liwei1dao
/
deepvoice_client
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Browse Source

Merge commit '025ac6b26d6abaeaf9821d451bdcfbd337575673' into new_dev

newdev_shunjiawei
lxm 1 year ago
parent
88c3d14a0a 025ac6b26d
commit
e361215196
15 changed files with 528 additions and 321 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 152
      lib/data/services/language_manager.dart
  2. 42
      lib/modules/settings/views/settings_view.dart
  3. 24
      local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentService.kt
  4. 13
      local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentServicePlugin.kt
  5. 1
      local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/BleAgent.kt
  6. 51
      local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServiceImpl.swift
  7. 4
      local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServicePlugin.swift
  8. 4
      local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrHelper.kt
  9. 141
      local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrToAsr.kt
  10. 46
      local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/tools/SimpleAudioReceiver.kt
  11. 174
      local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/IntegratedSpeechTranslationService.swift
  12. 56
      local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/VolcanoTranslationServiceImpl.swift
  13. 54
      local_plugins/azure_speech/ios/azure_speech/Sources/tools/SimpleAudioReceiver.swift
  14. 63
      local_plugins/ble_service/android/src/main/kotlin/com/yunqiinnovation/ble_service/BleService.kt
  15. 24
      local_plugins/ble_service/ios/ble_service/Sources/ble_service/BleService.swift

152
lib/data/services/language_manager.dart
View File

@ -5,22 +5,22 @@ import '../../core/utils/logger.dart';
class LanguageInfo {
/// 短代码，如 'zh', 'en'
final String shortCode;
/// 中文名称
final String chineseName;
/// 英文名称
final String englishName;
/// Azure ASR 语言代码，如 'zh-CN', 'en-US'
final String asrCode;
/// Azure TTS 语音名称
final String ttsVoiceName;
/// 备注信息
final String? note;
LanguageInfo({
required this.shortCode,
required this.chineseName,
@ -35,30 +35,30 @@ class LanguageInfo {
class LanguageManager extends GetxService {
// 单例实例
static LanguageManager? _instance;
// 单例工厂方法
factory LanguageManager() {
_instance ??= LanguageManager._internal();
return _instance!;
}
// 私有构造函数
LanguageManager._internal() {
_initLanguageMap();
}
// 语言信息映射表，以 ASR 代码为键
final Map<String, LanguageInfo> _languageMap = {};
// 中文名称到 ASR 代码的映射
final Map<String, String> _chineseNameToAsrCodeMap = {};
// 英文名称到 ASR 代码的映射
final Map<String, String> _englishNameToAsrCodeMap = {};
// 短代码到 ASR 代码的映射
final Map<String, String> _shortCodeToAsrCodeMap = {};
/// 初始化语言映射表
void _initLanguageMap() {
// 中文相关
@ -69,7 +69,7 @@ class LanguageManager extends GetxService {
asrCode: 'zh-CN',
ttsVoiceName: 'zh-CN-XiaoxiaoNeural',
);
_addLanguage(
shortCode: 'zh-Hant',
chineseName: '中文(繁体)',
@ -77,7 +77,7 @@ class LanguageManager extends GetxService {
asrCode: 'zh-TW',
ttsVoiceName: 'zh-TW-HsiaoChenNeural',
);
_addLanguage(
shortCode: 'zh-Hant-hk',
chineseName: '中文(香港繁体)',
@ -85,7 +85,7 @@ class LanguageManager extends GetxService {
asrCode: 'zh-HK',
ttsVoiceName: 'zh-HK-HiuMaanNeural',
);
// 英语相关
_addLanguage(
shortCode: 'en',
@ -94,7 +94,7 @@ class LanguageManager extends GetxService {
asrCode: 'en-US',
ttsVoiceName: 'en-US-AriaNeural',
);
// 日语相关
_addLanguage(
shortCode: 'ja',
@ -103,7 +103,7 @@ class LanguageManager extends GetxService {
asrCode: 'ja-JP',
ttsVoiceName: 'ja-JP-NanamiNeural',
);
// 韩语相关
_addLanguage(
shortCode: 'ko',
@ -112,7 +112,7 @@ class LanguageManager extends GetxService {
asrCode: 'ko-KR',
ttsVoiceName: 'ko-KR-SunHiNeural',
);
// 欧洲语言
_addLanguage(
shortCode: 'fr',
@ -121,7 +121,7 @@ class LanguageManager extends GetxService {
asrCode: 'fr-FR',
ttsVoiceName: 'fr-FR-DeniseNeural',
);
_addLanguage(
shortCode: 'es',
chineseName: '西班牙语',
@ -129,7 +129,7 @@ class LanguageManager extends GetxService {
asrCode: 'es-ES',
ttsVoiceName: 'es-ES-ElviraNeural',
);
_addLanguage(
shortCode: 'pt',
chineseName: '葡萄牙语',
@ -137,7 +137,7 @@ class LanguageManager extends GetxService {
asrCode: 'pt-BR',
ttsVoiceName: 'pt-BR-FranciscaNeural',
);
_addLanguage(
shortCode: 'it',
chineseName: '意大利语',
@ -145,7 +145,7 @@ class LanguageManager extends GetxService {
asrCode: 'it-IT',
ttsVoiceName: 'it-IT-ElsaNeural',
);
_addLanguage(
shortCode: 'de',
chineseName: '德语',
@ -153,7 +153,7 @@ class LanguageManager extends GetxService {
asrCode: 'de-DE',
ttsVoiceName: 'de-DE-KatjaNeural',
);
_addLanguage(
shortCode: 'ru',
chineseName: '俄语',
@ -161,7 +161,7 @@ class LanguageManager extends GetxService {
asrCode: 'ru-RU',
ttsVoiceName: 'ru-RU-SvetlanaNeural',
);
_addLanguage(
shortCode: 'pl',
chineseName: '波兰语',
@ -169,7 +169,7 @@ class LanguageManager extends GetxService {
asrCode: 'pl-PL',
ttsVoiceName: 'pl-PL-AgnieszkaNeural',
);
_addLanguage(
shortCode: 'nl',
chineseName: '荷兰语',
@ -177,7 +177,7 @@ class LanguageManager extends GetxService {
asrCode: 'nl-NL',
ttsVoiceName: 'nl-NL-ColetteNeural',
);
_addLanguage(
shortCode: 'sv',
chineseName: '瑞典语',
@ -185,7 +185,7 @@ class LanguageManager extends GetxService {
asrCode: 'sv-SE',
ttsVoiceName: 'sv-SE-SofieNeural',
);
_addLanguage(
shortCode: 'cs',
chineseName: '捷克语',
@ -193,7 +193,7 @@ class LanguageManager extends GetxService {
asrCode: 'cs-CZ',
ttsVoiceName: 'cs-CZ-VlastaNeural',
);
_addLanguage(
shortCode: 'el',
chineseName: '现代希腊语',
@ -201,7 +201,7 @@ class LanguageManager extends GetxService {
asrCode: 'el-GR',
ttsVoiceName: 'el-GR-AthinaNeural',
);
_addLanguage(
shortCode: 'ro',
chineseName: '罗马尼亚语',
@ -209,7 +209,7 @@ class LanguageManager extends GetxService {
asrCode: 'ro-RO',
ttsVoiceName: 'ro-RO-AlinaNeural',
);
_addLanguage(
shortCode: 'hu',
chineseName: '匈牙利语',
@ -217,7 +217,7 @@ class LanguageManager extends GetxService {
asrCode: 'hu-HU',
ttsVoiceName: 'hu-HU-NoemiNeural',
);
_addLanguage(
shortCode: 'uk',
chineseName: '乌克兰语',
@ -225,7 +225,7 @@ class LanguageManager extends GetxService {
asrCode: 'uk-UA',
ttsVoiceName: 'uk-UA-PolinaNeural',
);
_addLanguage(
shortCode: 'da',
chineseName: '丹麦语',
@ -233,7 +233,7 @@ class LanguageManager extends GetxService {
asrCode: 'da-DK',
ttsVoiceName: 'da-DK-ChristelNeural',
);
_addLanguage(
shortCode: 'fi',
chineseName: '芬兰语',
@ -241,7 +241,7 @@ class LanguageManager extends GetxService {
asrCode: 'fi-FI',
ttsVoiceName: 'fi-FI-NooraNeural',
);
_addLanguage(
shortCode: 'no',
chineseName: '挪威语',
@ -249,7 +249,7 @@ class LanguageManager extends GetxService {
asrCode: 'nb-NO',
ttsVoiceName: 'nb-NO-IselinNeural',
);
_addLanguage(
shortCode: 'hr',
chineseName: '克罗地亚语',
@ -257,7 +257,7 @@ class LanguageManager extends GetxService {
asrCode: 'hr-HR',
ttsVoiceName: 'hr-HR-GabrijelaNeural',
);
_addLanguage(
shortCode: 'ca',
chineseName: '加泰隆语',
@ -265,7 +265,7 @@ class LanguageManager extends GetxService {
asrCode: 'ca-ES',
ttsVoiceName: 'ca-ES-JoanaNeural',
);
// 亚洲语言
_addLanguage(
shortCode: 'ar',
@ -274,7 +274,7 @@ class LanguageManager extends GetxService {
asrCode: 'ar-EG',
ttsVoiceName: 'ar-EG-SalmaNeural',
);
_addLanguage(
shortCode: 'th',
chineseName: '泰语',
@ -282,7 +282,7 @@ class LanguageManager extends GetxService {
asrCode: 'th-TH',
ttsVoiceName: 'th-TH-PremwadeeNeural',
);
_addLanguage(
shortCode: 'vi',
chineseName: '越南语',
@ -290,7 +290,7 @@ class LanguageManager extends GetxService {
asrCode: 'vi-VN',
ttsVoiceName: 'vi-VN-HoaiMyNeural',
);
_addLanguage(
shortCode: 'id',
chineseName: '印尼语',
@ -298,7 +298,7 @@ class LanguageManager extends GetxService {
asrCode: 'id-ID',
ttsVoiceName: 'id-ID-GadisNeural',
);
_addLanguage(
shortCode: 'hi',
chineseName: '印地语',
@ -306,7 +306,7 @@ class LanguageManager extends GetxService {
asrCode: 'hi-IN',
ttsVoiceName: 'hi-IN-SwaraNeural',
);
_addLanguage(
shortCode: 'tr',
chineseName: '土耳其语',
@ -314,7 +314,7 @@ class LanguageManager extends GetxService {
asrCode: 'tr-TR',
ttsVoiceName: 'tr-TR-EmelNeural',
);
_addLanguage(
shortCode: 'ms',
chineseName: '马来语',
@ -322,7 +322,7 @@ class LanguageManager extends GetxService {
asrCode: 'ms-MY',
ttsVoiceName: 'ms-MY-YasminNeural',
);
_addLanguage(
shortCode: 'bn',
chineseName: '孟加拉语',
@ -330,7 +330,7 @@ class LanguageManager extends GetxService {
asrCode: 'bn-IN',
ttsVoiceName: 'bn-IN-TanishaaNeural',
);
_addLanguage(
shortCode: 'ta',
chineseName: '泰米尔语',
@ -338,7 +338,7 @@ class LanguageManager extends GetxService {
asrCode: 'ta-IN',
ttsVoiceName: 'ta-IN-PallaviNeural',
);
_addLanguage(
shortCode: 'te',
chineseName: '泰卢固语',
@ -346,7 +346,7 @@ class LanguageManager extends GetxService {
asrCode: 'te-IN',
ttsVoiceName: 'te-IN-ShrutiNeural',
);
_addLanguage(
shortCode: 'mr',
chineseName: '马拉提语',
@ -354,7 +354,7 @@ class LanguageManager extends GetxService {
asrCode: 'mr-IN',
ttsVoiceName: 'mr-IN-AarohiNeural',
);
_addLanguage(
shortCode: 'ur',
chineseName: '乌尔都语',
@ -362,7 +362,7 @@ class LanguageManager extends GetxService {
asrCode: 'ur-IN',
ttsVoiceName: 'ur-IN-GulNeural',
);
_addLanguage(
shortCode: 'fa',
chineseName: '波斯语',
@ -370,7 +370,7 @@ class LanguageManager extends GetxService {
asrCode: 'fa-IR',
ttsVoiceName: 'fa-IR-DilaraNeural',
);
// 非洲语言
_addLanguage(
shortCode: 'sw',
@ -379,10 +379,10 @@ class LanguageManager extends GetxService {
asrCode: 'sw-KE',
ttsVoiceName: 'sw-KE-ZuriNeural',
);
Logger.info('语言管理器初始化完成，共加载 ${_languageMap.length} 种语言');
}
/// 添加语言信息到映射表
void _addLanguage({
required String shortCode,
@ -400,133 +400,133 @@ class LanguageManager extends GetxService {
ttsVoiceName: ttsVoiceName,
note: note,
);
_languageMap[asrCode] = languageInfo;
_chineseNameToAsrCodeMap[chineseName] = asrCode;
_englishNameToAsrCodeMap[englishName] = asrCode;
_shortCodeToAsrCodeMap[shortCode] = asrCode;
}
/// 获取所有支持的语言信息
Map<String, LanguageInfo> getAllLanguages() {
return Map.from(_languageMap);
}
/// 获取所有支持的语言的中文名称
List<String> getAllChineseNames() {
return _chineseNameToAsrCodeMap.keys.toList();
}
/// 获取所有支持的语言的英文名称
List<String> getAllEnglishNames() {
return _englishNameToAsrCodeMap.keys.toList();
}
/// 根据 ASR 代码获取语言信息
LanguageInfo? getLanguageByAsrCode(String asrCode) {
return _languageMap[asrCode];
}
/// 根据中文名称获取语言信息
LanguageInfo? getLanguageByChineseName(String chineseName) {
final asrCode = _chineseNameToAsrCodeMap[chineseName];
if (asrCode == null) return null;
return _languageMap[asrCode];
}
/// 根据英文名称获取语言信息
LanguageInfo? getLanguageByEnglishName(String englishName) {
final asrCode = _englishNameToAsrCodeMap[englishName];
if (asrCode == null) return null;
return _languageMap[asrCode];
}
/// 根据短代码获取语言信息
LanguageInfo? getLanguageByShortCode(String shortCode) {
final asrCode = _shortCodeToAsrCodeMap[shortCode];
if (asrCode == null) return null;
return _languageMap[asrCode];
}
/// 根据中文名称获取 ASR 代码
String? getAsrCodeByChineseName(String chineseName) {
return _chineseNameToAsrCodeMap[chineseName];
}
/// 根据英文名称获取 ASR 代码
String? getAsrCodeByEnglishName(String englishName) {
return _englishNameToAsrCodeMap[englishName];
}
/// 根据短代码获取 ASR 代码
String? getAsrCodeByShortCode(String shortCode) {
return _shortCodeToAsrCodeMap[shortCode];
}
/// 根据 ASR 代码获取中文名称
String? getChineseNameByAsrCode(String asrCode) {
final languageInfo = _languageMap[asrCode];
return languageInfo?.chineseName;
}
/// 根据 ASR 代码获取英文名称
String? getEnglishNameByAsrCode(String asrCode) {
final languageInfo = _languageMap[asrCode];
return languageInfo?.englishName;
}
/// 根据 ASR 代码获取短代码
String? getShortCodeByAsrCode(String asrCode) {
final languageInfo = _languageMap[asrCode];
return languageInfo?.shortCode;
}
/// 根据 ASR 代码获取 TTS 语音名称
String? getTtsVoiceNameByAsrCode(String asrCode) {
final languageInfo = _languageMap[asrCode];
return languageInfo?.ttsVoiceName;
}
/// 根据中文名称获取 TTS 语音名称
String? getTtsVoiceNameByChineseName(String chineseName) {
final asrCode = _chineseNameToAsrCodeMap[chineseName];
if (asrCode == null) return null;
return _languageMap[asrCode]?.ttsVoiceName;
}
/// 根据短代码获取中文名称
String? getChineseNameByShortCode(String shortCode) {
final asrCode = _shortCodeToAsrCodeMap[shortCode];
if (asrCode == null) return null;
return _languageMap[asrCode]?.chineseName;
}
/// 根据短代码获取英文名称
String? getEnglishNameByShortCode(String shortCode) {
final asrCode = _shortCodeToAsrCodeMap[shortCode];
if (asrCode == null) return null;
return _languageMap[asrCode]?.englishName;
}
/// 根据短代码获取 TTS 语音名称
String? getTtsVoiceNameByShortCode(String shortCode) {
final asrCode = _shortCodeToAsrCodeMap[shortCode];
if (asrCode == null) return null;
return _languageMap[asrCode]?.ttsVoiceName;
}
/// 获取中文名称到 ASR 代码的映射
Map<String, String> getChineseNameToAsrCodeMap() {
return Map.from(_chineseNameToAsrCodeMap);
}
/// 获取英文名称到 ASR 代码的映射
Map<String, String> getEnglishNameToAsrCodeMap() {
return Map.from(_englishNameToAsrCodeMap);
}
/// 获取短代码到 ASR 代码的映射
Map<String, String> getShortCodeToAsrCodeMap() {
return Map.from(_shortCodeToAsrCodeMap);
}
}
}

42
lib/modules/settings/views/settings_view.dart
View File

@ -515,27 +515,27 @@ class SettingsView extends GetView<SettingsController> {
// ? Colors.white.withOpacity(0.1)
// : Colors.grey[200]),
//BLE测试
// _buildSimpleNavigationSetting(
// title: 'opus解码测试',
// subtitle: '测试opus解码',
// icon: Icons.bluetooth_searching,
// iconBgColor: isDarkMode
// ? Colors.green[900]!.withOpacity(0.3)
// : Colors.green[100]!,
// iconColor:
// isDarkMode ? Colors.green[300]! : Colors.green[600]!,
// titleColor: isDarkMode ? Colors.white : null,
// subtitleColor: isDarkMode ? Colors.white70 : null,
// onTap: () {
// Get.toNamed(Routes.opusTest);
// },
// isDarkMode: isDarkMode,
// ),
// Divider(
// height: 1,
// color: isDarkMode
// ? Colors.white.withOpacity(0.1)
// : Colors.grey[200]),
_buildSimpleNavigationSetting(
title: 'opus解码测试',
subtitle: '测试opus解码',
icon: Icons.bluetooth_searching,
iconBgColor: isDarkMode
? Colors.green[900]!.withOpacity(0.3)
: Colors.green[100]!,
iconColor:
isDarkMode ? Colors.green[300]! : Colors.green[600]!,
titleColor: isDarkMode ? Colors.white : null,
subtitleColor: isDarkMode ? Colors.white70 : null,
onTap: () {
Get.toNamed(Routes.opusTest);
},
isDarkMode: isDarkMode,
),
Divider(
height: 1,
color: isDarkMode
? Colors.white.withOpacity(0.1)
: Colors.grey[200]),
// BLE测试
// _buildSimpleNavigationSetting(

24
local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentService.kt
View File

@ -130,6 +130,8 @@ object AgentService : CoroutineScope {
//是否播放音乐中
private var isplaymusic = false;
//是否保留最终结果
private var isKeepResult = true;
// 打断防抖相关
private var lastInterruptTime = 0L
@ -265,6 +267,7 @@ object AgentService : CoroutineScope {
*/
fun dispose() {
try {
BleService.closeCodec()
stopRecognition()
stopTts()
stopAiStream()
@ -474,6 +477,7 @@ object AgentService : CoroutineScope {
if (isActive && isRecognitionActive && !isTtsSpeaking && !isAiStreaming) {
Log.d(TAG, "空闲检测触发自动停止，模式: $currentRecognitionMode")
BleService.closeCodec()
stopRecognition()
sendEvent(
"auto_stop",
@ -516,13 +520,18 @@ object AgentService : CoroutineScope {
* @param mode 识别模式：normal, ble_wakeup, phone_call, push_to_talk
*/
fun startRecognition(isExternalActive: Boolean, mode: String = "normal"): Boolean {
if(isExternalActive){
isKeepResult = false
BleService.openEncoder()//不发指令。只是录制opus
}
if (!isInitialized) {
Log.d(TAG, "服务未初始化")
return false
}
Log.d("TAG", "开始语音识别startRecognition?")
stopRecognition()
BleService.openEncoder()
// if (isRecognitionActive&&audioSourceType==AzureAsrHelper.AudioSourceType.MICROPHONE){
// Log.d(TAG, "开始语音识别stopRecognition")
// // BleService.closeCodec()
@ -559,6 +568,9 @@ object AgentService : CoroutineScope {
AzureAsrHelper.AudioSourceType.MICROPHONE
}
Log.d("TAG", "azureAsrHelper?.startContinuousRecognition")
if(isExternalActive){
isKeepResult = true
}
azureAsrHelper?.startContinuousRecognition(audioSourceType)
// 根据模式启动相应的空闲检测
@ -586,7 +598,7 @@ object AgentService : CoroutineScope {
val success = azureAsrHelper?.setupEventListeners(object :
AzureAsrHelper.ContinuousRecognizeCallback {
override fun onRecognizing(recognizing: String, detectedLanguage: String) {
if (recognizing.isNotEmpty()) {
if (recognizing.isNotEmpty()&&isKeepResult==true) {
// 检测到语音，更新状态
_hasSpeechDetected.set(true)
@ -630,7 +642,7 @@ object AgentService : CoroutineScope {
}
override fun onResult(result: String, detectedLanguage: String) {
if (result.isNotEmpty()) {
if (result.isNotEmpty()&&isKeepResult==true) {//加入是否结束要输出最后的结果判读
sendEvent(
"recognition_result", mapOf(
"text" to result,
@ -709,11 +721,9 @@ object AgentService : CoroutineScope {
Log.d(TAG, "停止语音识别，当前模式: $currentRecognitionMode")
try {
_isRecognitionActive.set(false)
azureAsrHelper?.stopContinuousRecognition()
BleService.closeCodec()
_isRecognitionActive.set(false)
stopIdleCheck()
// 重置识别模式
currentRecognitionMode = "normal"
return
@ -721,7 +731,6 @@ object AgentService : CoroutineScope {
Log.e(TAG, "停止语音识别异常: ${e.message}")
_isRecognitionActive.set(false)
stopIdleCheck()
// 重置识别模式
currentRecognitionMode = "normal"
return
@ -1055,6 +1064,7 @@ object AgentService : CoroutineScope {
)
)
if (name == "exit_interaction") {
BleService.closeCodec()
stopRecognition()
}
} catch (e: Exception) {

13
local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentServicePlugin.kt
View File

@ -14,6 +14,8 @@ import kotlinx.coroutines.MainScope
import kotlinx.coroutines.launch
import android.content.Intent
import android.os.Build
// 添加BleService的导入语句
import com.yunqiinnovation.ble_service.BleService
class AgentServicePlugin : FlutterPlugin, MethodCallHandler, EventChannel.StreamHandler,
AgentServiceListener {
@ -124,13 +126,14 @@ class AgentServicePlugin : FlutterPlugin, MethodCallHandler, EventChannel.Stream
result.error("START_ERROR", "启动对话失败: ${e.message}", null)
}
}
"recognizeCallback" -> {
val isStarted = AgentService.recognizeCallback()
result.success(isStarted)
}
"recognizeCallback" -> {
val isStarted = AgentService.recognizeCallback()
result.success(isStarted)
}
"stopConversation" -> {
try {
BleService.closeCodec()
AgentService.stopRecognition()
result.success(true)
} catch (e: Exception) {
@ -309,4 +312,4 @@ class AgentServicePlugin : FlutterPlugin, MethodCallHandler, EventChannel.Stream
eventChannel.setStreamHandler(null)
AgentService.removeListener(this)
}
}
}

1
local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/BleAgent.kt
View File

@ -77,6 +77,7 @@ Log.d(TAG, "手动启动语音识别: ")
Log.e(TAG, "语音代理未初始化")
return
}
BleService.closeCodec()
AgentService.stopRecognition()
return
}

51
local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServiceImpl.swift
View File

@ -69,7 +69,9 @@ class AgentServiceImpl: NSObject {
// 用途：解决 start 异步、stop 先于 onSessionStarted 到来时被忽略的问题
private var isStartingRecognition: Bool = false
private var stopRequestedDuringStart: Bool = false
//是否保留最终结果
private var isKeepResult = true;
var currsessionId: String = ""
var language: String = ""
private var serverurl = ""
@ -354,7 +356,8 @@ class AgentServiceImpl: NSObject {
if self.isRecognizing && !self.isSpeaking && !self.isAiStreaming {
os_log("空闲检测触发自动停止，模式: %{public}@", log: self.logger, type: .info, self.currentRecognitionMode)
BleService.shared.closeCodec()
// 关闭编码器 (设备侧)
BleService.shared.closeCodec()
self.stopRecognition()
self.sendEvent(name: "auto_stop", data: [
"reason": "idle_timeout",
@ -378,6 +381,14 @@ class AgentServiceImpl: NSObject {
}
}
func startRecognition(useBle: Bool = false, mode: String = "normal") -> Bool {
// 接收到唤醒信号，打开编码器 (设备侧)
print("ai启动语音\(useBle))")
var openResult = false // 添加openResult变量定义
if useBle{
isKeepResult = false
// 捕获openEncoder的返回值
openResult = BleService.shared.openEncoder() // 正确捕获返回值
}
if !isInitialized {
sendError("服务未初始化", code: "NOT_INITIALIZED")
return false
@ -388,13 +399,12 @@ class AgentServiceImpl: NSObject {
sendError("无麦克风权限", code: "PERMISSION_DENIED")
return false
}
}
stopTts();
if isRecognizing {
stopRecognition()
}
}
stopRecognition()
stopTts()
os_log("🎙️ 尝试打开编码器结果: %@", log: logger, type: .info, openResult ? "成功" : "失败")
wasMusicPlayingBeforeRecognition = MusicService.shared.isPlaying
// 如果音乐正在播放，则暂停
@ -416,7 +426,7 @@ class AgentServiceImpl: NSObject {
// 新增：标记为"启动中"，用于允许 stop 在启动未完成时也能生效
isStartingRecognition = true
print("ai启动语音=\(audioSourceType)=\(useBle)")
guard let success = azureAsrHelper?.startContinuousRecognition(
// callback: self,
@ -427,7 +437,9 @@ class AgentServiceImpl: NSObject {
sendError("启动语音识别失败", code: "RECOGNITION_START_ERROR")
return false
}
if(useBle){
isKeepResult = true
}
// 根据模式启动相应的空闲检测（此处 guard isRecognizing，会在 onSessionStarted 中再次启动）
startIdleCheckForMode(mode: mode)
@ -449,6 +461,7 @@ class AgentServiceImpl: NSObject {
/// 特性：
/// 1) 即便处于"启动中"（isStartingRecognition）也会打印日志并尝试停止
/// 2) 如果底层尚未完全进入可停止状态，记录一次待停止请求，待会话启动后立即停止
print("停止识别语音,isRecognizing=\(isRecognizing),isStartingRecognition=\(isStartingRecognition)")
if !(isRecognizing || isStartingRecognition) {
// 既不在识别中，也不在启动中，视为幂等处理
return true
@ -472,7 +485,6 @@ class AgentServiceImpl: NSObject {
MusicService.shared.resume()
}
// 重置识别模式
currentRecognitionMode = "normal"
@ -490,7 +502,7 @@ class AgentServiceImpl: NSObject {
return false
}
guard let audioStream = azureAsrHelper?.audioStream else {
os_log("音频流未初始化", log: logger,type: .error)
os_log("AI音频流未初始化", log: logger,type: .error)
return false
}
audioStream.saveAudioDataTo(data: audioData)
@ -1053,6 +1065,8 @@ private func jsonToString(_ json: [String: Any]) -> String? {
// 如果正在识别，需要先停止
let wasRecognizing = isRecognizing
if wasRecognizing {
// 关闭编码器 (设备侧)
BleService.shared.closeCodec()
_ = stopRecognition()
}
@ -1227,6 +1241,8 @@ private func jsonToString(_ json: [String: Any]) -> String? {
func dispose() -> Bool {
print("ai释放资源")
if isRecognizing {
// 关闭编码器 (设备侧)
BleService.shared.closeCodec()
stopRecognition()
}
@ -1388,6 +1404,8 @@ class ChatApiStreamCallback: StreamCallback {
print("onFunctionCall=functionCall\(functionCall)")
if (name == "exit_interaction") {
// 关闭编码器 (设备侧)
BleService.shared.closeCodec()
agentService.stopRecognition()
}
@ -1513,9 +1531,6 @@ extension AgentServiceImpl: BleService.Callback {
func onWakeupSignalReceived() {
stopTts()
stopRecognition()
startRecognition(useBle: true, mode: "ble_wakeup")
}
@ -1715,7 +1730,7 @@ class AudioPlayer {
extension AgentServiceImpl: AzureAsrHelper.ContinuousRecognizeCallback {
func onResult(_ text: String, _ detectedLanguage: String) {
if !text.isEmpty {
if !text.isEmpty && isKeepResult {
var data: [String: Any] = ["text": text]
if !detectedLanguage.isEmpty {
data["language"] = detectedLanguage
@ -1734,7 +1749,7 @@ extension AgentServiceImpl: AzureAsrHelper.ContinuousRecognizeCallback {
}
func onRecognizing(_ recognizing: String, _ detectedLanguage: String) {
if !recognizing.isEmpty {
if !recognizing.isEmpty && isKeepResult {
let previousHasSpeech = hasSpeechDetected
hasSpeechDetected = true
@ -1770,6 +1785,8 @@ extension AgentServiceImpl: AzureAsrHelper.ContinuousRecognizeCallback {
// 新增：若在启动期间已经收到停止请求，则立即执行停止
if stopRequestedDuringStart {
stopRequestedDuringStart = false
// 关闭编码器 (设备侧)
BleService.shared.closeCodec()
_ = stopRecognition()
}
}

4
local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServicePlugin.swift
View File

@ -2,7 +2,7 @@ import Flutter
import UIKit
import Foundation
import os.log
import ble_service
/// 代理服务Flutter插件，负责将Flutter调用桥接到原生实现
public class AgentServicePlugin: NSObject, FlutterPlugin {
@ -90,6 +90,8 @@ public class AgentServicePlugin: NSObject, FlutterPlugin {
return
}
print("stopConversation")
// 关闭编码器 (设备侧)
BleService.shared.closeCodec()
impl.stopRecognition()
result(true)

4
local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrHelper.kt
View File

@ -215,7 +215,7 @@ class AzureAsrHelper(private val context: Context) {
if (!audioStream.isInitialized()) {
audioStream.initAudioRecord()
}
audioStream.setupStream()
audioConfig = AudioConfig.fromStreamInput(audioStream.pushAudioStream)
} catch (e: Exception) {
@ -488,7 +488,7 @@ class AzureAsrHelper(private val context: Context) {
if (!audioStream.isInitialized()) {
audioStream.initAudioRecord()
}
audioStream.setupStream()
if (audioStream.recordfile == null) {
audioStream.recordfile = RecordFile()
}

141
local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrToAsr.kt
View File

@ -475,26 +475,57 @@ class IntegratedSpeechTranslationService(
}
/**
* 停止连续语音翻译
* 停止连续语音翻译（非阻塞）
* 在 IO 线程内顺序停止识别器、录音线程与文件，并对翻译/合成收尾增加统一超时保护，避免阻塞主线程与潜在 ANR。
*/
fun stopContinuousTranslation() {
try {
recognizer?.stopContinuousRecognitionAsync()
audioProcessor?.stopRecording()
recordfile1!!.closeFile(true)
Log.d(TAG, "停止连续语音翻译-开始")
// 先标记识别关闭，避免后续流程继续入队处理
serviceState.isRecognizing.set(false)
// 等待当前处理完成
while (serviceState.isTranslating.get() || serviceState.isSynthesizing.get()) {
Thread.sleep(100)
}
Log.d(TAG, "停止连续语音翻译")
// 在 IO 线程中做真实的停止动作，避免阻塞主线程
launch(Dispatchers.IO) {
// 1) 停止识别器：等待短超时，避免卡住
try {
recognizer?.stopContinuousRecognitionAsync()
?.get(1500, java.util.concurrent.TimeUnit.MILLISECONDS)
} catch (t: Throwable) {
Log.w(TAG, "停止识别器超时或失败", t)
}
// 2) 停止录音线程（打断并短等待退出）
runCatching { audioProcessor?.stopRecording() }
.onFailure { Log.w(TAG, "停止录音失败", it) }
// 3) 关闭录音文件（若开启过录音）
runCatching { recordfile1?.closeFile(true) }
.onFailure { Log.w(TAG, "关闭录音文件失败", it) }
// 4) 等待翻译/合成收尾，整体不超过 2s
val start = System.currentTimeMillis()
while ((serviceState.isTranslating.get() || serviceState.isSynthesizing.get())
&& System.currentTimeMillis() - start < 2000
) {
delay(100)
}
if (serviceState.isTranslating.get() || serviceState.isSynthesizing.get()) {
Log.w(TAG, "等待翻译/合成完成超时，强制标记结束")
}
// 5) 统一落幕：确保状态关闭
serviceState.isTranslating.set(false)
serviceState.isSynthesizing.set(false)
Log.d(TAG, "停止连续语音翻译-完成")
}
} catch (e: Exception) {
Log.e(TAG, "停止翻译失败", e)
eventCallback?.onError("Service", "停止失败: ${e.message}")
}
}
/**
* 处理翻译和合成流程
*/
@ -1184,58 +1215,57 @@ fun getSupportedLanguages(): List<String> {
}
/**
* 改进的资源释放方法
* 安全释放音频处理器资源
* 添加更强的异常处理和超时保护
*/
fun dispose() {
try {
Log.d(TAG, "开始释放音频处理器资源...")
// 1. 停止运行标志
isRunning.set(false)
// 1. 设置停止标志
if (!isRunning.compareAndSet(true, false)) {
Log.d(TAG, "音频处理器已停止，跳过释放")
return
}
// 2. 中断处理线程
processingThread?.let { thread ->
if (thread.isAlive) {
// 1) 安全停止处理线程
processingThread?.let { thread ->
if (thread.isAlive) {
try {
thread.interrupt()
try {
// 等待线程结束，最多等待2秒
thread.join(2000)
if (thread.isAlive) {
Log.w(TAG, "音频处理线程未能正常结束")
} else {
Log.d(TAG, "音频处理线程已结束")
}
} catch (e: InterruptedException) {
Log.w(TAG, "等待音频处理线程结束时被中断")
Thread.currentThread().interrupt()
thread.join(500)
if (thread.isAlive) {
Log.w(TAG, "音频处理线程未在500ms内结束")
// 不强制杀死，交由系统回收，但继续清理资源
} else {
Log.d(TAG, "音频处理线程已正常结束")
}
} catch (e: InterruptedException) {
Log.w(TAG, "等待线程结束时被中断", e)
Thread.currentThread().interrupt()
} catch (e: Exception) {
Log.e(TAG, "停止处理线程时发生异常", e)
}
}
processingThread = null
// 3. 关闭音频流
try {
pushAudioStream?.close()
pushAudioStream = null
Log.d(TAG, "音频流已关闭")
} catch (e: Exception) {
Log.e(TAG, "关闭音频流失败", e)
}
// 4. 清理音频队列
try {
val remainingData = audioQueue.size
audioQueue.clear()
Log.d(TAG, "音频队列已清理，丢弃 $remainingData 个数据包")
} catch (e: Exception) {
Log.e(TAG, "清理音频队列失败", e)
}
Log.d(TAG, "音频处理器资源释放完成")
} catch (e: Exception) {
Log.e(TAG, "释放音频处理器资源时发生异常", e)
}
processingThread = null
// 2) 关闭输入流
try {
pushAudioStream?.close()
} catch (t: Throwable) {
Log.w(TAG, "关闭 PushAudioInputStream 时异常", t)
} finally {
pushAudioStream = null
}
// 3) 清空队列
audioQueue.clear()
Log.d(TAG, "音频处理器资源释放完成")
} catch (e: Exception) {
Log.e(TAG, "释放音频处理器资源失败", e)
}
}
}
}
@ -1315,7 +1345,16 @@ class VolcanoTranslationServiceImpl :
"ms-MY" to "ms", // 马来语（马来西亚）
"id-ID" to "id", // 印尼语（印尼）
"fil-PH" to "tl", // 菲律宾语（菲律宾）
"sv-SE" to "sv", // 瑞典语（瑞典）
"ro-RO" to "ro", // 罗马尼亚语（罗马尼亚）
"hu-HU" to "hu", // 匈牙利语（匈牙利）
"da-DK" to "da", // 丹麦语（丹麦）
"fi-FI" to "fi", // 芬兰语（芬兰）
"nb-NO" to "nb", // 挪威语（挪威）
"hr-HR" to "hr", // 克罗地亚语（克罗地亚）
"ca-ES" to "ca", // 加泰隆语
"mr-IN" to "mr", // 马拉地语（印度）
"ur-IN" to "ur", // 乌尔都语（印度）
// 欧洲
"en-US" to "en", // 英语（美国）
"en-GB" to "en", // 英语（英国）

46
local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/tools/SimpleAudioReceiver.kt
View File

@ -88,8 +88,6 @@ class SimpleAudioReceiver(private val context: Context) {
* 初始化音频录制
*/
fun initAudioRecord() {
val format = getOptimalAudioFormat()
pushAudioStream = AudioInputStream.createPushStream(format)
isRunning.set(true)
Log.d(TAG, "initAudioRecord: ${isRunning}")
startWriteThread() // 再启动数据读取线程
@ -97,6 +95,17 @@ class SimpleAudioReceiver(private val context: Context) {
audioManager = context.getSystemService(Context.AUDIO_SERVICE) as AudioManager
originalAudioMode = audioManager?.mode ?: AudioManager.MODE_NORMAL
isInitialized = true
}
/**
* 设置音频流
*/
fun setupStream() {
// 安全关闭并清理之前的流
pushAudioStream?.close()
pushAudioStream = null
val format = getOptimalAudioFormat()
pushAudioStream = AudioInputStream.createPushStream(format)
}
fun isInitialized(): Boolean = isInitialized
/**
@ -143,7 +152,7 @@ class SimpleAudioReceiver(private val context: Context) {
if (bytesToWrite < data.size) data.copyOf(bytesToWrite) else data
try {
Log.d(TAG, "写入数据大小: ${finalData.size},isContinuousRecognitionActive:${isContinuousRecognitionActive},isRecord${isRecord}")
//Log.d(TAG, "写入数据大小: ${finalData.size},isContinuousRecognitionActive:${isContinuousRecognitionActive},isRecord${isRecord}")
if(audioDataCallback!=null){
audioDataCallback?.onAudio(finalData)
}
@ -183,22 +192,36 @@ class SimpleAudioReceiver(private val context: Context) {
writeQueue.offer(buffer.copyOf())
}
/**
* 开始音频输入
* @param audioSourceType 音频源类型
* @param callback 音频数据回调，可为空
*/
fun startAudioRecord(audioSourceType: AudioSourceType, callback: AudioDataCallback?) {
/**
* 开始音频输入
* @param audioSourceType 音频源类型
* @param callback 音频数据回调，可为空
*/
fun startAudioRecord(audioSourceType: AudioSourceType, callback: AudioDataCallback?) {
try {
this.audioSourceType = audioSourceType
this.audioDataCallback = callback
writeQueue.clear()
this.audioDataCallback = callback
isWriting.set(true)
if(writeQueue.isNotEmpty()){
Log.d("tag", "存下来的数据: ${writeQueue?.size}")
}
when (audioSourceType) {
AudioSourceType.MICROPHONE -> runMicrophoneCapture()
AudioSourceType.EXTERNAL -> runExternalCapture()
}
} catch (e: Exception) {
Log.e(TAG, "启动音频录制失败: ${e.message}", e)
// 清理资源
pushAudioStream?.close()
pushAudioStream = null
isWriting.set(false)
throw e // 重新抛出异常让调用者处理
}
}
// ... existing code ...
/**
* 配置通话音频模式以优化回声消除
*/
@ -367,6 +390,7 @@ class SimpleAudioReceiver(private val context: Context) {
*/
fun stopMicrophoneCapture() {
try {
writeQueue.clear()
isWriting.set(false)
// 停止录音

174
local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/IntegratedSpeechTranslationService.swift
View File

@ -605,13 +605,36 @@ import os.log
audioProcessor?.stopMicrophoneCapture()
recordFile?.closeFile(isSave: true)
// 等待当前处理完成
while serviceState.isTranslating || serviceState.isSynthesizing {
Thread.sleep(forTimeInterval: 0.1)
// 使用后台队列等待当前处理完成，避免阻塞主线程
DispatchQueue.global(qos: .userInitiated).async { [weak self] in
guard let self = self else { return }
// 设置超时时间，避免无限等待
let timeout = DispatchTime.now() + .seconds(5)
while (self.serviceState.isTranslating || self.serviceState.isSynthesizing) && DispatchTime.now() < timeout {
Thread.sleep(forTimeInterval: 0.1)
}
// 强制停止合成器（如果仍在运行）
if self.serviceState.isSynthesizing {
do {
try self.synthesizer?.stopSpeaking()
} catch {
os_log("停止语音合成失败: %@", log: self.log, type: .error, error.localizedDescription)
}
self.serviceState.isSynthesizing = false
DispatchQueue.main.async {
self.eventCallback?.onStateChanged(component: "Synthesis", isActive: false)
}
}
DispatchQueue.main.async {
os_log("停止连续语音翻译", log: self.log, type: .info)
}
}
os_log("停止连续语音翻译", log: log, type: .info)
} catch {
os_log("停止翻译失败: %@", log: log, type: .error, error.localizedDescription)
eventCallback?.onError(component: "Service", error: "停止失败: \(error.localizedDescription)")
@ -777,68 +800,93 @@ import os.log
let voice: String
switch language {
// 亚洲地区
case "zh-CN":
voice = "zh-CN-XiaoxiaoNeural"
case "zh-TW":
voice = "zh-TW-HsiaoNeural"
case "zh-HK":
voice = "zh-HK-HiuGaaiNeural"
case "ja-JP":
voice = "ja-JP-NanamiNeural"
case "ko-KR":
voice = "ko-KR-SunHiNeural"
case "hi-IN":
voice = "hi-IN-KalpanaNeural"
case "ta-IN":
voice = "ta-IN-SundarNeural"
case "th-TH":
voice = "th-TH-PattaraNeural"
case "vi-VN":
voice = "vi-VN-AnhNeural"
// 欧洲地区
case "en-US":
voice = "en-US-AriaNeural"
case "en-GB":
voice = "en-GB-SusanNeural"
case "fr-FR":
voice = "fr-FR-DeniseNeural"
case "de-DE":
voice = "de-DE-KatjaNeural"
case "es-ES":
voice = "es-ES-ElviraNeural"
case "it-IT":
voice = "it-IT-DonatellaNeural"
case "pl-PL":
voice = "pl-PL-JozefinaNeural"
case "ru-RU":
voice = "ru-RU-DariaNeural"
// 中文相关
case "zh-CN":
voice = "zh-CN-XiaoxiaoNeural"
case "zh-TW":
voice = "zh-TW-HsiaoChenNeural"
case "zh-HK":
voice = "zh-HK-HiuMaanNeural"
// 中东/非洲
case "ar-SA":
voice = "ar-SA-MajdNeural"
case "he-IL":
voice = "he-IL-NoaNeural"
case "sw-KE":
voice = "sw-KE-SabinaNeural"
// 英语相关
case "en-US":
voice = "en-US-AriaNeural"
// 美洲/大洋洲
case "pt-BR":
voice = "pt-BR-ManuelaNeural"
case "es-MX":
voice = "es-MX-CarlaNeural"
// 亚洲语言
case "ja-JP":
voice = "ja-JP-NanamiNeural"
case "ko-KR":
voice = "ko-KR-SunHiNeural"
case "th-TH":
voice = "th-TH-PremwadeeNeural"
case "vi-VN":
voice = "vi-VN-HoaiMyNeural"
case "id-ID":
voice = "id-ID-GadisNeural"
case "hi-IN":
voice = "hi-IN-SwaraNeural"
case "ms-MY":
voice = "ms-MY-YasminNeural"
case "bn-IN":
voice = "bn-IN-TanishaaNeural"
case "ta-IN":
voice = "ta-IN-PallaviNeural"
case "te-IN":
voice = "te-IN-ShrutiNeural"
case "mr-IN":
voice = "mr-IN-AarohiNeural"
case "ur-IN":
voice = "ur-IN-GulNeural"
case "tr-TR":
voice = "tr-TR-EmelNeural"
case "fa-IR":
voice = "fa-IR-DilaraNeural"
// 其他重要语言
case "tr-TR":
voice = "tr-TR-SerifeNeural"
case "uk-UA":
voice = "uk-UA-OksanaNeural"
case "fa-IR":
voice = "fa-IR-SaraNeural"
case "nl-NL":
voice = "nl-NL-LotteNeural"
// 欧洲语言
case "fr-FR":
voice = "fr-FR-DeniseNeural"
case "es-ES":
voice = "es-ES-ElviraNeural"
case "pt-BR":
voice = "pt-BR-FranciscaNeural"
case "it-IT":
voice = "it-IT-ElsaNeural"
case "de-DE":
voice = "de-DE-KatjaNeural"
case "ru-RU":
voice = "ru-RU-SvetlanaNeural"
case "pl-PL":
voice = "pl-PL-AgnieszkaNeural"
case "nl-NL":
voice = "nl-NL-ColetteNeural"
case "sv-SE":
voice = "sv-SE-SofieNeural"
case "cs-CZ":
voice = "cs-CZ-VlastaNeural"
case "el-GR":
voice = "el-GR-AthinaNeural"
case "ro-RO":
voice = "ro-RO-AlinaNeural"
case "hu-HU":
voice = "hu-HU-NoemiNeural"
case "uk-UA":
voice = "uk-UA-PolinaNeural"
case "da-DK":
voice = "da-DK-ChristelNeural"
case "fi-FI":
voice = "fi-FI-NooraNeural"
case "nb-NO":
voice = "nb-NO-IselinNeural"
case "hr-HR":
voice = "hr-HR-GabrijelaNeural"
case "ca-ES":
voice = "ca-ES-JoanaNeural"
// 阿拉伯语和非洲语言
case "ar-EG":
voice = "ar-EG-SalmaNeural"
case "sw-KE":
voice = "sw-KE-ZuriNeural"
default:
voice = "en-US-AriaNeural" // 默认回退英语
}

56
local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/VolcanoTranslationServiceImpl.swift
View File

@ -24,6 +24,8 @@ public class VolcanoTranslationServiceImpl: IntegratedSpeechTranslationService.T
private var region = "cn-north-1"
private var maxRetryAttempts = 3
private var timeout: TimeInterval = 10.0
private var isDisposed = false // 添加释放状态标志
private let disposeLock = NSLock() // 添加线程安全锁
// HTTP客户端
private lazy var urlSession: URLSession = {
@ -52,7 +54,16 @@ public class VolcanoTranslationServiceImpl: IntegratedSpeechTranslationService.T
"ms-MY": "ms", // 马来语（马来西亚）
"id-ID": "id", // 印尼语（印尼）
"fil-PH": "tl", // 菲律宾语（菲律宾）
"sv-SE": "sv", // 瑞典语（瑞典）
"ro-RO": "ro", // 罗马尼亚语（罗马尼亚）
"hu-HU": "hu", // 匈牙利语（匈牙利）
"da-DK": "da", // 丹麦语（丹麦）
"fi-FI": "fi", // 芬兰语（芬兰）
"nb-NO": "nb", // 挪威语（挪威）
"hr-HR": "hr", // 克罗地亚语（克罗地亚）
"ca-ES": "ca", // 加泰隆语
"mr-IN": "mr", // 马拉地语（印度）
"ur-IN": "ur", // 乌尔都语（印度）
// 欧洲
"en-US": "en", // 英语（美国）
"en-GB": "en", // 英语（英国）
@ -150,11 +161,25 @@ public class VolcanoTranslationServiceImpl: IntegratedSpeechTranslationService.T
/**
* 释放资源
* 添加线程安全保护，防止重复释放
*/
public func dispose() {
disposeLock.lock()
defer { disposeLock.unlock() }
if isDisposed {
os_log("翻译服务已经释放，跳过重复释放", log: log, type: .info)
return
}
os_log("开始释放翻译服务资源", log: log, type: .info)
isDisposed = true
isInitialized = false
// 取消并失效URLSession
urlSession.invalidateAndCancel()
os_log("火山翻译服务已释放", log: log, type: .info)
os_log("翻译服务资源释放完成", log: log, type: .info)
}
// MARK: - Private Methods
@ -203,6 +228,7 @@ public class VolcanoTranslationServiceImpl: IntegratedSpeechTranslationService.T
/**
* 执行翻译
* 添加服务状态检查，防止在已释放的会话中创建任务
*/
private func performTranslation(
text: String,
@ -210,6 +236,19 @@ public class VolcanoTranslationServiceImpl: IntegratedSpeechTranslationService.T
targetLanguage: String
) async -> IntegratedSpeechTranslationService.TranslationResult {
// 检查服务是否已被释放
disposeLock.lock()
let disposed = isDisposed
disposeLock.unlock()
if disposed {
os_log("翻译服务已释放，取消翻译请求", log: log, type: .error)
return IntegratedSpeechTranslationService.TranslationResult(
success: false,
error: "翻译服务已释放"
)
}
do {
// 转换语言代码
let sourceCode = getLanguageCode(languageCode: sourceLanguage)
@ -263,6 +302,19 @@ public class VolcanoTranslationServiceImpl: IntegratedSpeechTranslationService.T
os_log("发送翻译请求到: %@", log: log, type: .debug, url.absoluteString)
// 再次检查服务状态，防止在请求构建过程中被释放
disposeLock.lock()
let stillDisposed = isDisposed
disposeLock.unlock()
if stillDisposed {
os_log("翻译服务在请求过程中被释放，取消网络请求", log: log, type: .error)
return IntegratedSpeechTranslationService.TranslationResult(
success: false,
error: "翻译服务已释放"
)
}
// 发送请求
let (data, response) = try await urlSession.data(for: request)

54
local_plugins/azure_speech/ios/azure_speech/Sources/tools/SimpleAudioReceiver.swift
View File

@ -144,11 +144,11 @@ public class SimpleAudioReceiver: NSObject {
while self.isRunning {
if !self._isWriting {
// print("startWriteThread 111")
usleep(10_000)
continue
}
// print("startWriteThread 222")
guard let dataToWrite = self.writeQueue.take() else { continue }
do {
@ -157,14 +157,12 @@ public class SimpleAudioReceiver: NSObject {
self.audioDataCallback!.onAudio(dataToWrite)
}
if self.pushAudioStream != nil && isContinuousRecognitionActive {
try self.pushAudioStream?.write(dataToWrite)
}
if self.recordfile != nil && isRecord{
// print("写入recordfile数据长度: \(dataToWrite.count)")
recordfile?.saveAudioDataToWav(dataToWrite)
}
// print("写入数据长度: \(dataToWrite.count)")
} catch {
print("推送音频数据失败: \(error.localizedDescription)")
}
@ -183,9 +181,8 @@ public class SimpleAudioReceiver: NSObject {
if audioSourceType != .external {
return
}
// print("外部data=\(data)")
// 放入队列，由写线程写入
writeQueue.put(data)
self.writeQueue.put(data)
}
/**
@ -220,18 +217,22 @@ public class SimpleAudioReceiver: NSObject {
}
}
/**
* 运行外部音频捕获
* 修复OSStatus error -50问题
*/
private func runExternalCapture() {
if audioSourceType == .external {
do {
try audioSession.setCategory(.playback,
mode: .videoChat,
options: [ .mixWithOthers,.allowBluetoothA2DP // 允许蓝牙耳机，不占用hfp链路
]) // 添加音频优先级控制
try audioSession.overrideOutputAudioPort(.none)
try audioSession.setActive(true)
} catch {
print("设置音频会话失败: \(error.localizedDescription)")
}
// do {
// try audioSession.setCategory(.playback,
// mode: .videoChat,
// options: [ .mixWithOthers,.allowBluetoothA2DP // 允许蓝牙耳机，不占用hfp链路
// ]) // 添加音频优先级控制
// try audioSession.overrideOutputAudioPort(.none)
// try audioSession.setActive(true)
// } catch {
// print("设置音频会话失败: \(error.localizedDescription)")
// }
//pushAudioData(data: Data())
micCapture.stopCapture()
@ -336,7 +337,7 @@ public func restoreOriginalAudioState() {
public func stopMicrophoneCapture() {
self.writeQueue.clear()
_isWriting=false
micCapture.stopCapture()
print("stopMicrophoneCapture")
@ -517,6 +518,23 @@ private class LinkedBlockingQueue<T> {
// 通知有新元素可用
availableItems.signal()
}
/**
* 清空队列
*/
func clear() {
queueLock.lock()
defer { queueLock.unlock() }
// 检查队列是否已关闭
if isClosed {
return
}
// 添加元素（无容量限制，与Android一致）
queue.removeAll()
}
/**

63
local_plugins/ble_service/android/src/main/kotlin/com/yunqiinnovation/ble_service/BleService.kt
View File

@ -178,7 +178,7 @@ object BleService {
*/
fun initialize(appContext: Context): Boolean {
if (isInitialized) return true
try {
this.context = appContext.applicationContext
// 初始化蓝牙管理器和适配器
@ -190,6 +190,11 @@ object BleService {
// 初始化Handler
scanHandler = Handler(Looper.getMainLooper())
recordfile = RecordingFile(this.context)
recordfile!!.fileName = "不拆分"
recordfile1 = RecordingFile(this.context)
recordfile1!!.fileName = "重新压缩"
// 启动队列处理
startAudioQueueProcessing()
@ -202,13 +207,7 @@ object BleService {
Log.e(TAG, "OpusManager初始化失败: ${e.message}", e)
// 根据需要决定是否因为Opus初始化失败而返回false
}
recordfile = RecordingFile(this.context)
recordfile!!.fileName = "不拆分"
recordfile1 = RecordingFile(this.context)
recordfile1!!.fileName = "重新压缩"
isInitialized = true
Log.d(TAG, "BLE服务初始化成功")
return true
} catch (e: Exception) {
Log.e(TAG, "BLE服务初始化失败: ${e.message}", e)
@ -838,22 +837,7 @@ startBytesStatistics()
recordfile?.saveAudioDataToWav(data)
// 将完整的帧加入队列，由专门的线程处理
audioDataQueue.offer(data)
// // 将接收到的数据添加到缓存中
// synchronized(audioDataBuffer) {
// audioDataBuffer.addAll(data.toList())
// // 当缓存达到阈值时，提取完整的帧进行解码
// while (audioDataBuffer.size >= AUDIO_BUFFER_SIZE) {
// // 提取一个完整的帧（1280字节）
// val frameData = ByteArray(AUDIO_BUFFER_SIZE)
// for (i in 0 until AUDIO_BUFFER_SIZE) {
// frameData[i] = audioDataBuffer.removeAt(0)
// }
// Log.d(TAG, "缓存达到阈值，提取 ${frameData.size} 字节帧进行解码")
// }
// }
} else {
// Log.d(TAG, "Opus解码流未启动，忽略音频数据")
}
@ -1243,7 +1227,7 @@ startBytesStatistics()
// Opus 解码相关
// ======================================================================================================
private fun startOpusStreamDecoding(
public fun startOpusStreamDecoding(
hasHeader: Boolean = false, // 通常BLE传输的Opus没有文件头
channel: Int = 1,
sampleRate: Int = 16000, // 确认设备端Opus编码采样率
@ -1633,7 +1617,12 @@ startBytesStatistics()
) // 或其他默认声道
)
}
fun closeCodec1(): Boolean {
recordfile1!!.closeFile()
recordfile!!.closeFile()
return true
}
/**
* 控制编解码 - 打开解码
*/
@ -1706,16 +1695,20 @@ startBytesStatistics()
*/
fun openEncoder(): Boolean {
Log.i(TAG, "打开编码0xB1")
recordfile1!!.closeFile()
recordfile1!!.creatingFiles()
recordfile!!.closeFile()
recordfile!!.creatingFiles()
startOpusStreamDecoding(false, 1, 16000, 40)
// 目前仅发送命令通知设备开始编码。
return sendCommand(
BleConst.CMD_CONTROL_CODEC.toByte(),
byteArrayOf(
BleConst.CODEC_CONTROL_ENCODE_ON.toByte(),
BleConst.AUDIO_CHANNEL_RIGHT.toByte() // 或设备期望的声道
)
)
// // 目前仅发送命令通知设备开始编码。
// return sendCommand(
// BleConst.CMD_CONTROL_CODEC.toByte(),
// byteArrayOf(
// BleConst.CODEC_CONTROL_ENCODE_ON.toByte(),
// BleConst.AUDIO_CHANNEL_RIGHT.toByte() // 或设备期望的声道
// )
// )
return true
}
@ -2101,7 +2094,7 @@ class RecordingFile(private val context: Context) {
}
/**
* 保存音频数据到 WAV 文件（异步）
* 保存音频数据到 opus 文件（异步）
*/
internal fun saveAudioDataToWav(buffer: ByteArray) {
if (fos == null || currentAudioFile == null) return

24
local_plugins/ble_service/ios/ble_service/Sources/ble_service/BleService.swift
View File

@ -586,22 +586,24 @@ private var cmdReplyType: UInt8 = 0
}
/// 打开编码器
func openEncoder() -> Bool {
/// - Returns: 操作是否成功
public func openEncoder() -> Bool {
os_log("🎙️ 准备打开编码（设备侧）...", log: logger, type: .info)
// 确保Opus解码器处于活跃状态（如果不是则启动）
// if opusProcessor?.isDecoding() != true {
startOpusDecoding()
//}
let paramData = Data([BleConst.CODEC_CONTROL_ENCODE_ON, BleConst.AUDIO_CHANNEL_RIGHT])
let result = sendCommand(BleConst.CMD_CONTROL_CODEC, data: paramData)
// let paramData = Data([BleConst.CODEC_CONTROL_ENCODE_ON, BleConst.AUDIO_CHANNEL_RIGHT])
// let result = sendCommand(BleConst.CMD_CONTROL_CODEC, data: paramData)
if result {
os_log("✅ 编码器命令发送成功", log: logger, type: .info)
} else {
os_log("❌ 编码器命令发送失败", log: logger, type: .error)
}
// if result {
// os_log("✅ 编码器命令发送成功", log: logger, type: .info)
// } else {
// os_log("❌ 编码器命令发送失败", log: logger, type: .error)
// }
return result
//return result
return true
}
/// 打开解码器
@ -1108,9 +1110,7 @@ private var cmdReplyType: UInt8 = 0
notifyCharacteristic != nil ? "✅" : "❌",
audioCharacteristic != nil ? "✅" : "❌")
// 接收到唤醒信号，打开编码器 (设备侧)
let openResult = openEncoder()
os_log("🎙️ 尝试打开编码器结果: %@", log: logger, type: .info, openResult ? "成功" : "失败")
notifyWakeupSignalReceived()
default:

Write Preview
Loading…
Cancel
Save