diff --git a/local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentService.kt b/local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentService.kt index a1921b405..bb34a6d5c 100644 --- a/local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentService.kt +++ b/local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentService.kt @@ -15,6 +15,7 @@ import com.yunqiinnovation.user_portrait.UserPortrait import com.yunqiinnovation.location_service.LocationService import android.util.Log import kotlinx.coroutines.* +import kotlinx.coroutines.channels.Channel import java.util.Collections import kotlin.coroutines.CoroutineContext import com.yunqiinnovation.ble_service.BleService @@ -116,7 +117,7 @@ object AgentService : CoroutineScope { // 空闲检测相关 private var idleCheckJob: Job? = null - private val maxIdleSeconds = 10 // 最大空闲秒数 + private val maxIdleSeconds = 10 // 最大空闲秒数（保持原有设置） // 打断防抖相关 private var lastInterruptTime = 0L @@ -192,27 +193,31 @@ object AgentService : CoroutineScope { } // 初始化Azure ASR azureAsrHelper = AzureAsrHelper(context).apply { - // 使用协程处理异步初始化 - CoroutineScope(Dispatchers.IO).launch { - //获取语音识别的语种 - val supportedLanguagesConfig = config["supportedLanguages"] as? List - val supportedLanguages = supportedLanguagesConfig?.toTypedArray() - ?: arrayOf("zh-CN", "en-US") // 默认支持多语言 - - val asrInitSuccess = initialize( - subscriptionKey = config["azureSpeechKey"]?.toString() ?: "", - region = config["azureSpeechRegion"]?.toString() ?: "", - supportedLanguages = supportedLanguages, - audioSourceType = if (isExternalActive) AzureAsrHelper.AudioSourceType.EXTERNAL - else AzureAsrHelper.AudioSourceType.MICROPHONE - ) + // 使用AgentService的协程作用域，确保生命周期管理 + launch(Dispatchers.IO) { + try { + //获取语音识别的语种 + val supportedLanguagesConfig = config["supportedLanguages"] as? List + val supportedLanguages = supportedLanguagesConfig?.toTypedArray() + ?: arrayOf("zh-CN", "en-US") // 默认支持多语言 + + val asrInitSuccess = initialize( + subscriptionKey = config["azureSpeechKey"]?.toString() ?: "", + region = config["azureSpeechRegion"]?.toString() ?: "", + supportedLanguages = supportedLanguages, + audioSourceType = if (isExternalActive) AzureAsrHelper.AudioSourceType.EXTERNAL + else AzureAsrHelper.AudioSourceType.MICROPHONE + ) - if (asrInitSuccess) { - withContext(Dispatchers.Main) { - recognizeCallback() + if (asrInitSuccess) { + withContext(Dispatchers.Main) { + recognizeCallback() + } + } else { + Log.w(TAG, "Azure ASR初始化失败") } - } else { - Log.w(TAG, "Azure ASR初始化失败") + } catch (e: Exception) { + Log.e(TAG, "Azure ASR初始化异常", e) } } } @@ -244,35 +249,71 @@ object AgentService : CoroutineScope { } /** - * 释放所有资源 + * 释放所有资源 - 增强版本 */ fun dispose() { try { + Log.i(TAG, "开始释放AgentService资源...") + + // 1. 停止所有活动 stopRecognition() stopTts() stopAiStream() stopIdleCheck() + // 2. 停止事件处理器 + eventProcessorJob?.cancel() + eventProcessorJob = null + eventQueue.close() + + // 3. 释放音频播放器 audioPlayer?.release() audioPlayer = null - azureAsrHelper?.dispose() + // 4. 释放Azure ASR（确保完全清理） + azureAsrHelper?.let { helper -> + try { + helper.dispose() + } catch (e: Exception) { + Log.e(TAG, "释放Azure ASR异常", e) + } + } + azureAsrHelper = null - // 释放ChatAPI服务 + // 5. 释放ChatAPI服务 if (::chatApiService.isInitialized) { - chatApiService.cancelCurrentStream() - chatApiService.dispose() + try { + chatApiService.cancelCurrentStream() + chatApiService.dispose() + } catch (e: Exception) { + Log.e(TAG, "释放ChatAPI服务异常", e) + } } - // 释放TTS服务 - ttsService?.release() + // 6. 释放TTS服务 + ttsService?.let { tts -> + try { + tts.release() + } catch (e: Exception) { + Log.e(TAG, "释放TTS服务异常", e) + } + } ttsService = null + // 7. 取消所有协程并清理监听器 job.cancel() clearListeners() + + // 8. 重置状态 _isInitialized.set(false) + _isRecognitionActive.set(false) + _hasSpeechDetected.set(false) + _isTtsSpeaking.set(false) + _isAiStreaming.set(false) + + Log.i(TAG, "AgentService资源释放完成") } catch (e: Exception) { - Log.e(TAG, "释放资源异常: ${e.message}") + Log.e(TAG, "释放资源异常: ${e.message}", e) } } @@ -415,19 +456,31 @@ object AgentService : CoroutineScope { stopIdleCheck() // 先停止现有的检查 if (!isRecognitionActive) return + // BLE模式下使用稍长的超时时间，但不影响正常使用 + val timeoutSeconds = if (audioSourceType == AzureAsrHelper.AudioSourceType.EXTERNAL) { + maxIdleSeconds + 3 // BLE模式额外增加3秒，避免过度延长 + } else { + maxIdleSeconds + } + idleCheckJob = launch { try { - delay(maxIdleSeconds * 1000L) + Log.d(TAG, "启动空闲检测，超时时间: ${timeoutSeconds}秒，音频源: $audioSourceType") + delay(timeoutSeconds * 1000L) if (isActive && isRecognitionActive && !hasSpeechDetected && !isTtsSpeaking && !isAiStreaming) { + Log.i(TAG, "空闲超时，停止识别 - 超时时间: ${timeoutSeconds}秒") stopRecognition() sendEvent( "auto_stop", - mapOf("reason" to "idle_timeout", "seconds" to maxIdleSeconds) + mapOf("reason" to "idle_timeout", "seconds" to timeoutSeconds) ) + } else { + Log.d(TAG, "空闲检测结束，但条件不满足停止要求") } } catch (e: CancellationException) { // 协程被取消，正常行为 + Log.d(TAG, "空闲检测被取消") } catch (e: Exception) { Log.e(TAG, "空闲检测异常: ${e.message}") } @@ -460,26 +513,12 @@ object AgentService : CoroutineScope { Log.d(TAG, "服务未初始化") return false } - Log.d("TAG", "开始语音识别startRecognition?") + + Log.i(TAG, "开始语音识别 - 模式: ${if (isExternalActive) "BLE外部音频" else "设备麦克风"}") + + // 先停止当前识别 stopRecognition() - BleService.openEncoder() - // if (isRecognitionActive&&audioSourceType==AzureAsrHelper.AudioSourceType.MICROPHONE){ - // Log.d(TAG, "开始语音识别stopRecognition") - // // BleService.closeCodec() - // stopRecognition() - - // } - // if((isExternalActive&&audioSourceType==AzureAsrHelper.AudioSourceType.MICROPHONE)||!isRecognitionActive) - // { - // Log.d(TAG, "开始语音识别BleService.openEncoder") - // BleService.openEncoder() - - // } - // if(isRecognitionActive) - // { - // Log.d(TAG, "开始语音识别azureAsrHelper?.startContinuousRecognition(audioSourceType)") - // return true - // } + // 检查录音权限 if (!isExternalActive && !checkRecordAudioPermission()) { Log.e(TAG, "需要录音权限") @@ -487,22 +526,45 @@ object AgentService : CoroutineScope { return false } - _isRecognitionActive.set(true) - _hasSpeechDetected.set(false) + // 设置音频源类型 + audioSourceType = if (isExternalActive) { + AzureAsrHelper.AudioSourceType.EXTERNAL + } else { + AzureAsrHelper.AudioSourceType.MICROPHONE + } + + Log.i(TAG, "音频源类型设置为: $audioSourceType") try { - audioSourceType = if (isExternalActive) { - AzureAsrHelper.AudioSourceType.EXTERNAL - } else { - AzureAsrHelper.AudioSourceType.MICROPHONE + // 启动Azure ASR + val asrStarted = azureAsrHelper?.startContinuousRecognition(audioSourceType) ?: false + if (!asrStarted) { + Log.e(TAG, "Azure ASR启动失败") + return false + } + + // 如果是BLE模式，在ASR启动后再打开编码器 + if (isExternalActive) { + Log.i(TAG, "BLE模式：启动编码器") + // 使用协程延迟而不是阻塞主线程 + launch { + delay(50) // 给ASR一些启动时间 + val encoderOpened = BleService.openEncoder() + if (!encoderOpened) { + Log.w(TAG, "BLE编码器启动失败，但继续尝试识别") + } + } } - Log.d("TAG", "azureAsrHelper?.startContinuousRecognition") - azureAsrHelper?.startContinuousRecognition(audioSourceType) + _isRecognitionActive.set(true) + _hasSpeechDetected.set(false) + + Log.i(TAG, "语音识别启动成功") return true + } catch (e: Exception) { _isRecognitionActive.set(false) - Log.e(TAG, "启动语音识别失败: ${e.message}") + Log.e(TAG, "启动语音识别失败: ${e.message}", e) sendEvent( "error", mapOf( "code" to "RECOGNITION_START_ERROR", @@ -525,13 +587,17 @@ object AgentService : CoroutineScope { _hasSpeechDetected.set(true) if (!previousHasSpeech) { + Log.i(TAG, "首次检测到语音输入: $recognizing (音频源: $audioSourceType)") restartIdleCheck() } + // 实时发送识别结果，确保UI及时更新 + Log.v(TAG, "实时识别: $recognizing") sendEvent( "recognizing", mapOf( "text" to recognizing, - "language" to detectedLanguage + "language" to detectedLanguage, + "audioSource" to audioSourceType.toString() ) ) @@ -604,28 +670,42 @@ object AgentService : CoroutineScope { override fun onCanceled(reason: String, errorDetails: String) { _isRecognitionActive.set(false) stopIdleCheck() - Log.d(TAG, "recognition_canceled:") - BleService.closeCodec() + Log.w(TAG, "语音识别被取消: $reason, 详情: $errorDetails") + + // 如果是BLE模式，关闭编码器 + if (audioSourceType == AzureAsrHelper.AudioSourceType.EXTERNAL) { + BleService.closeCodec() + } + sendEvent( "recognition_canceled", mapOf( "reason" to reason, "details" to errorDetails ) ) + + // 重新设置回调，但不自动重启识别 recognizeCallback() } override fun onError(error: String) { _isRecognitionActive.set(false) stopIdleCheck() - BleService.closeCodec() - Log.d(TAG, "语音识别错误: $error") + Log.e(TAG, "语音识别错误: $error") + + // 如果是BLE模式，关闭编码器 + if (audioSourceType == AzureAsrHelper.AudioSourceType.EXTERNAL) { + BleService.closeCodec() + } + sendEvent( "error", mapOf( "code" to "RECOGNITION_ERROR", "message" to error ) ) + + // 重新设置回调，但不自动重启识别 recognizeCallback() } }) @@ -635,10 +715,68 @@ object AgentService : CoroutineScope { } } + // 音频数据推送统计 + private var audioDataCount = 0 + private var lastAudioDataTime = 0L + + // 性能监控 + private var lastMemoryCheckTime = 0L + private val MEMORY_CHECK_INTERVAL = 30000L // 30秒检查一次内存 + fun pushAudioData(data: ByteArray) { + val currentTime = System.currentTimeMillis() + audioDataCount++ + + // 每100个数据包或每5秒输出一次统计 + if (audioDataCount % 100 == 0 || (currentTime - lastAudioDataTime) > 5000) { + Log.i(TAG, "音频数据统计: 已推送${audioDataCount}包, 当前包大小: ${data.size} bytes, 音频源: $audioSourceType") + lastAudioDataTime = currentTime + + // 定期检查内存使用情况 + checkMemoryUsage(currentTime) + } + + // 添加调试日志，但保持原有逻辑的兼容性 + if (!isRecognitionActive) { + Log.w(TAG, "语音识别未激活，但仍尝试推送音频数据: ${data.size} bytes") + } + + if (audioSourceType != AzureAsrHelper.AudioSourceType.EXTERNAL) { + Log.w(TAG, "音频源类型不匹配: $audioSourceType，推送音频数据: ${data.size} bytes") + } + + // 保持原有逻辑，确保兼容性 azureAsrHelper?.audioStream?.saveAudioDataTo(data) } + /** + * 检查内存使用情况 + */ + private fun checkMemoryUsage(currentTime: Long) { + if (currentTime - lastMemoryCheckTime > MEMORY_CHECK_INTERVAL) { + lastMemoryCheckTime = currentTime + + val runtime = Runtime.getRuntime() + val usedMemory = runtime.totalMemory() - runtime.freeMemory() + val maxMemory = runtime.maxMemory() + val memoryUsagePercent = (usedMemory * 100 / maxMemory).toInt() + + Log.i(TAG, "内存使用情况: ${usedMemory / 1024 / 1024}MB / ${maxMemory / 1024 / 1024}MB (${memoryUsagePercent}%)") + + // 如果内存使用超过80%，发出警告 + if (memoryUsagePercent > 80) { + Log.w(TAG, "内存使用率过高: ${memoryUsagePercent}%，建议检查内存泄漏") + + // 发送内存警告事件 + sendEvent("memory_warning", mapOf( + "usedMemoryMB" to (usedMemory / 1024 / 1024), + "maxMemoryMB" to (maxMemory / 1024 / 1024), + "usagePercent" to memoryUsagePercent + )) + } + } + } + /** * 停止语音识别 */ @@ -1236,22 +1374,56 @@ object AgentService : CoroutineScope { } } - /** - * 发送事件 - */ - private fun sendEvent(eventName: String, data: Map) { - // 切换到主线程执行监听器回调，避免 UI 更新问题 - launch(Dispatchers.Main) { - listeners.forEach { listener -> + // 事件发送优化：使用单一协程处理事件队列 + private val eventQueue = Channel>>(capacity = Channel.UNLIMITED) + private var eventProcessorJob: Job? = null + + init { + // 启动事件处理协程 + startEventProcessor() + } + + private fun startEventProcessor() { + eventProcessorJob = launch(Dispatchers.Main) { + for ((eventName, data) in eventQueue) { try { - listener.onEvent(eventName, data) + // 创建监听器快照，避免并发修改 + val currentListeners = synchronized(listeners) { listeners.toList() } + + currentListeners.forEach { listener -> + try { + listener.onEvent(eventName, data) + } catch (e: Exception) { + Log.e(TAG, "向监听器发送事件失败: $eventName", e) + } + } } catch (e: Exception) { - Log.e(TAG, "向监听器发送事件失败", e) + Log.e(TAG, "处理事件失败: $eventName", e) } } } } + /** + * 发送事件 - 优化版本，使用事件队列避免主线程阻塞 + */ + private fun sendEvent(eventName: String, data: Map) { + try { + // 对于高频事件，进行限流处理 + if (eventName == "recognizing" && eventQueue.trySend(Pair(eventName, data)).isFailure) { + Log.v(TAG, "识别事件队列已满，跳过当前事件") + return + } + + // 非阻塞发送到事件队列 + if (!eventQueue.trySend(Pair(eventName, data)).isSuccess) { + Log.w(TAG, "事件队列已满，丢弃事件: $eventName") + } + } catch (e: Exception) { + Log.e(TAG, "发送事件异常: $eventName", e) + } + } + /** * 检查录音权限 */ diff --git a/local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentServicePlugin.kt b/local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentServicePlugin.kt index 60b919843..dc5428345 100644 --- a/local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentServicePlugin.kt +++ b/local_plugins/agent_service/android/src/main/kotlin/com/yunqiinnovation/agent_service/AgentServicePlugin.kt @@ -286,19 +286,43 @@ class AgentServicePlugin : FlutterPlugin, MethodCallHandler, EventChannel.Stream } /** - * 实现AgentServiceListener接口 + * 实现AgentServiceListener接口 - 优化版本 * 接收来自AgentService的事件，并转发到Flutter */ override fun onEvent(eventName: String, data: Map) { - // 将事件转换为JSON字符串 - val eventData = JSONObject().apply { - put("event", eventName) - put("data", JSONObject(data)) - } + try { + // 对于高频事件，进行优化处理 + val eventData = when (eventName) { + "recognizing" -> { + // 实时识别事件，简化数据结构 + JSONObject().apply { + put("event", eventName) + put("data", JSONObject().apply { + put("text", data["text"] ?: "") + put("language", data["language"] ?: "") + put("audioSource", data["audioSource"] ?: "") + }) + } + } + else -> { + // 其他事件，正常处理 + JSONObject().apply { + put("event", eventName) + put("data", JSONObject(data)) + } + } + } - // 在主线程发送事件到Flutter - MainScope().launch { - eventSink?.success(eventData.toString()) + // 使用协程在主线程发送事件 + MainScope().launch { + try { + eventSink?.success(eventData.toString()) + } catch (e: Exception) { + Log.e(TAG, "发送事件到Flutter失败: $eventName", e) + } + } + } catch (e: Exception) { + Log.e(TAG, "处理事件异常: $eventName", e) } } diff --git a/local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServiceImpl.swift b/local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServiceImpl.swift index 684616bc8..6854e8974 100644 --- a/local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServiceImpl.swift +++ b/local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServiceImpl.swift @@ -104,13 +104,34 @@ class AgentServiceImpl: NSObject { listenersLock.unlock() } + // 事件发送优化：使用队列避免主线程阻塞 + private let eventQueue = DispatchQueue(label: "com.yunqiinnovation.agent_service.events", qos: .userInitiated) + internal func sendEvent(name eventName: String, data: [String: Any]) { + // 对于高频事件，使用异步处理 + if eventName == "recognizing" { + eventQueue.async { [weak self] in + self?.dispatchEventToListeners(eventName: eventName, data: data) + } + } else { + // 其他事件在主线程处理 + DispatchQueue.main.async { [weak self] in + self?.dispatchEventToListeners(eventName: eventName, data: data) + } + } + } + + private func dispatchEventToListeners(eventName: String, data: [String: Any]) { listenersLock.lock() let currentListeners = self.listeners listenersLock.unlock() - + for listener in currentListeners { - listener.onEvent(eventName: eventName, data: data) + do { + listener.onEvent(eventName: eventName, data: data) + } catch { + os_log("发送事件失败: %{public}@", log: logger, type: .error, eventName) + } } } @@ -275,25 +296,39 @@ class AgentServiceImpl: NSObject { private func startIdleCheck() { stopIdleCheck() - + guard isRecognizing else { return } - + + // BLE模式下使用稍长的超时时间 + let timeoutSeconds: TimeInterval + if let azureHelper = azureAsrHelper, + azureHelper.audioSourceType == .external { + timeoutSeconds = maxIdleSeconds + 3 // BLE模式额外增加3秒 + } else { + timeoutSeconds = maxIdleSeconds + } + + os_log("启动空闲检测，超时时间: %.0f秒", log: logger, type: .debug, timeoutSeconds) + // 使用全局队列而不是主队列，避免后台挂起问题 let timer = DispatchSource.makeTimerSource(queue: DispatchQueue.global(qos: .utility)) idleTimer = timer - - timer.schedule(deadline: .now() + maxIdleSeconds) + + timer.schedule(deadline: .now() + timeoutSeconds) timer.setEventHandler { [weak self] in // 使用更安全的方式检查self是否存在 DispatchQueue.main.async { [weak self] in guard let self = self else { return } - + if self.isRecognizing && !self.hasSpeechDetected && !self.isSpeaking && !self.isAiStreaming { + os_log("空闲超时，停止识别 - 超时时间: %.0f秒", log: self.logger, type: .info, timeoutSeconds) self.stopRecognition() self.sendEvent(name: "auto_stop", data: [ - "reason": "idle_timeout", - "seconds": self.maxIdleSeconds + "reason": "idle_timeout", + "seconds": timeoutSeconds ]) + } else { + os_log("空闲检测结束，但条件不满足停止要求", log: self.logger, type: .debug) } } } @@ -315,34 +350,49 @@ class AgentServiceImpl: NSObject { sendError("服务未初始化", code: "NOT_INITIALIZED") return false } - + + os_log("开始语音识别 - 模式: %{public}@", log: logger, type: .info, useBle ? "BLE外部音频" : "设备麦克风") + if !useBle { guard AVAudioSession.sharedInstance().recordPermission == .granted else { sendError("无麦克风权限", code: "PERMISSION_DENIED") return false } } - + if isRecognizing { stopRecognition() } - + let audioSourceType: AzureAsrHelper.AudioSourceType = useBle ? .external : .microphone - + if useBle { AudioSessionHub.shared.begin(.playback) } else { AudioSessionHub.shared.begin(.voice) } -print("ai启动语音=\(audioSourceType)=\(useBle)") + + os_log("音频源类型设置为: %{public}@", log: logger, type: .info, audioSourceType == .external ? "EXTERNAL" : "MICROPHONE") + guard let success = azureAsrHelper?.startContinuousRecognition( - // callback: self, audioSourceType: audioSourceType ), success else { sendError("启动语音识别失败", code: "RECOGNITION_START_ERROR") return false } - + + // 如果是BLE模式，在ASR启动后再打开编码器 + if useBle { + os_log("BLE模式：启动编码器", log: logger, type: .info) + DispatchQueue.global().asyncAfter(deadline: .now() + 0.05) { + let encoderOpened = BleService.shared.openEncoder() + if !encoderOpened { + os_log("BLE编码器启动失败，但继续尝试识别", log: self.logger, type: .warning) + } + } + } + + os_log("语音识别启动成功", log: logger, type: .info) return true } func recognizeCallback(useBle: Bool = false) -> Bool { @@ -363,13 +413,15 @@ print("ai启动语音=\(audioSourceType)=\(useBle)") } func pushAudioData(_ audioData: Data) -> Bool { + // 添加调试日志，但保持兼容性 if !isRecognizing || !isInitialized { + os_log("语音识别未激活或服务未初始化，但仍尝试推送音频数据: %d bytes", log: logger, type: .debug, audioData.count) + } + + guard let audioStream = azureAsrHelper?.audioStream else { + os_log("音频流未初始化", log: logger, type: .error) return false } - guard let audioStream = azureAsrHelper?.audioStream else { - os_log("音频流未初始化", log: logger,type: .error) - return false -} audioStream.saveAudioDataTo(data: audioData) return true @@ -1398,6 +1450,8 @@ extension AgentServiceImpl: AzureAsrHelper.ContinuousRecognizeCallback { } func onCanceled(_ reason: String, _ errorDetails: String) { + os_log("语音识别被取消: %{public}@, 详情: %{public}@", log: logger, type: .warning, reason, errorDetails) + var data: [String: Any] = [:] if !reason.isEmpty { data["reason"] = reason @@ -1405,13 +1459,27 @@ extension AgentServiceImpl: AzureAsrHelper.ContinuousRecognizeCallback { if !errorDetails.isEmpty { data["details"] = errorDetails } - + + // 如果是BLE模式，关闭编码器 + if let azureHelper = azureAsrHelper, + azureHelper.audioSourceType == .external { + BleService.shared.closeCodec() + } + sendEvent(name: "recognition_canceled", data: data) isRecognizing = false stopIdleCheck() } - + func onError(_ error: String) { + os_log("语音识别错误: %{public}@", log: logger, type: .error, error) + + // 如果是BLE模式，关闭编码器 + if let azureHelper = azureAsrHelper, + azureHelper.audioSourceType == .external { + BleService.shared.closeCodec() + } + let data: [String: Any] = ["message": error.isEmpty ? "未知错误" : error] sendEvent(name: "error", data: data) isRecognizing = false diff --git a/local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrHelper.kt b/local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrHelper.kt index 02d909d4a..d5b8dae6c 100644 --- a/local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrHelper.kt +++ b/local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrHelper.kt @@ -146,6 +146,17 @@ class AzureAsrHelper(private val context: Context) { // 设置分段策略为时间模式 setProperty("Speech_SegmentationStrategy", "Time") + + // 针对EXTERNAL音频源的实时性优化 + if (audioSourceType == AudioSourceType.EXTERNAL) { + // 减少缓冲延迟 + setProperty("SpeechServiceConnection_ReceiveTimeout", "1000") + setProperty("SpeechServiceConnection_SendTimeout", "1000") + // 启用实时模式 + setProperty("SpeechServiceConnection_EnableRealTimeRecognition", "true") + // 减少音频缓冲 + setProperty("SpeechServiceConnection_AudioBufferSizeMs", "100") + } } // 录音文件类 recordfile = RecordFile; @@ -706,7 +717,9 @@ class AzureAsrHelper(private val context: Context) { // 情况2：不在录制但队列有数据 -> 从队列获取 else if (writeQueue.isNotEmpty()) { data = writeQueue.poll() - Log.d("tag", "写入数据: ${data?.size}") + if (data != null) { + Log.v("tag", "从队列写入数据: ${data.size} bytes, 队列剩余: ${writeQueue.size}") + } bytesToWrite = data?.size ?: 0 } @@ -724,7 +737,8 @@ class AzureAsrHelper(private val context: Context) { Log.e("tag", "写入失败: ${e.message}") } } else { - Thread.yield() // 避免空转消耗CPU + // 使用更短的等待时间，提高响应性 + Thread.sleep(1) // 1ms等待，避免空转但保持响应性 } } } catch (e: Exception) { @@ -744,11 +758,22 @@ class AzureAsrHelper(private val context: Context) { * 外部音频输入 */ fun saveAudioDataTo(buffer: ByteArray) { - if (audioSourceType == AudioSourceType.MICROPHONE) return - // 放入队列，由写线程写入 - writeQueue.offer(buffer.copyOf()) + // 检查队列大小，避免积压 + if (writeQueue.size > 50) { // 限制队列大小 + Log.w("AudioStream", "音频队列积压严重(${writeQueue.size})，清理旧数据") + // 清理一半的旧数据，保持实时性 + repeat(writeQueue.size / 2) { + writeQueue.poll() + } + } + + // 放入队列，由写线程写入 + val offered = writeQueue.offer(buffer.copyOf()) + if (!offered) { + Log.w("AudioStream", "音频队列已满，丢弃数据包: ${buffer.size} bytes") + } } /** diff --git a/local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureAsrHelper.swift b/local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureAsrHelper.swift index 3106c1236..d8ee80e97 100644 --- a/local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureAsrHelper.swift +++ b/local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureAsrHelper.swift @@ -690,22 +690,27 @@ public class AzureAsrHelper: NSObject { guard let self = self else { return } while self.isRunning { - // print("是否写入: \(self.isWriting)") if !self.isWriting { - usleep(10_000) + usleep(1_000) // 减少等待时间，提高响应性 continue } - - guard let dataToWrite = self.writeQueue.take() else { continue } - - // print("写入数据长度: \(dataToWrite.count)") - do { - try self.pushAudioStream?.write(dataToWrite) - self.onAudioData?(dataToWrite) -} catch { - print("推送音频数据失败: \(error.localizedDescription)") -} + guard let dataToWrite = self.writeQueue.take() else { + usleep(1_000) // 短暂等待，避免空转 + continue + } + + // 添加队列状态日志 + if self.writeQueue.count > 10 { + print("音频写入队列积压: \(self.writeQueue.count)") + } + + do { + try self.pushAudioStream?.write(dataToWrite) + self.onAudioData?(dataToWrite) + } catch { + print("推送音频数据失败: \(error.localizedDescription)") + } } } } @@ -720,7 +725,16 @@ public class AzureAsrHelper: NSObject { if audioSourceType != .external { return } - // print("外部data=\(data)") + + // 检查队列大小，避免积压 + if writeQueue.count > 50 { + os_log("音频队列积压严重(%d)，清理旧数据", log: OSLog.default, type: .warning, writeQueue.count) + // 清理一半的旧数据，保持实时性 + for _ in 0..<(writeQueue.count / 2) { + _ = writeQueue.take() + } + } + // 放入队列，由写线程写入 writeQueue.put(data) }