deepvoice_client/lib/data/services/flutter_asr_service.dart

import 'dart:async';
import 'package:flutter/foundation.dart';
import 'package:get/get.dart';
import 'package:speech_to_text/speech_to_text.dart';
import 'package:speech_to_text/speech_recognition_result.dart';
import '../../core/utils/logger.dart';

/// ASR状态枚举
enum AsrState {
  notInitialized, // 未初始化
  initialized,    // 已初始化
  listening,      // 正在监听
  notListening,   // 未监听
  error           // 错误状态
}

/// Flutter ASR服务，负责本地语音转文本功能
class FlutterAsrService extends GetxService {
  // 语音识别实例
  late SpeechToText _speech;
  
  // 当前ASR状态
  final Rx<AsrState> _asrState = AsrState.notInitialized.obs;
  AsrState get asrState => _asrState.value;
  
  // 可用的语音识别语言
  final RxList<LocaleName> _locales = <LocaleName>[].obs;
  List<LocaleName> get locales => _locales;
  
  // 当前选择的语言
  final RxString _currentLocale = ''.obs;
  String get currentLocale => _currentLocale.value;
  
  // 识别结果
  final RxString _lastWords = ''.obs;
  String get lastWords => _lastWords.value;
  
  // 识别准确度
  final RxDouble _confidence = 0.0.obs;
  double get confidence => _confidence.value;
  
  // 是否可用
  final RxBool _isAvailable = false.obs;
  bool get isAvailable => _isAvailable.value;
  
  // 是否正在监听
  bool get isListening => _speech.isListening;
  
  // 识别结果流
  final _recognitionResultsController = StreamController<String>.broadcast();
  Stream<String> get recognitionResults => _recognitionResultsController.stream;
  
  @override
  void onInit() {
    super.onInit();
    _initSpeech();
  }
  
  /// 初始化语音识别
  Future<void> _initSpeech() async {
    _speech = SpeechToText();
    
    try {
      final available = await _speech.initialize(
        onStatus: _onStatusChanged,
        onError: _onErrorListener,
        debugLogging: kDebugMode,
      );
      
      _isAvailable.value = available;
      
      if (available) {
        _asrState.value = AsrState.initialized;
        
        // 获取可用的语言
        final locales = await _speech.locales();
        _locales.assignAll(locales);
        
        // 默认选择系统语言，如果支持的话
        final systemLocale = await _speech.systemLocale();
        if (systemLocale != null) {
          _currentLocale.value = systemLocale.localeId;
        } else if (_locales.isNotEmpty) {
          _currentLocale.value = _locales.first.localeId;
        }
        
        Logger.info('语音识别初始化成功，可用语言数量: ${_locales.length}');
      } else {
        _asrState.value = AsrState.error;
        Logger.error('语音识别不可用');
      }
    } catch (e) {
      _asrState.value = AsrState.error;
      _isAvailable.value = false;
      Logger.error('语音识别初始化失败: $e');
    }
  }
  
  /// 开始监听
  /// 
  /// [listenFor] 监听时长，默认30秒
  /// [pauseFor] 暂停时长，默认3秒，超过此时长无语音输入会自动停止
  /// [partialResults] 是否返回部分结果，默认true
  /// [onDevice] 是否使用设备本地识别，默认false
  /// [listenMode] 监听模式，默认设备默认模式
  Future<bool> startListening({
    Duration listenFor = const Duration(seconds: 30),
    Duration pauseFor = const Duration(seconds: 3),
    bool partialResults = true,
    bool onDevice = false,
    ListenMode listenMode = ListenMode.deviceDefault,
  }) async {
    if (!_isAvailable.value) {
      Logger.error('语音识别不可用，无法开始监听');
      return false;
    }
    
    // 清空上次的识别结果
    _lastWords.value = '';
    _confidence.value = 0.0;
    
    try {
      final started = await _speech.listen(
        onResult: _onSpeechResult,
        listenFor: listenFor,
        pauseFor: pauseFor,
        localeId: _currentLocale.value,
        listenOptions: SpeechListenOptions(
          partialResults: partialResults,
          onDevice: onDevice,
          listenMode: listenMode,
        ),
      );
      
      if (started) {
        _asrState.value = AsrState.listening;
        Logger.info('开始语音识别监听');
      } else {
        Logger.error('开始语音识别监听失败');
      }
      
      return started;
    } catch (e) {
      Logger.error('开始语音识别监听出错: $e');
      return false;
    }
  }
  
  /// 停止监听
  Future<void> stopListening() async {
    if (!_speech.isListening) return;
    
    try {
      await _speech.stop();
      _asrState.value = AsrState.notListening;
      Logger.info('停止语音识别监听');
    } catch (e) {
      Logger.error('停止语音识别监听出错: $e');
    }
  }
  
  /// 取消监听
  Future<void> cancelListening() async {
    if (!_speech.isListening) return;
    
    try {
      await _speech.cancel();
      _lastWords.value = '';
      _confidence.value = 0.0;
      _asrState.value = AsrState.notListening;
      Logger.info('取消语音识别监听');
    } catch (e) {
      Logger.error('取消语音识别监听出错: $e');
    }
  }
  
  /// 设置当前语言
  void setLocale(String localeId) {
    if (_currentLocale.value != localeId) {
      _currentLocale.value = localeId;
      Logger.info('设置语音识别语言: $localeId');
    }
  }
  
  /// 语音识别结果回调
  void _onSpeechResult(SpeechRecognitionResult result) {
    final resultText = result.recognizedWords;
    _lastWords.value = resultText;
    
    if (result.finalResult) {
      // 计算置信度 - 仅在最终结果时有效
      if (result.alternates.isNotEmpty) {
        _confidence.value = result.alternates.first.confidence;
      }
      
      // 发送最终结果到流
      _recognitionResultsController.add(resultText);
      
      Logger.info('语音识别最终结果: "$resultText", 置信度: ${_confidence.value}');
      _asrState.value = AsrState.notListening;
    } else {
      // 部分结果
      Logger.info('语音识别部分结果: "$resultText"');
    }
  }
  
  /// 状态变化回调
  void _onStatusChanged(String status) {
    Logger.info('语音识别状态变化: $status');
    
    switch (status) {
      case 'listening':
        _asrState.value = AsrState.listening;
        break;
      case 'notListening':
        _asrState.value = AsrState.notListening;
        break;
      case 'done':
        _asrState.value = AsrState.notListening;
        break;
    }
  }
  
  /// 错误回调
  void _onErrorListener(dynamic error) {
    Logger.error('语音识别错误: ${error.errorMsg}');
    _asrState.value = AsrState.error;
  }
  
  @override
  void onClose() {
    _speech.cancel();
    _recognitionResultsController.close();
    super.onClose();
  }
}