修改语音识别功能

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/AudioKeyWord.java

@@ -18,15 +18,15 @@ public class AudioKeyWord {
     public AudioKeyWord() {
         values = new HashMap<>();
         //地址
-        values.put(MessageKeywords.LOCATION.getType(), new HashSet<>());
+        values.put(AudioKeywordType.LOCATION.getType(), new HashSet<>());
         //联系人
-        values.put(MessageKeywords.CONTACT.getType(), new HashSet<>());
+        values.put(AudioKeywordType.CONTACT.getType(), new HashSet<>());
         //联系电话
-        values.put(MessageKeywords.CONTACT_NUMBER.getType(), new HashSet<>());
+        values.put(AudioKeywordType.CONTACT_NUMBER.getType(), new HashSet<>());
         //事发单位
-        values.put(MessageKeywords.UNIT.getType(), new HashSet<>());
+        values.put(AudioKeywordType.UNIT.getType(), new HashSet<>());
         //重点部位
-        values.put(MessageKeywords.PART.getType(), new HashSet<>());
+        values.put(AudioKeywordType.PART.getType(), new HashSet<>());
     }
 
     public Map<String, Set<String>> getValues() {

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/MessageType.java → amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/AudioKeywordType.java

 package com.yeejoin.amos.boot.module.jcs.biz.audioToText;
 
 /**
- * 语音转文字消息类型
- * mqtt 发送消息类型是加上[_phoneNumber]即接电话用户的电话号码，实现区分
- */
-public enum MessageType {
-    RECORD("audioRecord"), KEYWORD("audioKeyWord");
-
-    private String name;
-
-    MessageType(String name) {
-        this.name = name;
-    }
-
-    public String getName() {
-        return name;
-    }
-
-    public void setName(String name) {
-        this.name = name;
-    }
-}
-
-/**
  * 语音转文字关键字查找表
  */
-enum MessageKeywords {
+enum AudioKeywordType {
     LOCATION("location", new String[]{"地址", "位置", "地点",}),
     CONTACT("contact", new String[]{"联系人", "联系", "在",}),
     CONTACT_NUMBER("contactNumber", new String[]{"号码是", "号码", "手机是", "手机", "手机号", "手机号是"}),
@@ -35,7 +13,7 @@ enum MessageKeywords {
     private String type;
     private String[] keyword;
 
-    MessageKeywords(String type, String[] keyword) {
+    AudioKeywordType(String type, String[] keyword) {
         this.type = type;
         this.keyword = keyword;
     }

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/AudioTopic.java

+package com.yeejoin.amos.boot.module.jcs.biz.audioToText;
+
+/**
+ * 语音转文字消息类型
+ * mqtt 发送消息类型是加上[_phoneNumber]即接电话用户的电话号码，实现区分
+ */
+public enum AudioTopic {
+    RECORD("audioRecord"), KEYWORD("audioKeyWord");
+
+    private String name;
+
+    AudioTopic(String name) {
+        this.name = name;
+    }
+
+    public String getName() {
+        return name;
+    }
+
+    public void setName(String name) {
+        this.name = name;
+    }
+}
+

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/FileSpeechTranscriberListener.java

+package com.yeejoin.amos.boot.module.jcs.biz.audioToText;
+
+import com.alibaba.nls.client.protocol.asr.SpeechTranscriberListener;
+import com.alibaba.nls.client.protocol.asr.SpeechTranscriberResponse;
+
+/**
+ * 文件语音识别 回调函数
+ */
+public class FileSpeechTranscriberListener extends SpeechTranscriberListener {
+    @Override
+    public void onTranscriberStart(SpeechTranscriberResponse response) {
+
+    }
+
+    @Override
+    public void onSentenceBegin(SpeechTranscriberResponse response) {
+
+    }
+
+    @Override
+    public void onSentenceEnd(SpeechTranscriberResponse response) {
+
+    }
+
+    @Override
+    public void onTranscriptionResultChange(SpeechTranscriberResponse response) {
+
+    }
+
+    @Override
+    public void onTranscriptionComplete(SpeechTranscriberResponse response) {
+
+    }
+
+    @Override
+    public void onFail(SpeechTranscriberResponse response) {
+
+    }
+
+    void responseLogger(SpeechTranscriberResponse response) {
+
+    }
+}

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/SpeechTranscriberListener.java → amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/RealTimeSpeechTranscriberListener.java

 package com.yeejoin.amos.boot.module.jcs.biz.audioToText;
 
+import com.alibaba.nls.client.protocol.asr.SpeechTranscriberListener;
 import com.alibaba.nls.client.protocol.asr.SpeechTranscriberResponse;
 import com.fasterxml.jackson.databind.ObjectMapper;
 import org.eclipse.paho.client.mqttv3.MqttException;
@@ -17,8 +18,8 @@ import java.util.regex.Pattern;
 /**
  * 实时语音识别 回调函数
  */
-public class SpeechTranscriberListener extends com.alibaba.nls.client.protocol.asr.SpeechTranscriberListener {
-    private static final Logger logger = LoggerFactory.getLogger(SpeechTranscriberListener.class);
+public class RealTimeSpeechTranscriberListener extends SpeechTranscriberListener {
+    private static final Logger logger = LoggerFactory.getLogger(RealTimeSpeechTranscriberListener.class);
     public static final int RESULT_SUCCESS_CODE = 20000000;
     private final String myNumber;
     private final String typeNumber;
@@ -26,7 +27,7 @@ public class SpeechTranscriberListener extends com.alibaba.nls.client.protocol.a
     private final List<AudioRecord> audioRecords;
     private final AudioKeyWord audioKeyWords;
 
-    public SpeechTranscriberListener(
+    public RealTimeSpeechTranscriberListener(
             String myNumber,
             String typeNumber,
             EmqKeeper emqKeeper,
@@ -116,8 +117,8 @@ public class SpeechTranscriberListener extends com.alibaba.nls.client.protocol.a
                 + ", status: " + response.getStatus()
                 + ",result:" + response.getTransSentenceText());
         try {
-            emqKeeper.getMqttClient().publish(MessageType.RECORD.getName() + "/" + myNumber, "".getBytes(), 2, true);
-            emqKeeper.getMqttClient().publish(MessageType.KEYWORD.getName() + "/" + myNumber, "".getBytes(), 2, true);
+            emqKeeper.getMqttClient().publish(AudioTopic.RECORD.getName() + "/" + myNumber, "".getBytes(), 2, true);
+            emqKeeper.getMqttClient().publish(AudioTopic.KEYWORD.getName() + "/" + myNumber, "".getBytes(), 2, true);
         } catch (MqttException e) {
             e.printStackTrace();
         }
@@ -169,7 +170,7 @@ public class SpeechTranscriberListener extends com.alibaba.nls.client.protocol.a
         try {
             ObjectMapper objectMapper = new ObjectMapper();
             byte[] bytes = objectMapper.writeValueAsString(audioRecord).getBytes();
-            emqKeeper.getMqttClient().publish(MessageType.RECORD.getName() + "/" + myNumber, bytes, 2, false);
+            emqKeeper.getMqttClient().publish(AudioTopic.RECORD.getName() + "/" + myNumber, bytes, 2, false);
         } catch (Exception e) {
             e.printStackTrace();
             logger.error("发送音频识别结果消息异常，原因：" + e.getMessage());
@@ -192,11 +193,11 @@ public class SpeechTranscriberListener extends com.alibaba.nls.client.protocol.a
             Pattern pattern = Pattern.compile("\\d{5,}");
             Matcher matcher = pattern.matcher(result);
             while (matcher.find()) {
-                audioKeyWords.getValues().get(MessageKeywords.CONTACT_NUMBER.getType()).add(matcher.group());
+                audioKeyWords.getValues().get(AudioKeywordType.CONTACT_NUMBER.getType()).add(matcher.group());
             }
             //寻找其他类型关键字
-            for (MessageKeywords messageKeyword : MessageKeywords.values()) {
-                if (messageKeyword == MessageKeywords.CONTACT_NUMBER) continue;
+            for (AudioKeywordType messageKeyword : AudioKeywordType.values()) {
+                if (messageKeyword == AudioKeywordType.CONTACT_NUMBER) continue;
                 for (String keyword : messageKeyword.getKeyword()) {
                     int keywordIndex = result.indexOf(keyword);
                     //问答式：截取关键字，从关键字的下一个[,/。]号开始截取到下下一个[,/。]号结束
@@ -213,7 +214,7 @@ public class SpeechTranscriberListener extends com.alibaba.nls.client.protocol.a
             }
             ObjectMapper objectMapper = new ObjectMapper();
             byte[] bytes = objectMapper.writeValueAsString(audioKeyWords.getValues()).getBytes();
-            emqKeeper.getMqttClient().publish(MessageType.KEYWORD.getName() + "/" + myNumber, bytes, 2, false);
+            emqKeeper.getMqttClient().publish(AudioTopic.KEYWORD.getName() + "/" + myNumber, bytes, 2, false);
         } catch (Exception e) {
             e.printStackTrace();
             logger.error("发送音频关键字消息异常，原因：" + e.getMessage());

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/SocketClient.java

@@ -8,6 +8,7 @@ import org.springframework.stereotype.Component;
 import java.io.File;
 import java.io.FileInputStream;
 import java.io.OutputStream;
+import java.net.InetAddress;
 import java.net.InetSocketAddress;
 import java.net.Socket;
 import java.util.concurrent.TimeUnit;
@@ -37,7 +38,7 @@ public class SocketClient {
 
         Socket socket = new Socket();
         try {
-            socket.connect(new InetSocketAddress("127.0.0.1", port));
+            socket.connect(new InetSocketAddress(InetAddress.getLocalHost().getHostAddress(), port));
             OutputStream outputStream = socket.getOutputStream();
             FileInputStream fis = new FileInputStream(new File(testFilePath[type]));
             byte[] b = new byte[4096];

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/streamToText/AbstractInputStream2Text.java

+package com.yeejoin.amos.boot.module.jcs.biz.audioToText.streamToText;
+
+import java.io.InputStream;
+
+/**
+ * 语音文件输入流，转文字
+ */
+public abstract class AbstractInputStream2Text {
+    private InputStream inputStream;
+
+    /**
+     * 获取输入流
+     */
+    abstract InputStream getInputStream();
+
+    /**
+     * 转换操作
+     */
+    void doTranslate() {
+        inputStream = getInputStream();
+        if (inputStream == null) throw new RuntimeException("inputStream is null");
+
+    }
+}

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/streamToText/LocalFileInputStream2Text.java

+package com.yeejoin.amos.boot.module.jcs.biz.audioToText.streamToText;
+
+import java.io.InputStream;
+
+/**
+ * 本地文件语音输入流转文字
+ */
+public class LocalFileInputStream2Text extends AbstractInputStream2Text {
+
+
+    @Override
+    InputStream getInputStream() {
+        return null;
+    }
+
+}

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/streamToText/NetWorkInputStream2Text.java

+package com.yeejoin.amos.boot.module.jcs.biz.audioToText.streamToText;
+
+import java.io.InputStream;
+
+/**
+ * 网络语音文件输入流转文字
+ */
+public class NetWorkInputStream2Text extends AbstractInputStream2Text {
+
+
+    @Override
+    InputStream getInputStream() {
+        return null;
+    }
+
+}

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/AudioToText.java → amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/audioToText/streamToText/RealTimeStream2Text.java

-package com.yeejoin.amos.boot.module.jcs.biz.audioToText;
+package com.yeejoin.amos.boot.module.jcs.biz.audioToText.streamToText;
 
+import com.yeejoin.amos.boot.module.jcs.biz.audioToText.AudioKeyWord;
+import com.yeejoin.amos.boot.module.jcs.biz.audioToText.AudioRecord;
+import com.yeejoin.amos.boot.module.jcs.biz.audioToText.RealTimeSpeechTranscriberListener;
 import com.yeejoin.amos.speech.AppSpeechTranscriber;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -14,11 +17,11 @@ import java.util.HashMap;
 import java.util.List;
 
 /**
- * 语音转文字
+ * 实时语音转文字
  */
 @Component
-public class AudioToText {
-    private static final Logger logger = LoggerFactory.getLogger(AudioToText.class);
+public class RealTimeStream2Text {
+    private static final Logger logger = LoggerFactory.getLogger(RealTimeStream2Text.class);
     public static int serverPort = 10001;
 
     @Autowired
@@ -30,7 +33,7 @@ public class AudioToText {
      * @param myNumber     我的电话号码
      * @param callerNumber 呼入电话号码
      */
-    public HashMap<String, Integer> convert(String myNumber, String callerNumber) {
+    public HashMap<String, Object> doTranslate(String cid, String myNumber, String callerNumber) {
         //启动两个监听端口监听推送进来的2路语音流
         ServerSocket serverSocket1 = initServerSocketPort();
         ServerSocket serverSocket2 = initServerSocketPort();
@@ -41,17 +44,17 @@ public class AudioToText {
         logger.warn("myNumber监听的端口为：" + serverSocket1.getLocalPort() + "  callerNumber监听的端口为：" + serverSocket2.getLocalPort());
         //我的语音流
         Thread thread1 = new Thread(() -> {
-            new AppSpeechTranscriber(new SpeechTranscriberListener(myNumber, myNumber, emqKeeper, audioRecords, audioKeyWord), serverSocket1).process();
+            new AppSpeechTranscriber(new RealTimeSpeechTranscriberListener(myNumber, myNumber, emqKeeper, audioRecords, audioKeyWord), serverSocket1).process();
         });
         //呼入的语音流
         Thread thread2 = new Thread(() -> {
-            new AppSpeechTranscriber(new SpeechTranscriberListener(myNumber, callerNumber, emqKeeper, audioRecords, audioKeyWord), serverSocket2).process();
+            new AppSpeechTranscriber(new RealTimeSpeechTranscriberListener(myNumber, callerNumber, emqKeeper, audioRecords, audioKeyWord), serverSocket2).process();
         });
         thread1.setUncaughtExceptionHandler(new SubUncaughtExceptionHandler(serverSocket1));
         thread2.setUncaughtExceptionHandler(new SubUncaughtExceptionHandler(serverSocket2));
         thread1.start();
         thread2.start();
-        HashMap<String, Integer> map = new HashMap<>();
+        HashMap<String, Object> map = new HashMap<>();
         map.put(myNumber, serverSocket1.getLocalPort());
         map.put(callerNumber, serverSocket2.getLocalPort());
         return map;

amos-boot-module/amos-boot-module-biz/amos-boot-module-jcs-biz/src/main/java/com/yeejoin/amos/boot/module/jcs/biz/controller/Audio2TextController.java

 package com.yeejoin.amos.boot.module.jcs.biz.controller;
 
-import com.yeejoin.amos.boot.module.jcs.biz.audioToText.AudioToText;
+import com.yeejoin.amos.boot.module.jcs.biz.audioToText.streamToText.RealTimeStream2Text;
 import com.yeejoin.amos.boot.module.jcs.biz.audioToText.SocketClient;
 import io.swagger.annotations.Api;
 import io.swagger.annotations.ApiOperation;
@@ -21,11 +21,12 @@ import java.util.concurrent.TimeUnit;
 public class Audio2TextController {
 
     @Autowired
-    AudioToText audioToText;
+    RealTimeStream2Text audio2Text;
 
     @Autowired
     SocketClient socketClient;
 
+
     /**
      * 测试语音转文字第一步
      *
@@ -34,8 +35,8 @@ public class Audio2TextController {
     @TycloudOperation(ApiLevel = UserType.AGENCY)
     @GetMapping("/startConvert")
     @ApiOperation(httpMethod = "GET", value = "测试语音转文字第一步", notes = "number为当前用户的手机号")
-    public HashMap<String, Integer> startConvert(@RequestParam String myNumber, @RequestParam String callerNumber) {
-        return audioToText.convert(myNumber, callerNumber);
+    public HashMap<String, Object> startConvert(@RequestParam String myNumber, @RequestParam String callerNumber) {
+        return audio2Text.doTranslate("", myNumber, callerNumber);
     }
 
     /**
@@ -58,12 +59,12 @@ public class Audio2TextController {
     @TycloudOperation(ApiLevel = UserType.AGENCY)
     @GetMapping("/startConvertAndSendAudio")
     @ApiOperation(httpMethod = "GET", value = "测试语音转文字融合接口", notes = "测试语音转文字融合接口")
-    public HashMap<String, Integer> startConvertAndSendAudio(@RequestParam String myNumber, @RequestParam String callerNumber) {
-        HashMap<String, Integer> convert = audioToText.convert(myNumber, callerNumber);
+    public HashMap<String, Object> startConvertAndSendAudio(@RequestParam String myNumber, @RequestParam String callerNumber) {
+        HashMap<String, Object> convert = audio2Text.doTranslate("", myNumber, callerNumber);
         try {
             TimeUnit.SECONDS.sleep(1);
-            socketClient.process(convert.get(myNumber), 2);
-            socketClient.process(convert.get(callerNumber), 3);
+            socketClient.process((Integer) convert.get(myNumber), 2);
+            socketClient.process((Integer) convert.get(callerNumber), 3);
         } catch (InterruptedException e) {
             e.printStackTrace();
         }
@@ -83,9 +84,10 @@ public class Audio2TextController {
      */
     @TycloudOperation(ApiLevel = UserType.AGENCY)
     @GetMapping("/startConvertText")
-    @ApiOperation(httpMethod = "GET", value = "静听电话回调后端开始转文字", notes = "静听电话回调后端开始转文字")
+    @ApiOperation(httpMethod = "GET", value = "接听电话回调后端开始转文字", notes = "接听电话回调后端开始转文字")
     public void startConvertText(String cid, String myPhone, String caller) {
 
     }
 }
 
+

amos-boot-utils/amos-boot-utils-speech/src/main/java/com/yeejoin/amos/speech/AppSpeechTranscriber.java

@@ -9,6 +9,7 @@ import org.slf4j.LoggerFactory;
 
 import java.io.IOException;
 import java.io.InputStream;
+import java.net.InetAddress;
 import java.net.ServerSocket;
 import java.net.Socket;
 import java.util.Timer;
@@ -51,7 +52,7 @@ public class AppSpeechTranscriber {
         SpeechTranscriber transcriber = null;
         try {
             //启动ServerSocket等待接收音频数据,只接受一次请求
-            logger.warn("serverSocket已启动，地址：" + serverSocket.getInetAddress().getHostAddress()
+            logger.warn("serverSocket已启动，地址：" + InetAddress.getLocalHost().getHostAddress()
                     + "监听端口：" + serverSocket.getLocalPort() + "  等待语音融合系统推送数据...");
             Socket socket = serverSocket.accept();
             timerTask.cancel();