器：

// 示例代码省略，具体实现根据所选插件而定

完整示例代码

以下是完整的示例代码，展示了如何使用vosk_flutter插件进行语音识别。

import 'dart:io';

import 'package:flutter/material.dart';
import 'package:record/record.dart';
import 'package:vosk_flutter/vosk_flutter.dart';
import 'package:vosk_flutter_example/test_screen.dart';

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

class MyApp extends StatelessWidget {
  const MyApp({Key? key}) : super(key: key);

  [@override](/user/override)
  Widget build(BuildContext context) {
    return const MaterialApp(
      home: VoskFlutterDemo(),
    );
  }
}

class VoskFlutterDemo extends StatefulWidget {
  const VoskFlutterDemo({Key? key}) : super(key: key);

  [@override](/user/override)
  State<VoskFlutterDemo> createState() => _VoskFlutterDemoState();
}

class _VoskFlutterDemoState extends State<VoskFlutterDemo> {
  static const _textStyle = TextStyle(fontSize: 30, color: Colors.black);
  static const _modelName = 'vosk-model-small-en-us-0.15';
  static const _sampleRate = 16000;

  final _vosk = VoskFlutterPlugin.instance();
  final _modelLoader = ModelLoader();
  final _recorder = Record();

  String? _fileRecognitionResult;
  String? _error;
  Model? _model;
  Recognizer? _recognizer;
  SpeechService? _speechService;

  bool _recognitionStarted = false;

  [@override](/user/override)
  void initState() {
    super.initState();

    _modelLoader
        .loadModelsList()
        .then((modelsList) =>
            modelsList.firstWhere((model) => model.name == _modelName))
        .then((modelDescription) =>
            _modelLoader.loadFromNetwork(modelDescription.url)) // 加载模型
        .then(
            (modelPath) => _vosk.createModel(modelPath)) // 创建模型对象
        .then((model) => setState(() => _model = model))
        .then((_) => _vosk.createRecognizer(
            model: _model!, sampleRate: _sampleRate)) // 创建识别器
        .then((value) => _recognizer = value)
        .then((recognizer) {
      if (Platform.isAndroid) {
        _vosk
            .initSpeechService(_recognizer!) // 初始化语音服务
            .then((speechService) =>
                setState(() => _speechService = speechService))
            .catchError((e) => setState(() => _error = e.toString()));
      }
    }).catchError((e) {
      setState(() => _error = e.toString());
      return null;
    });
  }

  [@override](/user/override)
  Widget build(BuildContext context) {
    if (_error != null) {
      return Scaffold(
          body: Center(child: Text("Error: $_error", style: _textStyle)));
    } else if (_model == null) {
      return const Scaffold(
          body: Center(child: Text("Loading model...", style: _textStyle)));
    } else if (Platform.isAndroid && _speechService == null) {
      return const Scaffold(
        body: Center(
          child: Text("Initializing speech service...", style: _textStyle),
        ),
      );
    } else {
      return Platform.isAndroid ? _androidExample() : _commonExample();
    }
  }

  Widget _androidExample() {
    return Scaffold(
      body: Center(
        child: Column(
          mainAxisAlignment: MainAxisAlignment.center,
          children: [
            ElevatedButton(
                onPressed: () async {
                  if (_recognitionStarted) {
                    await _speechService!.stop();
                  } else {
                    await _speechService!.start();
                  }
                  setState(() => _recognitionStarted = !_recognitionStarted);
                },
                child: Text(_recognitionStarted
                    ? "Stop recognition"
                    : "Start recognition")),
            StreamBuilder(
                stream: _speechService!.onPartial(),
                builder: (context, snapshot) => Text(
                    "Partial result: ${snapshot.data.toString()}",
                    style: _textStyle)),
            StreamBuilder(
                stream: _speechService!.onResult(),
                builder: (context, snapshot) => Text(
                    "Result: ${snapshot.data.toString()}",
                    style: _textStyle)),
          ],
        ),
      ),
    );
  }

  Widget _commonExample() {
    return Scaffold(
      body: Center(
        child: Column(
          mainAxisAlignment: MainAxisAlignment.center,
          children: [
            ElevatedButton(
                onPressed: () async {
                  if (_recognitionStarted) {
                    await _stopRecording();
                  } else {
                    await _recordAudio();
                  }
                  setState(() => _recognitionStarted = !_recognitionStarted);
                },
                child: Text(
                    _recognitionStarted ? "Stop recording" : "Record audio")),
            Text("Final recognition result: $_fileRecognitionResult",
                style: _textStyle),
          ],
        ),
      ),
    );
  }

  Future<void> _recordAudio() async {
    try {
      await _recorder.start(
          samplingRate: 16000, encoder: AudioEncoder.wav, numChannels: 1);
    } catch (e) {
      _error = e.toString() +
          '\n\n Make sure fmedia(https://stsaz.github.io/fmedia/)'
              ' is installed on Linux';
    }
  }

  Future<void> _stopRecording() async {
    try {
      final filePath = await _recorder.stop();
      if (filePath != null) {
        final bytes = File(filePath).readAsBytesSync();
        _recognizer!.acceptWaveformBytes(bytes);
        _fileRecognitionResult = await _recognizer!.getFinalResult();
      }
    } catch (e) {
      _error = e.toString() +
          '\n\n Make sure fmedia(https://stsaz.github.io/fmedia/)'
              ' is installed on Linux';
    }
  }
}