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 5c90a7c67..3106c1236 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 @@ -100,14 +100,9 @@ public class AzureAsrHelper: NSObject { // 创建语音配置 speechConfig = try SPXSpeechConfiguration(subscription: subscriptionKey, region: region) // 针对翻译场景的性能优化 - optimizeForTranslation() + // optimizeForTranslation() if isAutoDetectLanguage { - // 针对翻译场景优化的语言检测配置 speechConfig?.setPropertyTo("Continuous", byName: "SpeechServiceConnection_LanguageIdMode") - - // 设置语言检测的置信度阈值，提高准确性 - speechConfig?.setPropertyTo("0.7", byName: "SpeechServiceConnection_AutoDetectSourceLanguageConfidenceThreshold") - // 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: ",")) } else { @@ -150,55 +145,7 @@ public class AzureAsrHelper: NSObject { } } - /** - * 检测输入文本的语言，从支持的语言列表中选择最匹配的 - * - * @param text 需要检测的文本 - * @return 从supportedLanguages中选择的最匹配语言代码 - */ - public func detectTextLanguage(_ text: String) -> String? { - // 如果只有一种支持的语言，直接返回 - if supportedLanguages.count == 1 { - return supportedLanguages[0] - } - - // 获取所有可能的语言及其置信度 - let languageResults = LanguageDetector.detectLanguageWithConfidence(from: text) - - // 遍历检测结果，找到第一个在supportedLanguages中的语言 - for result in languageResults { - let azureLanguageCode = LanguageDetector.convertToAzureLanguageCode(result.language) - if supportedLanguages.contains(azureLanguageCode) { - return azureLanguageCode - } - } - - // 如果没有找到匹配的语言，返回第一个支持的语言作为默认值 - return supportedLanguages.first - } - - /** - * 检测文本语言并返回详细信息 - * - * @param text 需要检测的文本 - * @return 包含语言代码、置信度和语言名称的字典 - */ - public func detectTextLanguageWithDetails(_ text: String) -> [String: Any]? { - let results = LanguageDetector.detectLanguageWithConfidence(from: text) - guard let topResult = results.first else { - return nil - } - - let azureLanguageCode = LanguageDetector.convertToAzureLanguageCode(topResult.language) - let languageName = LanguageDetector.getLanguageName(from: topResult.language) - - return [ - "languageCode": azureLanguageCode, - "confidence": topResult.confidence, - "languageName": languageName, - "originalCode": topResult.language - ] - } + /** * 开始连续语音识别 * @@ -249,28 +196,7 @@ public class AzureAsrHelper: NSObject { return false } } - // /** - // * 获取检测到的语言 - // */ - // private func getDetectedLanguage(from result: SPXSpeechRecognitionResult) -> String { - // if !isAutoDetectLanguage { - // return "" - // } - - // // 尝试从属性中获取语言 - // if let properties = result.properties, - // let language = properties.getPropertyByName("SpeechServiceResponse_RecognitionLanguage") { - // return language - // } - - // // 尝试另一种方式获取语言 - // do { - // let langResult = try SPXAutoDetectSourceLanguageResult(result) - // return langResult.language ?? "" - // } catch { - // return "" - // } - // } + /** * 停止连续语音识别 @@ -388,11 +314,20 @@ public class AzureAsrHelper: NSObject { // 创建识别器 if isAutoDetectLanguage { // 使用正确的初始化方法 - let autoDetectConfig = try SPXAutoDetectSourceLanguageConfiguration(supportedLanguages) - print("autoDetectConfig: \(autoDetectConfig)") + 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") + // 如果创建失败，使用默认的单语言配置 + recognizer = try SPXSpeechRecognizer( + speechConfiguration: speechConfig!, + audioConfiguration: audioConfig! + ) + return true + } + + print("autoDetectConfig: \(autoDetectSourceLanguageConfig)") recognizer = try SPXSpeechRecognizer( speechConfiguration: speechConfig!, - autoDetectSourceLanguageConfiguration: autoDetectConfig, + autoDetectSourceLanguageConfiguration: autoDetectSourceLanguageConfig, audioConfiguration: audioConfig! ) } else { @@ -484,60 +419,7 @@ public class AzureAsrHelper: NSObject { return currentLanguage } - /** - * 针对翻译场景的性能优化配置 - */ - private func optimizeForTranslation() { - guard let speechConfig = speechConfig else { return } - - // 启用实时转录，减少延迟 - speechConfig.setPropertyTo("true", byName: "SpeechServiceConnection_EnableAudioLogging") - speechConfig.setPropertyTo("INTERACTIVE", byName: "SpeechServiceConnection_RecoMode") - speechConfig.setPropertyTo("PCM", byName: "Speech_PushStreamFormat") - // 优化语音端点检测，适合翻译场景 - speechConfig.setPropertyTo("300", byName: "SpeechServiceConnection_InitialSilenceTimeoutMs") - speechConfig.setPropertyTo("500", byName: "SpeechServiceConnection_EndSilenceTimeoutMs") - speechConfig.setPropertyTo("500", byName: "Speech_SegmentationSilenceTimeoutMs") - // 启用标点符号，提高翻译质量 - speechConfig.setPropertyTo("true", byName: "SpeechServiceResponse_RequestProfanityFilterTrueFalse") - // 设置分段策略为时间模式 - speechConfig.setPropertyTo("Time", byName: "Speech_SegmentationStrategy") - - // 【优化】语言检测配置 - 减少检测延迟，实现快速响应 - if isAutoDetectLanguage { - // 降低语言检测置信度阈值，提高检测灵敏度 - speechConfig.setPropertyTo("0.3", byName: "SpeechServiceConnection_AutoDetectSourceLanguageConfidenceThreshold") - - // 使用Continuous模式，在每句话结束后重新检测语言 - speechConfig.setPropertyTo("Continuous", byName: "SpeechServiceConnection_LanguageIdMode") - - // 启用快速语言检测 - speechConfig.setPropertyTo("true", byName: "SpeechServiceConnection_EnableLanguageDetectionFast") - - // 【关键优化】设置语言检测的最小音频长度为100毫秒，实现快速响应 - speechConfig.setPropertyTo("100", byName: "SpeechServiceConnection_LanguageDetectionMinAudioLengthMs") - - // 【新增】设置语言检测的最大等待时间，避免长时间等待 - speechConfig.setPropertyTo("2000", byName: "SpeechServiceConnection_LanguageDetectionMaxWaitTimeMs") - - // 强制在每个语音段开始时重新检测语言 - speechConfig.setPropertyTo("true", byName: "SpeechServiceConnection_ForceLanguageDetectionOnNewSegment") - - // 【新增】启用句子级别的语言检测 - speechConfig.setPropertyTo("true", byName: "SpeechServiceConnection_EnableLanguageDetectionPerSentence") - - // 【新增】设置语言检测的频率 - 每个识别结果后都检测 - speechConfig.setPropertyTo("PerResult", byName: "SpeechServiceConnection_LanguageDetectionFrequency") - - // 【新增】启用实时语言检测，减少缓冲延迟 - speechConfig.setPropertyTo("true", byName: "SpeechServiceConnection_EnableRealtimeLanguageDetection") - - // 【新增】设置语言检测的采样间隔为200毫秒 - speechConfig.setPropertyTo("200", byName: "SpeechServiceConnection_LanguageDetectionSamplingIntervalMs") - } - - os_log("翻译场景优化配置已应用，包含语言检测优化", log: log, type: .info) - } + /** * 动态切换语言（翻译场景中的语言切换） @@ -588,13 +470,16 @@ public class AzureAsrHelper: NSObject { let result = event.result guard !(result.text?.isEmpty ?? true) else { return } - // let detectedLanguage = self.getDetectedLanguage(from: result) - - // 检测文本语言，避免重复调用 - let detectedLanguage = self.detectTextLanguage(result.text ?? "") ?? "zh-CN" + // 使用现有的 getDetectedLanguage 方法替代 result.language + let detectedLanguage = self.getDetectedLanguage(from: result) + let text = result.text + + print("正在识别事件=检测到语言: \(detectedLanguage), 识别中: \(text)") + // // 检测文本语言，避免重复调用 + // let detectedLanguage = self.detectTextLanguage(result.text ?? "") ?? "zh-CN" - // 输出检测到的语言信息 - print("检测到的语言: \(detectedLanguage)") + // // 输出检测到的语言信息 + // print("检测到的语言: \(detectedLanguage)") DispatchQueue.main.async { callback.onRecognizing(result.text ?? "", detectedLanguage) @@ -608,11 +493,16 @@ public class AzureAsrHelper: NSObject { guard result.reason == .recognizedSpeech, !(result.text?.isEmpty ?? true) else { return } + // 使用现有的 getDetectedLanguage 方法替代 result.language + let detectedLanguage = self.getDetectedLanguage(from: result) + let text = result.text + + print("识别完成事件=检测到语言: \(detectedLanguage), 识别中: \(text)") // 检测文本语言，避免重复调用 - let detectedLanguage = self.detectTextLanguage(result.text ?? "") ?? "zh-CN" + // let detectedLanguage = self.detectTextLanguage(result.text ?? "") ?? "zh-CN" - // 输出检测到的语言信息 - print("检测到的语言: \(detectedLanguage)") + // // 输出检测到的语言信息 + // print("检测到的语言: \(detectedLanguage)") // 在主线程回调结果 DispatchQueue.main.async { @@ -1175,3 +1065,54 @@ private func runMicrophoneCapture() { func onError(_ error: String) } } + +// //ios 语言检测 +// /** +// * 检测输入文本的语言，从支持的语言列表中选择最匹配的 +// * +// * @param text 需要检测的文本 +// * @return 从supportedLanguages中选择的最匹配语言代码 +// */ +// public func detectTextLanguage(_ text: String) -> String? { +// // 如果只有一种支持的语言，直接返回 +// if supportedLanguages.count == 1 { +// return supportedLanguages[0] +// } + +// // 获取所有可能的语言及其置信度 +// let languageResults = LanguageDetector.detectLanguageWithConfidence(from: text) + +// // 遍历检测结果，找到第一个在supportedLanguages中的语言 +// for result in languageResults { +// let azureLanguageCode = LanguageDetector.convertToAzureLanguageCode(result.language) +// if supportedLanguages.contains(azureLanguageCode) { +// return azureLanguageCode +// } +// } + +// // 如果没有找到匹配的语言，返回第一个支持的语言作为默认值 +// return supportedLanguages.first +// } + +// /** +// * 检测文本语言并返回详细信息 +// * +// * @param text 需要检测的文本 +// * @return 包含语言代码、置信度和语言名称的字典 +// */ +// public func detectTextLanguageWithDetails(_ text: String) -> [String: Any]? { +// let results = LanguageDetector.detectLanguageWithConfidence(from: text) +// guard let topResult = results.first else { +// return nil +// } + +// let azureLanguageCode = LanguageDetector.convertToAzureLanguageCode(topResult.language) +// let languageName = LanguageDetector.getLanguageName(from: topResult.language) + +// return [ +// "languageCode": azureLanguageCode, +// "confidence": topResult.confidence, +// "languageName": languageName, +// "originalCode": topResult.language +// ] +// } \ No newline at end of file