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

Merge branch 'new_dev' of https://github.com/deepcloud2048/deep_voice into new_dev

weicu
liwei1dao 8 months ago
parent
1ea0a66467 f10cdf628c
commit
e0d6abd444
7 changed files with 670 additions and 523 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 2
      lib/modules/translation/controllers/translation_controller.dart
  2. 30
      local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrHelper.kt
  3. 7
      local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureSpeechPlugin.kt
  4. 76
      local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/XunFeiAsrHelper.kt
  5. 229
      local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureAsrHelper.swift
  6. 10
      local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureSpeechPlugin.swift
  7. 47
      local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/XunFeiAsrHelper.swift

2
lib/modules/translation/controllers/translation_controller.dart
View File

@ -964,8 +964,10 @@ class TranslationController extends GetxController with WidgetsBindingObserver {
} }
/// 开始语音识别 /// 开始语音识别
/// 同声翻译、通话翻译、多媒体翻译的语音识别调用入口
Future<void> startRecognition() async { Future<void> startRecognition() async {
Logger.info('开始语音识别'); Logger.info('开始语音识别');
//检查用户是否登录且积分是否足够，不足则提示充值
if (User.isLoggedIn() && User.instance.tradeintegral <= 0) { if (User.isLoggedIn() && User.instance.tradeintegral <= 0) {
showRechargeDialog('translation'); showRechargeDialog('translation');
return; return;

30
local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureAsrHelper.kt
View File

@ -127,10 +127,24 @@ class AzureAsrHelper(private val context: Context) {
this.supportedLanguages = supportedLanguages this.supportedLanguages = supportedLanguages
} }
Log.i(
tag,
"lxm---ASR initialize: supportedLanguages=${supportedLanguages.joinToString(",")}, audioSourceType=$audioSourceType"
)
val xunfeiConfigReady = this.xunfeiAppId.isNotBlank() && val xunfeiConfigReady = this.xunfeiAppId.isNotBlank() &&
this.xunfeiAccessKeyId.isNotBlank() && this.xunfeiAccessKeyId.isNotBlank() &&
this.xunfeiAccessKeySecret.isNotBlank() this.xunfeiAccessKeySecret.isNotBlank()
this.useXunfei = supportedLanguages.all { it == "zh-CN" || it == "en-US" || it == "zh-HK"|| it == "en-TW" } && xunfeiConfigReady val onlySupportedByXunfei = supportedLanguages.all {
it == "zh-CN" || it == "en-US" || it == "zh-HK" || it == "zh-TW"
}
val chineseVariantsCount = supportedLanguages
.map { it.trim().lowercase() }
.filter { it == "zh" || it.startsWith("zh-") }
.distinct()
.size
val hasMultipleChineseVariants = chineseVariantsCount >= 2
this.useXunfei = onlySupportedByXunfei && xunfeiConfigReady && !hasMultipleChineseVariants
// 根据支持的语言数量决定是否启用自动语言检测 // 根据支持的语言数量决定是否启用自动语言检测
this.isAutoDetectLanguage = supportedLanguages.size >= 2 this.isAutoDetectLanguage = supportedLanguages.size >= 2
@ -140,6 +154,11 @@ class AzureAsrHelper(private val context: Context) {
this.currentLanguage = supportedLanguages[0] this.currentLanguage = supportedLanguages[0]
} }
Log.i(
tag,
"lxm---ASR initialize derived: isAutoDetectLanguage=$isAutoDetectLanguage, currentLanguage=$currentLanguage, provider=${getAsrProvider()}, xunfeiConfigured=$xunfeiConfigReady"
)
// 创建语音配置 // 创建语音配置
speechConfig = SpeechConfig.fromSubscription(subscriptionKey, region).apply { speechConfig = SpeechConfig.fromSubscription(subscriptionKey, region).apply {
if (isAutoDetectLanguage) { if (isAutoDetectLanguage) {
@ -314,7 +333,14 @@ class AzureAsrHelper(private val context: Context) {
} }
try { try {
xunFeiAsrHelper?.start(callback, currentLanguage, isAutoDetectLanguage, isRemoveFirstPunctuation) Log.d(tag, "startContinuousRecognition: ${audioStream.isContinuousRecognitionActive}")
xunFeiAsrHelper?.start(
callback,
currentLanguage,
supportedLanguages.toList(),
isAutoDetectLanguage,
isRemoveFirstPunctuation
)
this.audioSourceType = audioSourceType this.audioSourceType = audioSourceType
if (!audioStream.isContinuousRecognitionActive) { if (!audioStream.isContinuousRecognitionActive) {

7
local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureSpeechPlugin.kt
View File

@ -423,6 +423,13 @@ class AzureSpeechPlugin : BleService.Callback, FlutterPlugin {
val xunfeiAccessKeySecret = call.argument<String>("xunfeiAccessKeySecret") ?: "" val xunfeiAccessKeySecret = call.argument<String>("xunfeiAccessKeySecret") ?: ""
try { try {
val isXunfeiConfigured = xunfeiAppId.isNotBlank() &&
xunfeiAccessKeyId.isNotBlank() &&
xunfeiAccessKeySecret.isNotBlank()
FileLogger.i(
tag,
"lxm---ASR initialize args: supportedLanguages=${supportedLanguages.joinToString(",")}, useExternalAudio=$useExternalAudio, region=$region, xunfeiConfigured=$isXunfeiConfigured"
)
FileLogger.d(tag, "选择音频源类型: ${useExternalAudio}") // FileLogger.d(tag, "选择音频源类型: ${useExternalAudio}") //
// 选择音频源类型 // 选择音频源类型
val audioSourceType = if (useExternalAudio) { val audioSourceType = if (useExternalAudio) {

76
local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/XunFeiAsrHelper.kt
View File

@ -34,6 +34,7 @@ class XunFeiAsrHelper(private val context: Context) {
private var lastIntermediateResult = "" private var lastIntermediateResult = ""
private var currentLanguage = "zh-CN" private var currentLanguage = "zh-CN"
private var supportedLanguages: List<String> = emptyList()
private var isAutoDetectLanguage = false private var isAutoDetectLanguage = false
// 是否第去除一个文字的标点符号 // 是否第去除一个文字的标点符号
@ -58,7 +59,13 @@ class XunFeiAsrHelper(private val context: Context) {
return appId.isNotBlank() && accessKeyId.isNotBlank() && accessKeySecret.isNotBlank() return appId.isNotBlank() && accessKeyId.isNotBlank() && accessKeySecret.isNotBlank()
} }
fun start(callback: AzureAsrHelper.ContinuousRecognizeCallback, language: String, isAutoDetect: Boolean, isRemoveFirstPunctuation: Boolean) { fun start(
callback: AzureAsrHelper.ContinuousRecognizeCallback,
language: String,
supportedLanguages: List<String>,
isAutoDetect: Boolean,
isRemoveFirstPunctuation: Boolean
) {
try { try {
webSocket?.cancel() webSocket?.cancel()
webSocket = null webSocket = null
@ -70,6 +77,7 @@ class XunFeiAsrHelper(private val context: Context) {
lastIntermediateResult = "" lastIntermediateResult = ""
currentLanguage = language currentLanguage = language
this.supportedLanguages = supportedLanguages
isAutoDetectLanguage = isAutoDetect isAutoDetectLanguage = isAutoDetect
removeFirstPunctuation = isRemoveFirstPunctuation removeFirstPunctuation = isRemoveFirstPunctuation
@ -99,15 +107,7 @@ class XunFeiAsrHelper(private val context: Context) {
override fun onClosed(webSocket: WebSocket, code: Int, reason: String) { override fun onClosed(webSocket: WebSocket, code: Int, reason: String) {
Log.d(tag, "WebSocket Closed: $code / $reason") Log.d(tag, "WebSocket Closed: $code / $reason")
if (currentSessionId.isNotEmpty() && lastIntermediateResult.isNotEmpty()) { if (currentSessionId.isNotEmpty() && lastIntermediateResult.isNotEmpty()) {
val detectedLanguage = if (isAutoDetectLanguage) { val detectedLanguage = selectDetectedLanguage(lastIntermediateResult)
if (lastIntermediateResult.contains(Regex("[\\u4e00-\\u9fa5]"))) {
"zh-CN"
} else {
"en-US"
}
} else {
currentLanguage
}
callback.onResult(currentSessionId, lastIntermediateResult, detectedLanguage) callback.onResult(currentSessionId, lastIntermediateResult, detectedLanguage)
} }
callback.onSessionStopped(currentSessionId) callback.onSessionStopped(currentSessionId)
@ -219,15 +219,7 @@ class XunFeiAsrHelper(private val context: Context) {
return return
} }
val detectedLanguage = if (isAutoDetectLanguage) { val detectedLanguage = selectDetectedLanguage(resultTextTemp)
if (resultTextTemp.contains(Regex("[\\u4e00-\\u9fa5]"))) {
"zh-CN"
} else {
"en-US"
}
} else {
currentLanguage
}
if (st.optInt("type") == 0) { if (st.optInt("type") == 0) {
callback.onResult(currentSessionId, resultTextTemp, detectedLanguage) callback.onResult(currentSessionId, resultTextTemp, detectedLanguage)
@ -254,6 +246,52 @@ class XunFeiAsrHelper(private val context: Context) {
} }
} }
private fun normalizeLanguageCode(code: String): String {
val trimmed = code.trim()
if (trimmed.isEmpty()) return trimmed
val parts = trimmed.split("-")
val lang = parts.getOrNull(0)?.lowercase() ?: trimmed.lowercase()
val region = parts.getOrNull(1)?.uppercase()
return if (region.isNullOrEmpty()) lang else "$lang-$region"
}
private fun findSupportedExact(target: String): String? {
val targetNorm = normalizeLanguageCode(target)
return supportedLanguages.firstOrNull { normalizeLanguageCode(it) == targetNorm }
}
private fun findSupportedPrefix(prefix: String): String? {
val p = prefix.lowercase()
return supportedLanguages.firstOrNull { normalizeLanguageCode(it).startsWith(p) }
}
private fun pickPreferredChineseLanguage(): String {
return findSupportedExact("zh-CN")
?: findSupportedExact("zh-HK")
?: findSupportedExact("zh-TW")
?: findSupportedPrefix("zh")
?: currentLanguage
}
private fun pickPreferredEnglishLanguage(): String {
return findSupportedExact("en-US")
?: findSupportedPrefix("en")
?: currentLanguage
}
// 选择检测到的语言
// 如果开启了自动检测语言，根据文本内容判断是否为中文
// 如果是中文，选择中文语言；否则选择英文语言
// 如果未开启自动检测语言，返回当前语言
private fun selectDetectedLanguage(text: String): String {
if (!isAutoDetectLanguage) return currentLanguage
val hasChinese = text.contains(Regex("[\\u4e00-\\u9fa5]"))
return if (hasChinese) pickPreferredChineseLanguage() else pickPreferredEnglishLanguage()
}
private fun getWebSocketUrl(): String { private fun getWebSocketUrl(): String {
val baseWsUrl = "wss://office-api-ast-dx.iflyaisol.com/ast/communicate/v1" val baseWsUrl = "wss://office-api-ast-dx.iflyaisol.com/ast/communicate/v1"

229
local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureAsrHelper.swift
View File

@ -1,18 +1,17 @@
import Foundation
import AVFoundation import AVFoundation
import Foundation
import MicrosoftCognitiveServicesSpeech import MicrosoftCognitiveServicesSpeech
import speech
import os.log import os.log
import speech
// 移除 Network 导入，因为我们将使用独立的 NetworkStateMonitor // 移除 Network 导入，因为我们将使用独立的 NetworkStateMonitor
// import Network // import Network
// MARK: - 基于微软Azure语音服务的ASR实现 // MARK: - 基于微软Azure语音服务的ASR实现
/** /// Azure ASR Helper
* Azure ASR Helper ///
* /// 基于微软Azure语音服务的ASR实现
* 基于微软Azure语音服务的ASR实现 /// 参考文档: https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-recognize-speech
* 参考文档: https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-recognize-speech
*/
public class AzureAsrHelper: NSObject { public class AzureAsrHelper: NSObject {
private let tag = "AzureAsrHelper" private let tag = "AzureAsrHelper"
// 日志对象 // 日志对象
@ -49,7 +48,6 @@ public class AzureAsrHelper: NSObject {
private var currsessionId = "" private var currsessionId = ""
// 配置参数 // 配置参数
private var currentLanguage = "zh-CN" private var currentLanguage = "zh-CN"
private var supportedLanguages = ["zh-CN"] private var supportedLanguages = ["zh-CN"]
@ -81,10 +79,6 @@ public class AzureAsrHelper: NSObject {
// 音频处理 // 音频处理
public var audioStream: SimpleAudioReceiver? public var audioStream: SimpleAudioReceiver?
public func asrProvider() -> String { public func asrProvider() -> String {
return useXunfei ? "xunfei" : "azure" return useXunfei ? "xunfei" : "azure"
} }
@ -134,14 +128,42 @@ public class AzureAsrHelper: NSObject {
} }
// 检查是否只包含中文或英文，如果是则使用讯飞 // 检查是否只包含中文或英文，如果是则使用讯飞
let xunfeiConfigReady = !self.xunfeiAppId.isEmpty && let xunfeiConfigReady =
!self.xunfeiAccessKeyId.isEmpty && !self.xunfeiAppId.isEmpty && !self.xunfeiAccessKeyId.isEmpty
!self.xunfeiAccessKeySecret.isEmpty && !self.xunfeiAccessKeySecret.isEmpty
self.useXunfei = supportedLanguages.allSatisfy { $0 == "zh-CN" || $0 == "en-US" || $0 == "zh-HK" || $0 == "en-TW" } && xunfeiConfigReady let onlySupportedByXunfei = supportedLanguages.allSatisfy {
$0 == "zh-CN" || $0 == "en-US" || $0 == "zh-HK" || $0 == "zh-TW"
}
let chineseVariants = Set(
supportedLanguages
.map { $0.trimmingCharacters(in: .whitespacesAndNewlines).lowercased() }
.filter { $0 == "zh" || $0.hasPrefix("zh-") }
)
let hasMultipleChineseVariants = chineseVariants.count >= 2
self.useXunfei =
onlySupportedByXunfei && xunfeiConfigReady && !hasMultipleChineseVariants
os_log(
"ASR helper initialize: supportedLanguages=%{public}@, audioSourceType=%{public}@, xunfeiConfigured=%{public}@, provider=%{public}@",
log: log,
type: .info,
supportedLanguages.joined(separator: ","),
String(describing: audioSourceType),
String(xunfeiConfigReady),
asrProvider()
)
// 根据支持的语言数量决定是否启用自动语言检测 // 根据支持的语言数量决定是否启用自动语言检测
self.isAutoDetectLanguage = supportedLanguages.count >= 2 self.isAutoDetectLanguage = supportedLanguages.count >= 2
os_log(
"ASR helper initialize derived: isAutoDetectLanguage=%{public}@, currentLanguage=%{public}@",
log: log,
type: .info,
String(self.isAutoDetectLanguage),
self.currentLanguage
)
// 如果只有一种语言，设置为当前语言 // 如果只有一种语言，设置为当前语言
if !isAutoDetectLanguage && !supportedLanguages.isEmpty { if !isAutoDetectLanguage && !supportedLanguages.isEmpty {
self.currentLanguage = supportedLanguages[0] self.currentLanguage = supportedLanguages[0]
@ -152,9 +174,12 @@ public class AzureAsrHelper: NSObject {
// 针对翻译场景的性能优化 // 针对翻译场景的性能优化
// optimizeForTranslation() // optimizeForTranslation()
if isAutoDetectLanguage { if isAutoDetectLanguage {
speechConfig?.setPropertyTo("Continuous", by: SPXPropertyId.speechServiceConnectionLanguageIdMode) speechConfig?.setPropertyTo(
"Continuous", by: SPXPropertyId.speechServiceConnectionLanguageIdMode)
// 不在这里设置语言，让SPXAutoDetectSourceLanguageConfiguration处理 // 不在这里设置语言，让SPXAutoDetectSourceLanguageConfiguration处理
os_log("启用语言检测模式，支持语言: %{public}@", log: log, type: .info, supportedLanguages.joined(separator: ",")) os_log(
"启用语言检测模式，支持语言: %{public}@", log: log, type: .info,
supportedLanguages.joined(separator: ","))
} else { } else {
speechConfig?.speechRecognitionLanguage = currentLanguage speechConfig?.speechRecognitionLanguage = currentLanguage
os_log("固定语言模式，当前语言: %{public}@", log: log, type: .info, currentLanguage) os_log("固定语言模式，当前语言: %{public}@", log: log, type: .info, currentLanguage)
@ -188,7 +213,6 @@ public class AzureAsrHelper: NSObject {
private func handleNetworkAvailable() { private func handleNetworkAvailable() {
os_log("网络已恢复", log: log, type: .info) os_log("网络已恢复", log: log, type: .info)
if !self.isNetworkRecovering { if !self.isNetworkRecovering {
self.isNetworkRecovering = true self.isNetworkRecovering = true
os_log("网络恢复，准备重新启动识别", log: self.log, type: .info) os_log("网络恢复，准备重新启动识别", log: self.log, type: .info)
@ -196,7 +220,6 @@ public class AzureAsrHelper: NSObject {
} }
/** /**
* 处理网络连接丢失（带防抖机制） * 处理网络连接丢失（带防抖机制）
*/ */
@ -205,7 +228,9 @@ public class AzureAsrHelper: NSObject {
// 防抖机制：如果距离上次触发时间小于防抖间隔，则忽略本次触发 // 防抖机制：如果距离上次触发时间小于防抖间隔，则忽略本次触发
if currentTime - lastNetworkLostTime < networkLostDebounceInterval { if currentTime - lastNetworkLostTime < networkLostDebounceInterval {
os_log("网络丢失事件被防抖过滤，距离上次触发仅%{public}dms", log: log, type: .debug, currentTime - lastNetworkLostTime) os_log(
"网络丢失事件被防抖过滤，距离上次触发仅%{public}dms", log: log, type: .debug,
currentTime - lastNetworkLostTime)
return return
} }
@ -225,7 +250,7 @@ public class AzureAsrHelper: NSObject {
self.lastNetworkLostTime = 0 self.lastNetworkLostTime = 0
self.isNetworkRecovering = true self.isNetworkRecovering = true
} else { } else {
self.continuousCallback?.onError(sessionid: currsessionId,1000, "网络连接不可用，请检查网络设置") self.continuousCallback?.onError(sessionid: currsessionId, 1000, "网络连接不可用，请检查网络设置")
os_log("网络仍未恢复，继续检测", log: self.log, type: .debug) os_log("网络仍未恢复，继续检测", log: self.log, type: .debug)
self.networkLostCount = 0 self.networkLostCount = 0
self.stopContinuousRecognition() self.stopContinuousRecognition()
@ -233,7 +258,6 @@ public class AzureAsrHelper: NSObject {
} }
} }
/** /**
* 预初始化音频组件 * 预初始化音频组件
* 在初始化阶段就准备好音频流和相关组件，减少首次启动延迟 * 在初始化阶段就准备好音频流和相关组件，减少首次启动延迟
@ -251,7 +275,6 @@ public class AzureAsrHelper: NSObject {
} }
/** /**
* 异步启动音频录制（带防抖与状态二次校验） * 异步启动音频录制（带防抖与状态二次校验）
* *
@ -273,13 +296,15 @@ public class AzureAsrHelper: NSObject {
audioSourceType: AudioSourceType, audioSourceType: AudioSourceType,
audioDataCallback: SimpleAudioReceiver.AudioDataCallback? = nil, audioDataCallback: SimpleAudioReceiver.AudioDataCallback? = nil,
completion: @escaping (Bool) -> Void completion: @escaping (Bool) -> Void
) { ) {
let currentTime = Int64(Date().timeIntervalSince1970 * 1000) let currentTime = Int64(Date().timeIntervalSince1970 * 1000)
// 防抖：距离上次启动过短，取消旧任务 // 防抖：距离上次启动过短，取消旧任务
if currentTime - lastAudioStartTime < audioOperationDebounceInterval { if currentTime - lastAudioStartTime < audioOperationDebounceInterval {
audioStartWorkItem?.cancel() audioStartWorkItem?.cancel()
os_log("音频启动操作被防抖过滤，距离上次启动仅%{public}dms", log: log, type: .debug, currentTime - lastAudioStartTime) os_log(
"音频启动操作被防抖过滤，距离上次启动仅%{public}dms", log: log, type: .debug,
currentTime - lastAudioStartTime)
} else { } else {
// 即使未触发防抖，也保证旧任务不会残留 // 即使未触发防抖，也保证旧任务不会残留
audioStartWorkItem?.cancel() audioStartWorkItem?.cancel()
@ -312,7 +337,9 @@ public class AzureAsrHelper: NSObject {
// 准备启动 // 准备启动
self.isAudioStarting = true self.isAudioStarting = true
self.pendingStopRequest = false // 启动前清空旧的停止请求 self.pendingStopRequest = false // 启动前清空旧的停止请求
let success = self.performAudioStart(mod: mod, audioSourceType: audioSourceType, audioDataCallback: audioDataCallback) let success = self.performAudioStart(
mod: mod, audioSourceType: audioSourceType, audioDataCallback: audioDataCallback
)
self.isAudioStarting = false self.isAudioStarting = false
self.isAudioStarted = success self.isAudioStarted = success
@ -339,7 +366,7 @@ public class AzureAsrHelper: NSObject {
deadline: .now() + .milliseconds(Int(audioOperationDebounceInterval)), deadline: .now() + .milliseconds(Int(audioOperationDebounceInterval)),
execute: workItem execute: workItem
) )
} }
/** /**
* 执行音频启动操作 * 执行音频启动操作
@ -350,7 +377,7 @@ public class AzureAsrHelper: NSObject {
audioDataCallback: SimpleAudioReceiver.AudioDataCallback? audioDataCallback: SimpleAudioReceiver.AudioDataCallback?
) -> Bool { ) -> Bool {
do { do {
if(!audioStream!.isContinuousRecognitionActive){ if !audioStream!.isContinuousRecognitionActive {
audioStream?.startAudioRecord( audioStream?.startAudioRecord(
mod: mod, mod: mod,
audioSourceType: audioSourceType == .microphone ? .microphone : .external, audioSourceType: audioSourceType == .microphone ? .microphone : .external,
@ -423,7 +450,7 @@ public class AzureAsrHelper: NSObject {
audioSourceType: AudioSourceType = .microphone, audioSourceType: AudioSourceType = .microphone,
audioDataCallback: SimpleAudioReceiver.AudioDataCallback? = nil, audioDataCallback: SimpleAudioReceiver.AudioDataCallback? = nil,
isRemoveFirstPunctuation: Bool = true isRemoveFirstPunctuation: Bool = true
) -> Bool { ) -> Bool {
guard speechConfig != nil else { guard speechConfig != nil else {
print("语音服务未初始化") print("语音服务未初始化")
return false return false
@ -469,15 +496,23 @@ public class AzureAsrHelper: NSObject {
} }
// 启动讯飞识别 // 启动讯飞识别
xunFeiAsrHelper?.start(callback: callback, language: currentLanguage, isAutoDetect: isAutoDetectLanguage, isRemoveFirstPunctuation: isRemoveFirstPunctuation) xunFeiAsrHelper?.start(
callback: callback,
language: currentLanguage,
supportedLanguages: supportedLanguages,
isAutoDetect: isAutoDetectLanguage,
isRemoveFirstPunctuation: isRemoveFirstPunctuation)
// 创建代理回调，将音频数据转发给讯飞 // 创建代理回调，将音频数据转发给讯飞
let proxyCallback = AudioDataCallbackProxy(original: audioDataCallback) { [weak self] data in let proxyCallback = AudioDataCallbackProxy(original: audioDataCallback) {
[weak self] data in
self?.xunFeiAsrHelper?.sendAudio(frameBuffer: data) self?.xunFeiAsrHelper?.sendAudio(frameBuffer: data)
} }
// 异步启动音频处理 // 异步启动音频处理
startAudioRecordAsync(mod: mod, audioSourceType: audioSourceType, audioDataCallback: proxyCallback) { [weak self] (success: Bool) in startAudioRecordAsync(
mod: mod, audioSourceType: audioSourceType, audioDataCallback: proxyCallback
) { [weak self] (success: Bool) in
guard let self = self else { return } guard let self = self else { return }
if success { if success {
@ -506,20 +541,28 @@ public class AzureAsrHelper: NSObject {
// self.audioStream?.isContinuousRecognitionActive = true // self.audioStream?.isContinuousRecognitionActive = true
// 异步启动音频处理 // 异步启动音频处理
startAudioRecordAsync(mod: mod, audioSourceType: audioSourceType, audioDataCallback: audioDataCallback) { [weak self] (success: Bool) in startAudioRecordAsync(
mod: mod, audioSourceType: audioSourceType, audioDataCallback: audioDataCallback
) { [weak self] (success: Bool) in
guard let self = self else { return } guard let self = self else { return }
if success { if success {
// 只有在音频启动成功后才设置状态为 true // 只有在音频启动成功后才设置状态为 true
self.audioStream?.isContinuousRecognitionActive = true self.audioStream?.isContinuousRecognitionActive = true
print("startContinuousRecognition:\(self.audioStream?.isContinuousRecognitionActive)") print(
"startContinuousRecognition:\(self.audioStream?.isContinuousRecognitionActive)"
)
// 启动语音识别 // 启动语音识别
do { do {
try self.recognizer?.startContinuousRecognition() try self.recognizer?.startContinuousRecognition()
os_log("连续识别已启动，音频源: %{public}@", log: self.log, type: .info, audioSourceType == .microphone ? "麦克风" : "外部") os_log(
"连续识别已启动，音频源: %{public}@", log: self.log, type: .info,
audioSourceType == .microphone ? "麦克风" : "外部")
} catch { } catch {
os_log("启动连续识别失败: %{public}@", log: self.log, type: .error, error.localizedDescription) os_log(
"启动连续识别失败: %{public}@", log: self.log, type: .error,
error.localizedDescription)
self.stopContinuousRecognition() self.stopContinuousRecognition()
self.audioStream?.isContinuousRecognitionActive = false self.audioStream?.isContinuousRecognitionActive = false
} }
@ -535,9 +578,7 @@ public class AzureAsrHelper: NSObject {
audioStream?.isContinuousRecognitionActive = false audioStream?.isContinuousRecognitionActive = false
return false return false
} }
} }
/** /**
* 停止连续语音识别（优化：确保取消未执行的启动任务） * 停止连续语音识别（优化：确保取消未执行的启动任务）
@ -566,7 +607,6 @@ public class AzureAsrHelper: NSObject {
return false return false
} }
do { do {
// 关键：取消尚未执行的启动任务，防止 stop 后音频又被启动 // 关键：取消尚未执行的启动任务，防止 stop 后音频又被启动
@ -698,7 +738,6 @@ public class AzureAsrHelper: NSObject {
} }
} }
/** /**
* 设置识别器 * 设置识别器
*/ */
@ -734,7 +773,10 @@ public class AzureAsrHelper: NSObject {
// 创建识别器 // 创建识别器
if isAutoDetectLanguage { if isAutoDetectLanguage {
// 使用正确的初始化方法 // 使用正确的初始化方法
guard let autoDetectSourceLanguageConfig = try? SPXAutoDetectSourceLanguageConfiguration(supportedLanguages) else { guard
let autoDetectSourceLanguageConfig =
try? SPXAutoDetectSourceLanguageConfiguration(supportedLanguages)
else {
print("创建自动语言检测配置失败，使用默认语言配置") print("创建自动语言检测配置失败，使用默认语言配置")
// 如果创建失败，使用默认的单语言配置 // 如果创建失败，使用默认的单语言配置
recognizer = try SPXSpeechRecognizer( recognizer = try SPXSpeechRecognizer(
@ -764,7 +806,6 @@ public class AzureAsrHelper: NSObject {
} }
} }
/** /**
* 设置麦克风流 - 使用推流方式 * 设置麦克风流 - 使用推流方式
* 优化：避免重复初始化已存在的音频流 * 优化：避免重复初始化已存在的音频流
@ -787,8 +828,6 @@ public class AzureAsrHelper: NSObject {
} }
} }
/** /**
* 从语音识别结果中获取检测到的语言 * 从语音识别结果中获取检测到的语言
* @param result 语音识别结果 * @param result 语音识别结果
@ -803,8 +842,10 @@ public class AzureAsrHelper: NSObject {
// 备用方案：从属性中获取语言 // 备用方案：从属性中获取语言
if let properties = result.properties, if let properties = result.properties,
let language = properties.getPropertyByName("speechServiceResponse_RecognitionLanguage"), let language = properties.getPropertyByName(
!language.isEmpty { "speechServiceResponse_RecognitionLanguage"),
!language.isEmpty
{
print("从属性中检测到语言: \(language)") print("从属性中检测到语言: \(language)")
return language return language
} }
@ -823,8 +864,6 @@ public class AzureAsrHelper: NSObject {
return currentLanguage return currentLanguage
} }
/** /**
* 动态切换语言（翻译场景中的语言切换） * 动态切换语言（翻译场景中的语言切换）
* *
@ -857,7 +896,9 @@ public class AzureAsrHelper: NSObject {
} }
} }
os_log("语言已切换: %{public}@ -> %{public}@", log: log, type: .info, validatedSourceLang, validatedTargetLang) os_log(
"语言已切换: %{public}@ -> %{public}@", log: log, type: .info, validatedSourceLang,
validatedTargetLang)
return true return true
} }
/** /**
@ -870,7 +911,7 @@ public class AzureAsrHelper: NSObject {
self.continuousCallback = callback self.continuousCallback = callback
// 重设识别器 // 重设识别器
if (!setupRecognizer()) { if !setupRecognizer() {
return false return false
} }
guard let recognizer = recognizer else { return false } guard let recognizer = recognizer else { return false }
@ -888,7 +929,8 @@ public class AzureAsrHelper: NSObject {
print("正在识别事件=检测到语言: \(detectedLanguage), 识别中: \(text)") print("正在识别事件=检测到语言: \(detectedLanguage), 识别中: \(text)")
DispatchQueue.main.async { DispatchQueue.main.async {
callback.onRecognizing(sessionid: self.currsessionId,result.text ?? "", detectedLanguage) callback.onRecognizing(
sessionid: self.currsessionId, result.text ?? "", detectedLanguage)
} }
} }
@ -897,7 +939,8 @@ public class AzureAsrHelper: NSObject {
guard let self = self else { return } guard let self = self else { return }
let result = event.result let result = event.result
guard result.reason == .recognizedSpeech, guard result.reason == .recognizedSpeech,
!(result.text?.isEmpty ?? true) else { return } !(result.text?.isEmpty ?? true)
else { return }
// 使用现有的 getDetectedLanguage 方法替代 result.language // 使用现有的 getDetectedLanguage 方法替代 result.language
let detectedLanguage = self.getDetectedLanguage(from: result) let detectedLanguage = self.getDetectedLanguage(from: result)
@ -907,7 +950,8 @@ public class AzureAsrHelper: NSObject {
// 在主线程回调结果 // 在主线程回调结果
DispatchQueue.main.async { DispatchQueue.main.async {
callback.onResult(sessionid: self.currsessionId,result.text ?? "", detectedLanguage) callback.onResult(
sessionid: self.currsessionId, result.text ?? "", detectedLanguage)
self.currsessionId = UUID().uuidString self.currsessionId = UUID().uuidString
} }
@ -939,14 +983,12 @@ public class AzureAsrHelper: NSObject {
os_log("识别取消: %{public}@", log: self.log, type: .error, errorDetails) os_log("识别取消: %{public}@", log: self.log, type: .error, errorDetails)
callback.onCanceled(sessionid: self.currsessionId, reason, errorDetails)
callback.onCanceled(sessionid: self.currsessionId,reason, errorDetails)
} }
return true return true
} }
/** /**
* 禁用耳机进行拾音 * 禁用耳机进行拾音
*/ */
@ -980,10 +1022,13 @@ public class AzureAsrHelper: NSObject {
* @param filePath 录音文件路径 * @param filePath 录音文件路径
* @param audioDataCallback 音频数据回调接口 * @param audioDataCallback 音频数据回调接口
*/ */
public func enableRecord( audioSourceType: AudioSourceType = .microphone,filePath: String, audioDataCallback: SimpleAudioReceiver.AudioDataCallback? = nil) { public func enableRecord(
audioSourceType: AudioSourceType = .microphone, filePath: String,
audioDataCallback: SimpleAudioReceiver.AudioDataCallback? = nil
) {
os_log("开启录音: %{public}@", log: log, type: .info, filePath) os_log("开启录音: %{public}@", log: log, type: .info, filePath)
self.audioSourceType = audioSourceType self.audioSourceType = audioSourceType
if(audioSourceType == .external){ if audioSourceType == .external {
os_log("外部音频源不在这里录音", log: log, type: .info) os_log("外部音频源不在这里录音", log: log, type: .info)
return return
} }
@ -997,12 +1042,14 @@ public class AzureAsrHelper: NSObject {
audioStream?.recordfile = RecordFile() audioStream?.recordfile = RecordFile()
} }
// 修复：移除多余的 audioDataCallback 参数 // 修复：移除多余的 audioDataCallback 参数
// 修复：添加类型转换 // 修复：添加类型转换
if(!audioStream!.isContinuousRecognitionActive){ if !audioStream!.isContinuousRecognitionActive {
// 异步启动音频处理 // 异步启动音频处理
startAudioRecordAsync(mod: "normal", audioSourceType: audioSourceType, audioDataCallback: audioDataCallback) { [weak self] (success: Bool) in startAudioRecordAsync(
mod: "normal", audioSourceType: audioSourceType,
audioDataCallback: audioDataCallback
) { [weak self] (success: Bool) in
guard let self = self else { return } guard let self = self else { return }
} }
@ -1015,9 +1062,9 @@ public class AzureAsrHelper: NSObject {
/** /**
* 移动文件到新路径 * 移动文件到新路径
*/ */
public func moveFile(sourcePath: String, destPath: String)-> Bool { public func moveFile(sourcePath: String, destPath: String) -> Bool {
if(audioSourceType == .external){ if audioSourceType == .external {
os_log(" 外部音频源不在这里移动文件", log: log, type: .info) os_log(" 外部音频源不在这里移动文件", log: log, type: .info)
return true return true
} }
@ -1025,12 +1072,11 @@ public class AzureAsrHelper: NSObject {
return true return true
} }
/** /**
* 重命名指定路径的音频文件 * 重命名指定路径的音频文件
*/ */
public func renameFile(filePath: String, newName: String)-> Bool { public func renameFile(filePath: String, newName: String) -> Bool {
if(audioSourceType == .external){ if audioSourceType == .external {
os_log(" 外部音频源不在这里重命名文件", log: log, type: .info) os_log(" 外部音频源不在这里重命名文件", log: log, type: .info)
return true return true
} }
@ -1038,13 +1084,11 @@ public class AzureAsrHelper: NSObject {
return true return true
} }
/** /**
* 暂停录音 * 暂停录音
*/ */
public func pauseRecord() { public func pauseRecord() {
if(audioSourceType == .external){ if audioSourceType == .external {
os_log(" 外部音频源不在这里暂停录音", log: log, type: .info) os_log(" 外部音频源不在这里暂停录音", log: log, type: .info)
return return
} }
@ -1059,7 +1103,7 @@ public class AzureAsrHelper: NSObject {
* 继续录音 * 继续录音
*/ */
public func resumeRecord() { public func resumeRecord() {
if(audioSourceType == .external){ if audioSourceType == .external {
os_log(" 外部音频源不在这里继续录音", log: log, type: .info) os_log(" 外部音频源不在这里继续录音", log: log, type: .info)
return return
} }
@ -1074,7 +1118,7 @@ public class AzureAsrHelper: NSObject {
* 关闭录音 * 关闭录音
*/ */
public func stopRecord(isSave: Bool) { public func stopRecord(isSave: Bool) {
if(audioSourceType == .external){ if audioSourceType == .external {
os_log(" 外部音频源不在这里关闭录音", log: log, type: .info) os_log(" 外部音频源不在这里关闭录音", log: log, type: .info)
return return
} }
@ -1082,17 +1126,13 @@ public class AzureAsrHelper: NSObject {
return return
} }
audioStream.isRecord = false audioStream.isRecord = false
if (!audioStream.isContinuousRecognitionActive) { if !audioStream.isContinuousRecognitionActive {
audioStream.stopMicrophoneCapture() audioStream.stopMicrophoneCapture()
} }
audioStream.recordfile?.isPause = false // audioStream.recordfile?.isPause = false //
audioStream.recordfile?.closeFile(isSave: true) // audioStream.recordfile?.closeFile(isSave: true) //
} }
// MARK: - 回调函数 // MARK: - 回调函数
/** /**
* 连续识别回调接口 * 连续识别回调接口
@ -1104,7 +1144,7 @@ public class AzureAsrHelper: NSObject {
* @param text 识别的文本 * @param text 识别的文本
* @param detectedLanguage 检测到的语言 * @param detectedLanguage 检测到的语言
*/ */
func onResult(sessionid:String,_ text: String, _ detectedLanguage: String) func onResult(sessionid: String, _ text: String, _ detectedLanguage: String)
/** /**
* 识别进行中调用 * 识别进行中调用
@ -1112,17 +1152,17 @@ public class AzureAsrHelper: NSObject {
* @param recognizing 正在识别的文本 * @param recognizing 正在识别的文本
* @param detectedLanguage 检测到的语言 * @param detectedLanguage 检测到的语言
*/ */
func onRecognizing(sessionid:String ,_ recognizing: String, _ detectedLanguage: String) func onRecognizing(sessionid: String, _ recognizing: String, _ detectedLanguage: String)
/** /**
* 会话开始时调用 * 会话开始时调用
*/ */
func onSessionStarted(sessionid:String) func onSessionStarted(sessionid: String)
/** /**
* 会话结束时调用 * 会话结束时调用
*/ */
func onSessionStopped(sessionid:String) func onSessionStopped(sessionid: String)
/** /**
* 识别取消时调用 * 识别取消时调用
@ -1130,46 +1170,46 @@ public class AzureAsrHelper: NSObject {
* @param reason 取消原因 * @param reason 取消原因
* @param errorDetails 错误详情 * @param errorDetails 错误详情
*/ */
func onCanceled(sessionid:String ,_ reason: String, _ errorDetails: String) func onCanceled(sessionid: String, _ reason: String, _ errorDetails: String)
/** /**
* 识别出错时调用 * 识别出错时调用
* *
* @param error 错误信息 * @param error 错误信息
*/ */
func onError(sessionid:String ,_ code: Int,_ error: String) func onError(sessionid: String, _ code: Int, _ error: String)
} }
/** /**
* 检查网络状态 * 检查网络状态
* @return 网络是否可用 * @return 网络是否可用
*/ */
private func checkNetworkStatus() -> Bool { private func checkNetworkStatus() -> Bool {
// 直接使用 NetworkStateMonitor 的 checkNetworkStatus 方法 // 直接使用 NetworkStateMonitor 的 checkNetworkStatus 方法
let isAvailable = networkMonitor.checkNetworkStatus() let isAvailable = networkMonitor.checkNetworkStatus()
if !isAvailable { if !isAvailable {
os_log("网络不可用，无法启动语音识别", log: log, type: .error) os_log("网络不可用，无法启动语音识别", log: log, type: .error)
continuousCallback?.onError(sessionid: self.currsessionId,1000,"网络连接不可用，请检查网络设置") continuousCallback?.onError(sessionid: self.currsessionId, 1000, "网络连接不可用，请检查网络设置")
} }
return isAvailable return isAvailable
} }
/** /**
* 判断是否为网络相关错误 * 判断是否为网络相关错误
* @param errorDetails 错误详情 * @param errorDetails 错误详情
* @return 是否为网络错误 * @return 是否为网络错误
*/ */
private func isNetworkRelatedError(_ errorDetails: String) -> Bool { private func isNetworkRelatedError(_ errorDetails: String) -> Bool {
let networkErrorKeywords = [ let networkErrorKeywords = [
"network", "connection", "timeout", "unreachable", "network", "connection", "timeout", "unreachable",
"网络", "连接", "超时", "不可达", "网络", "连接", "超时", "不可达",
"ConnectionFailure", "NetworkError", "TimeoutError" "ConnectionFailure", "NetworkError", "TimeoutError",
] ]
let lowercaseError = errorDetails.lowercased() let lowercaseError = errorDetails.lowercased()
return networkErrorKeywords.contains { lowercaseError.contains($0.lowercased()) } return networkErrorKeywords.contains { lowercaseError.contains($0.lowercased()) }
} }
} }
// MARK: - NetworkStateListener实现 // MARK: - NetworkStateListener实现
@ -1184,8 +1224,6 @@ extension AzureAsrHelper: NetworkStateMonitor.NetworkStateListener {
} }
func onNetworkLost() { func onNetworkLost() {
os_log("网络连接丢失", log: log, type: .error) os_log("网络连接丢失", log: log, type: .error)
@ -1215,6 +1253,3 @@ private class AudioDataCallbackProxy: SimpleAudioReceiver.AudioDataCallback {
xunfeiCallback?(data) xunfeiCallback?(data)
} }
} }

10
local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureSpeechPlugin.swift
View File

@ -343,6 +343,16 @@ private func sendAstEvent(_ event: [String: Any]) {
let xunfeiAppId = args["xunfeiAppId"] as? String ?? "" let xunfeiAppId = args["xunfeiAppId"] as? String ?? ""
let xunfeiAccessKeyId = args["xunfeiAccessKeyId"] as? String ?? "" let xunfeiAccessKeyId = args["xunfeiAccessKeyId"] as? String ?? ""
let xunfeiAccessKeySecret = args["xunfeiAccessKeySecret"] as? String ?? "" let xunfeiAccessKeySecret = args["xunfeiAccessKeySecret"] as? String ?? ""
let isXunfeiConfigured = !xunfeiAppId.isEmpty && !xunfeiAccessKeyId.isEmpty && !xunfeiAccessKeySecret.isEmpty
os_log(
"ASR initialize args: supportedLanguages=%{public}@, region=%{public}@, xunfeiConfigured=%{public}@",
log: log,
type: .info,
supportedLanguages.joined(separator: ","),
region,
String(isXunfeiConfigured)
)
// 初始化ASR引擎 // 初始化ASR引擎
let success = azureAsrHelper.initialize( let success = azureAsrHelper.initialize(

47
local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/XunFeiAsrHelper.swift
View File

@ -23,6 +23,7 @@ class XunFeiAsrHelper: NSObject, URLSessionWebSocketDelegate {
private var lastIntermediateResult = "" private var lastIntermediateResult = ""
private var currentLanguage = "zh-CN" private var currentLanguage = "zh-CN"
private var supportedLanguages: [String] = []
private var isAutoDetectLanguage = false private var isAutoDetectLanguage = false
// 是否第去除一个文字的标点符号 // 是否第去除一个文字的标点符号
@ -47,11 +48,12 @@ class XunFeiAsrHelper: NSObject, URLSessionWebSocketDelegate {
return !appId.isEmpty && !accessKeyId.isEmpty && !accessKeySecret.isEmpty return !appId.isEmpty && !accessKeyId.isEmpty && !accessKeySecret.isEmpty
} }
func start(callback: AzureAsrHelper.ContinuousRecognizeCallback, language: String, isAutoDetect: Bool, isRemoveFirstPunctuation: Bool) { func start(callback: AzureAsrHelper.ContinuousRecognizeCallback, language: String, supportedLanguages: [String], isAutoDetect: Bool, isRemoveFirstPunctuation: Bool) {
stop() stop()
self.callback = callback self.callback = callback
self.currentLanguage = language self.currentLanguage = language
self.supportedLanguages = supportedLanguages
self.isAutoDetectLanguage = isAutoDetect self.isAutoDetectLanguage = isAutoDetect
self.removeFirstPunctuation = isRemoveFirstPunctuation self.removeFirstPunctuation = isRemoveFirstPunctuation
self.currentSessionId = UUID().uuidString self.currentSessionId = UUID().uuidString
@ -268,16 +270,43 @@ class XunFeiAsrHelper: NSObject, URLSessionWebSocketDelegate {
} }
} }
private func detectLanguage(from text: String) -> String { private func normalizeLanguageCode(_ code: String) -> String {
if isAutoDetectLanguage { let trimmed = code.trimmingCharacters(in: .whitespacesAndNewlines)
if text.range(of: "[\\u4e00-\\u9fa5]", options: .regularExpression) != nil { if trimmed.isEmpty { return trimmed }
return "zh-CN" let parts = trimmed.split(separator: "-")
} else { let lang = parts.first?.lowercased() ?? trimmed.lowercased()
return "en-US" let region = parts.count > 1 ? parts[1].uppercased() : ""
return region.isEmpty ? lang : "\(lang)-\(region)"
} }
} else {
return currentLanguage private func findSupportedExact(_ target: String) -> String? {
let targetNorm = normalizeLanguageCode(target)
return supportedLanguages.first { normalizeLanguageCode($0) == targetNorm }
}
private func findSupportedPrefix(_ prefix: String) -> String? {
let p = prefix.lowercased()
return supportedLanguages.first { normalizeLanguageCode($0).hasPrefix(p) }
} }
private func pickPreferredChineseLanguage() -> String {
return findSupportedExact("zh-CN")
?? findSupportedExact("zh-HK")
?? findSupportedExact("zh-TW")
?? findSupportedPrefix("zh")
?? currentLanguage
}
private func pickPreferredEnglishLanguage() -> String {
return findSupportedExact("en-US")
?? findSupportedPrefix("en")
?? currentLanguage
}
private func detectLanguage(from text: String) -> String {
guard isAutoDetectLanguage else { return currentLanguage }
let hasChinese = text.range(of: "[\\u4e00-\\u9fa5]", options: .regularExpression) != nil
return hasChinese ? pickPreferredChineseLanguage() : pickPreferredEnglishLanguage()
} }
private func getWebSocketUrl() -> String { private func getWebSocketUrl() -> String {

Write Preview
Loading…
Cancel
Save