使用Android Recognizer Intent实现低交互连续语音识别的异常检测问题

问题核心分析 #

你当前使用的ACTION_RECOGNIZE_SPEECH是单次识别Intent，当出现「Didn't catch that. Try speaking again.」这类识别失败时，系统对话框会自行关闭，你需要通过onActivityResult捕获返回状态，同时调整实现逻辑来支持连续识别。以下是具体改进方案：

方案一：基于原Intent优化错误捕获与连续识别 #

通过重写onActivityResult方法，识别系统返回的失败状态，触发新一轮识别流程：

@Override
protected void onActivityResult(int requestCode, int resultCode, Intent data) {
    super.onActivityResult(requestCode, resultCode, data);
    if (requestCode == RESULT_SPEECH) {
        // 识别成功，处理结果后启动下一次识别
        if (resultCode == RESULT_OK && data != null) {
            ArrayList<String> result = data.getStringArrayListExtra(RecognizerIntent.EXTRA_RESULTS);
            handleRecognitionResult(result.get(0));
            startSpeechRecognition();
        } 
        // 识别失败（含未捕获语音、用户取消）
        else if (resultCode == RESULT_CANCELED) {
            String errorMsg = data.getStringExtra(RecognizerIntent.EXTRA_ERROR_MESSAGE);
            // 针对「未捕获语音」的错误直接重启识别
            if (errorMsg != null && errorMsg.contains("Didn't catch that")) {
                startSpeechRecognition();
            } else {
                // 其他取消场景（如用户手动关闭对话框），可按需决定是否重启
                // startSpeechRecognition();
            }
        }
    }
}

// 封装识别启动逻辑，便于重复调用
private void startSpeechRecognition() {
    Intent intent = new Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH);
    intent.putExtra(RecognizerIntent.EXTRA_PARTIAL_RESULTS, true);
    intent.putExtra(RecognizerIntent.EXTRA_LANGUAGE_MODEL, RecognizerIntent.LANGUAGE_MODEL_FREE_FORM);
    String lang = "en-US";
    intent.putExtra(RecognizerIntent.EXTRA_LANGUAGE, lang);
    // 移除无用参数：EXTRA_ONLY_RETURN_LANGUAGE_PREFERENCE是用于获取语言偏好，而非设置识别语言
    startActivityForResult(intent, RESULT_SPEECH);
}

额外优化点 #

• 移除EXTRA_ONLY_RETURN_LANGUAGE_PREFERENCE参数，避免干扰正常识别流程；
• 可添加重试次数限制，防止无限循环重启识别消耗资源；
• 增加语音提示或视觉反馈，让用户感知系统处于监听状态。

方案二：使用SpeechRecognizer实现低交互连续识别 #

如果需要完全去掉系统对话框的低交互体验，推荐直接使用SpeechRecognizer类，它支持后台监听，且能更精准地捕获错误类型：

private SpeechRecognizer speechRecognizer;
private Intent recognizerIntent;

@Override
protected void onCreate(Bundle savedInstanceState) {
    super.onCreate(savedInstanceState);
    // 初始化SpeechRecognizer
    speechRecognizer = SpeechRecognizer.createSpeechRecognizer(this);
    speechRecognizer.setRecognitionListener(new RecognitionListener() {
        @Override
        public void onReadyForSpeech(Bundle params) {}

        @Override
        public void onBeginningOfSpeech() {}

        @Override
        public void onRmsChanged(float rmsdB) {}

        @Override
        public void onBufferReceived(byte[] buffer) {}

        @Override
        public void onEndOfSpeech() {}

        @Override
        public void onError(int error) {
            // 精准捕获错误类型
            switch (error) {
                case SpeechRecognizer.ERROR_NO_MATCH:
                case SpeechRecognizer.ERROR_SPEECH_TIMEOUT:
                    // 对应「未捕获语音」的场景，直接重启监听
                    startListening();
                    break;
                default:
                    // 处理其他错误（如无麦克风权限、识别服务不可用）
                    break;
            }
        }

        @Override
        public void onResults(Bundle results) {
            ArrayList<String> result = results.getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION);
            if (result != null && !result.isEmpty()) {
                handleRecognitionResult(result.get(0));
                // 识别完成后立即重启，实现连续监听
                startListening();
            }
        }

        @Override
        public void onPartialResults(Bundle partialResults) {
            // 处理部分识别结果
            ArrayList<String> partialResult = partialResults.getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION);
            if (partialResult != null && !partialResult.isEmpty()) {
                handlePartialResult(partialResult.get(0));
            }
        }

        @Override
        public void onEvent(int eventType, Bundle params) {}
    });

    // 初始化识别配置
    recognizerIntent = new Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH);
    recognizerIntent.putExtra(RecognizerIntent.EXTRA_LANGUAGE_MODEL, RecognizerIntent.LANGUAGE_MODEL_FREE_FORM);
    recognizerIntent.putExtra(RecognizerIntent.EXTRA_LANGUAGE, "en-US");
    recognizerIntent.putExtra(RecognizerIntent.EXTRA_PARTIAL_RESULTS, true);
}

// 启动语音监听
private void startListening() {
    if (speechRecognizer != null) {
        speechRecognizer.startListening(recognizerIntent);
    }
}

// 停止语音监听
private void stopListening() {
    if (speechRecognizer != null) {
        speechRecognizer.stopListening();
    }
}

@Override
protected void onDestroy() {
    super.onDestroy();
    // 释放资源
    if (speechRecognizer != null) {
        speechRecognizer.destroy();
    }
}

SpeechRecognizer的优势 #

• 无系统对话框，全程后台监听，符合低交互需求；
• 错误回调精准，可直接区分「未捕获语音」「超时」等场景；
• 识别流程完全可控，能实现真正的连续语音识别。

内容的提问来源于stack exchange，提问作者Mr Chan