Android原生/React Native语音识别触发任务及语音计数Counter app开发咨询

1.1 Android Native 实现方案 #

• 权限配置
  先在AndroidManifest.xml中添加必要权限：

  <uses-permission android:name="android.permission.RECORD_AUDIO" />
  <uses-permission android:name="android.permission.INTERNET" /> <!-- 云端识别需此权限 -->
  

• 实时语音识别与匹配
  用系统自带的SpeechRecognizer实现基础识别逻辑：

  SpeechRecognizer recognizer = SpeechRecognizer.createSpeechRecognizer(context);
  recognizer.setRecognitionListener(new RecognitionListener() {
      @Override
      public void onResults(Bundle results) {
          ArrayList<String> matches = results.getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION);
          if (matches != null && !matches.isEmpty()) {
              String recognizedText = matches.get(0);
              // 匹配目标语句并执行任务
              if (recognizedText.equalsIgnoreCase("打开相机")) {
                  Intent cameraIntent = new Intent(MediaStore.ACTION_IMAGE_CAPTURE);
                  startActivity(cameraIntent);
              }
          }
      }
  
      // 实现其他必填回调（onReadyForSpeech、onError等）
  });
  
  // 启动识别
  Intent recognizerIntent = new Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH);
  recognizerIntent.putExtra(RecognizerIntent.EXTRA_LANGUAGE_MODEL, RecognizerIntent.LANGUAGE_MODEL_FREE_FORM);
  recognizerIntent.putExtra(RecognizerIntent.EXTRA_LANGUAGE, Locale.getDefault());
  recognizer.startListening(recognizerIntent);
  

• 持续监听优化
  在onResults或onError回调中重新启动识别，实现持续监听，注意在页面销毁时调用recognizer.destroy()释放资源。
• 离线识别替代
  若无需云端支持，可集成PocketSphinx等离线语音库，将目标语句的关键词模型打包进应用，实现本地匹配。

1.2 React Native 实现方案 #

• 依赖安装
  使用@react-native-voice/voice库处理语音识别：

  npm install @react-native-voice/voice --save
  

  同时在iOS的Info.plist和Android的AndroidManifest.xml中配置对应录音权限。
• 核心代码实现

  import Voice from '@react-native-voice/voice';
  import { useEffect, useState } from 'react';
  
  const VoiceListener = () => {
      useEffect(() => {
          Voice.onSpeechResults = handleSpeechResults;
          Voice.onSpeechError = handleSpeechError;
  
          // 请求权限并启动识别
          Voice.requestPermissions().then(() => Voice.start('zh-CN'));
  
          // 清理资源
          return () => {
              Voice.destroy().then(Voice.removeAllListeners);
          };
      }, []);
  
      const handleSpeechResults = (event) => {
          const recognizedText = event.value[0];
          if (recognizedText.toLowerCase() === "启动任务") {
              // 执行指定任务
              console.log("任务已启动");
          }
      };
  
      const handleSpeechError = () => {
          // 出错后重启识别
          Voice.start('zh-CN');
      };
  
      return null;
  };
  

核心流程 #

用户录制目标语句 → 应用持续监听语音 → 匹配到目标语句自动计数

具体实现步骤 #

步骤1：录音功能实现 #

• Android Native
  用MediaRecorder实现录音并保存本地：

  MediaRecorder recorder = new MediaRecorder();
  recorder.setAudioSource(MediaRecorder.AudioSource.MIC);
  recorder.setOutputFormat(MediaRecorder.OutputFormat.MPEG_4);
  recorder.setAudioEncoder(MediaRecorder.AudioEncoder.AAC);
  String savePath = getExternalFilesDir(null) + "/target_voice.mp4";
  recorder.setOutputFile(savePath);
  recorder.prepare();
  recorder.start();
  // 停止录音时调用recorder.stop()和recorder.release()
  

• React Native
  使用react-native-audio-recorder-player库：

  npm install react-native-audio-recorder-player --save
  

  代码示例：

  import AudioRecorderPlayer from 'react-native-audio-recorder-player';
  const audioRecorderPlayer = new AudioRecorderPlayer();
  const [targetVoicePath, setTargetVoicePath] = useState('');
  
  const startRecording = async () => {
      const path = await audioRecorderPlayer.startRecorder();
      console.log("录音路径：", path);
  };
  
  const stopRecording = async () => {
      const result = await audioRecorderPlayer.stopRecorder();
      setTargetVoicePath(result);
  };
  

步骤2：目标语句校准 #

将录制的音频转成文本（用上述语音识别API），保存为目标匹配文本——这种方式实现简单，适合短句匹配；若需更高准确率，可提取音频MFCC特征，用余弦相似度对比实时音频与录制音频的特征（需借助音频处理库实现）。

步骤3：实时监听与计数更新 #

沿用语音监听逻辑，匹配到目标文本时更新计数，添加防抖机制避免重复触发：

const [count, setCount] = useState(0);
const [lastTriggerTime, setLastTriggerTime] = useState(0);
const COOLDOWN_TIME = 1000; // 1秒冷却

const handleSpeechResults = (event) => {
    const recognizedText = event.value[0];
    const now = Date.now();
    if (recognizedText === targetText && now - lastTriggerTime > COOLDOWN_TIME) {
        setCount(prev => prev + 1);
        setLastTriggerTime(now);
    }
};

步骤4：UI设计 #

• 核心元素：当前计数显示、录音按钮（设置目标语句）、监听启停按钮
• 状态提示：显示“正在监听”“匹配成功”等实时状态

内容的提问来源于stack exchange，提问作者InfiniteForLoop