Java 实现socket VAD通讯客户端

以下是基于Java的WebSocket ASR客户端实现，支持实时音频采集、VAD检测和标点符号恢复：

1. 基础Java客户端实现

import javax.sound.sampled.*;
import org.java_websocket.client.WebSocketClient;
import org.java_websocket.handshake.ServerHandshake;
import java.net.URI;
import java.nio.ByteBuffer;
import java.util.concurrent.ArrayBlockingQueue;public class AsrClient {private static final int SAMPLE_RATE = 16000;private static final int CHUNK_SIZE = 1024;private static final float SILENCE_THRESHOLD = 0.02f;private static final int SILENCE_TIMEOUT_MS = 1500;private final ArrayBlockingQueue<byte[]> audioQueue = new ArrayBlockingQueue<>(100);private volatile boolean isRecording = false;public static void main(String[] args) throws Exception {new AsrClient().start("ws://localhost:8765");}public void start(String serverUri) throws Exception {// WebSocket连接WSClient wsClient = new WSClient(new URI(serverUri));wsClient.connect();// 音频采集线程Thread captureThread = new Thread(() -> {try {AudioFormat format = new AudioFormat(SAMPLE_RATE, 16, 1, true, false);DataLine.Info info = new DataLine.Info(TargetDataLine.class, format);if (!AudioSystem.isLineSupported(info)) {System.err.println("不支持的音频格式");return;}TargetDataLine line = (TargetDataLine) AudioSystem.getLine(info);line.open(format);line.start();byte[] buffer = new byte[CHUNK_SIZE];System.out.println("开始录音...");while (isRecording) {int bytesRead = line.read(buffer, 0, buffer.length);if (bytesRead > 0) {audioQueue.put(buffer.clone());}}line.close();} catch (Exception e) {e.printStackTrace();}});// 处理线程Thread processThread = new Thread(() -> {try {int silenceCounter = 0;ByteArrayOutputStream audioBuffer = new ByteArrayOutputStream();while (isRecording || !audioQueue.isEmpty()) {byte[] chunk = audioQueue.poll(100, TimeUnit.MILLISECONDS);if (chunk == null) continue;// VAD检测boolean hasVoice = isVoiceActive(chunk);if (hasVoice) {silenceCounter = 0;audioBuffer.write(chunk);} else if (audioBuffer.size() > 0) {silenceCounter += CHUNK_SIZE * 1000 / SAMPLE_RATE;if (silenceCounter >= SILENCE_TIMEOUT_MS) {// 发送音频数据wsClient.send(audioBuffer.toByteArray());audioBuffer.reset();silenceCounter = 0;}}}} catch (Exception e) {e.printStackTrace();}});// 启动线程isRecording = true;captureThread.start();processThread.start();// 等待结束System.in.read();isRecording = false;captureThread.join();processThread.join();wsClient.close();}private boolean isVoiceActive(byte[] audio) {// 简单的能量检测VADshort[] samples = new short[audio.length / 2];ByteBuffer.wrap(audio).order(ByteOrder.LITTLE_ENDIAN).asShortBuffer().get(samples);double sum = 0;for (short sample : samples) {sum += sample * sample;}double rms = Math.sqrt(sum / samples.length) / 32768.0;return rms > SILENCE_THRESHOLD;}class WSClient extends WebSocketClient {public WSClient(URI serverUri) {super(serverUri);}@Overridepublic void onOpen(ServerHandshake handshakedata) {System.out.println("连接已建立");// 发送客户端信息send("{\"type\":\"client_info\",\"sample_rate\":16000}");}@Overridepublic void onMessage(String message) {System.out.println("识别结果: " + message);}@Overridepublic void onClose(int code, String reason, boolean remote) {System.out.println("连接关闭: " + reason);}@Overridepublic void onError(Exception ex) {ex.printStackTrace();}}
}

2. 需要添加的依赖（Maven）

<dependencies><!-- WebSocket客户端 --><dependency><groupId>org.java-websocket</groupId><artifactId>Java-WebSocket</artifactId><version>1.5.3</version></dependency><!-- 音频处理 --><dependency><groupId>com.googlecode.soundlibs</groupId><artifactId>tritonus-share</artifactId><version>0.3.7-2</version></dependency>
</dependencies>

3. 高级功能扩展版

import org.json.JSONObject;
import java.util.Base64;public class EnhancedAsrClient extends AsrClient {@Overridepublic void start(String serverUri) throws Exception {// 添加鉴权头Map<String, String> headers = new HashMap<>();headers.put("Authorization", "Bearer your_token");WSClient wsClient = new WSClient(new URI(serverUri), headers);// ...其余代码同基础版...}class EnhancedWSClient extends WSClient {private final Gson gson = new Gson();private int messageId = 0;public EnhancedWSClient(URI serverUri, Map<String, String> headers) {super(serverUri);if (headers != null) {this.addHeaders(headers);}}@Overridepublic void onMessage(String message) {try {JsonObject result = gson.fromJson(message, JsonObject.class);if (result.has("text")) {System.out.printf("[%s] 结果: %s\n", result.get("time").getAsString(),result.get("text").getAsString());}} catch (Exception e) {System.out.println("原始消息: " + message);}}public void sendAudio(byte[] audio) {JSONObject msg = new JSONObject();msg.put("id", ++messageId);msg.put("type", "audio");msg.put("data", Base64.getEncoder().encodeToString(audio));send(msg.toString());}}
}

4. 关键功能说明

1. 音频采集：

   • 使用javax.sound.sampled包实现跨平台音频采集
   • 支持16kHz单声道PCM格式

2. VAD检测：

   // 基于RMS能量的简单VAD
   double rms = Math.sqrt(sum / samples.length) / 32768.0;
   return rms > SILENCE_THRESHOLD;
   

3. 协议设计：

   • 连接时发送采样率信息
   • 支持JSON格式的消息交换
   • 音频数据Base64编码传输

4. 性能优化：

   • 使用双缓冲队列避免阻塞音频线程
   • 单独的VAD处理线程

5. 使用说明

1. 启动服务端：

   python server.py  # 使用之前提供的Python服务端
   

2. 运行Java客户端：

   mvn compile exec:java -Dexec.mainClass="AsrClient"
   

3. 参数配置：

   // 可调参数
   private static final float SILENCE_THRESHOLD = 0.02f;  // 静音阈值
   private static final int CHUNK_SIZE = 1024;          // 每次发送的音频块大小
   

6. 常见问题解决

问题1：无法找到音频设备

• 解决方案：检查系统录音权限，或指定具体设备：

  Mixer.Info[] mixers = AudioSystem.getMixerInfo();
  // 打印可用设备后指定具体设备
  line = AudioSystem.getTargetDataLine(format, mixers[0]);
  

问题2：WebSocket连接不稳定

• 解决方案：添加重连机制

  void reconnect() {try {Thread.sleep(5000);this.reconnectBlocking();} catch (Exception e) {e.printStackTrace();}
  }
  

问题3：延迟过高

• 优化建议：
  • 减小CHUNK_SIZE（但会增加CPU负载）
  • 使用OPUS编码压缩音频（需服务端支持）

这个Java客户端实现了与Python服务端的完整交互流程，支持实时音频处理和标点恢复功能。根据实际需求可以扩展更多功能如：音频压缩、自适应VAD阈值、多语言支持等。