1:对接阿里asr
1.1:pom
<dependency><groupId>com.alibaba.nls</groupId><artifactId>nls-sdk-recognizer</artifactId><version>2.2.1</version>
</dependency>
1.2:生成token
package com.dahuyou.ali.asr.generatetoken;import com.alibaba.nls.client.AccessToken;import java.io.IOException;/*** 生成token* program argument参数配置:"LTAI5tNg9N*****R28Zazv" "bAgAvjZwc5HVr******ADEAa"** Token: 6599217b19214759*****42ddf0f8016, expire time: 1726774011*/
public class GenerateToken {public static void main(String[] args) {if (args.length < 2) {System.err.println("CreateTokenDemo need params: <accessKeyId> <accessKeySecret>");System.exit(-1);}String accessKeyId = args[0];String accessKeySecret = args[1];System.out.println("accessKeyId="+accessKeyId+"; accessKeySecret="+accessKeySecret);AccessToken accessToken = new AccessToken(accessKeyId, accessKeySecret);try {accessToken.apply();System.out.println("Token: " + accessToken.getToken() + ", expire time: " + accessToken.getExpireTime());} catch (IOException e) {e.printStackTrace();}}
}
其中accessKeyId和accessKeySecret通过阿里云后台获取:
1.3:在线asr
package com.dahuyou.ali.asr;import java.io.File;
import java.io.FileInputStream;import com.alibaba.nls.client.protocol.InputFormatEnum;
import com.alibaba.nls.client.protocol.NlsClient;
import com.alibaba.nls.client.protocol.SampleRateEnum;
import com.alibaba.nls.client.protocol.asr.SpeechRecognizer;
import com.alibaba.nls.client.protocol.asr.SpeechRecognizerListener;
import com.alibaba.nls.client.protocol.asr.SpeechRecognizerResponse;import org.slf4j.Logger;
import org.slf4j.LoggerFactory;/*** 此示例演示了* ASR一句话识别API调用* 通过本地文件模拟实时流发送* 识别耗时计算* (仅作演示,需用户根据实际情况实现)*/
public class SpeechRecognizerDemo {private static final Logger logger = LoggerFactory.getLogger(SpeechRecognizerDemo.class);private String appKey;NlsClient client;public SpeechRecognizerDemo(String appKey, String token, String url) {this.appKey = appKey;//TODO 重要提示 创建NlsClient实例,应用全局创建一个即可,生命周期可和整个应用保持一致,默认服务地址为阿里云线上服务地址if(url.isEmpty()) {client = new NlsClient(token);}else {client = new NlsClient(url, token);}}// 传入自定义参数private static SpeechRecognizerListener getRecognizerListener(int myOrder, String userParam) {SpeechRecognizerListener listener = new SpeechRecognizerListener() {//识别出中间结果.服务端识别出一个字或词时会返回此消息.仅当setEnableIntermediateResult(true)时,才会有此类消息返回@Overridepublic void onRecognitionResultChanged(SpeechRecognizerResponse response) {//事件名称 RecognitionResultChanged、 状态码(20000000 表示识别成功)、语音识别文本System.out.println("name: " + response.getName() + ", status: " + response.getStatus() + ", result: " + response.getRecognizedText());}//识别完毕@Overridepublic void onRecognitionCompleted(SpeechRecognizerResponse response) {//事件名称 RecognitionCompleted, 状态码 20000000 表示识别成功, getRecognizedText是识别结果文本System.out.println("name: " + response.getName() + ", status: " + response.getStatus() + ", result: " + response.getRecognizedText());}@Overridepublic void onStarted(SpeechRecognizerResponse response) {System.out.println("myOrder: " + myOrder + "; myParam: " + userParam + "; task_id: " + response.getTaskId());}@Overridepublic void onFail(SpeechRecognizerResponse response) {// TODO 重要提示: task_id很重要,是调用方和服务端通信的唯一ID标识,当遇到问题时,需要提供此task_id以便排查System.out.println("task_id: " + response.getTaskId() + ", status: " + response.getStatus() + ", status_text: " + response.getStatusText());}};return listener;}/// 根据二进制数据大小计算对应的同等语音长度/// sampleRate 仅支持8000或16000public static int getSleepDelta(int dataSize, int sampleRate) {// 仅支持16位采样int sampleBytes = 16;// 仅支持单通道int soundChannel = 1;return (dataSize * 10 * 8000) / (160 * sampleRate);}public void process(String filepath, int sampleRate) {SpeechRecognizer recognizer = null;try {// 传递用户自定义参数String myParam = "user-param";int myOrder = 1234;SpeechRecognizerListener listener = getRecognizerListener(myOrder, myParam);recognizer = new SpeechRecognizer(client, listener);recognizer.setAppKey(appKey);//设置音频编码格式 TODO 如果是opus文件,请设置为 InputFormatEnum.OPUSrecognizer.setFormat(InputFormatEnum.PCM);//设置音频采样率if(sampleRate == 16000) {recognizer.setSampleRate(SampleRateEnum.SAMPLE_RATE_16K);} else if(sampleRate == 8000) {recognizer.setSampleRate(SampleRateEnum.SAMPLE_RATE_8K);}//设置是否返回中间识别结果recognizer.setEnableIntermediateResult(true);//此方法将以上参数设置序列化为json发送给服务端,并等待服务端确认long now = System.currentTimeMillis();recognizer.start();logger.info("ASR start latency : " + (System.currentTimeMillis() - now) + " ms");File file = new File(filepath);FileInputStream fis = new FileInputStream(file);byte[] b = new byte[3200];int len;while ((len = fis.read(b)) > 0) {logger.info("send data pack length: " + len);recognizer.send(b, len);// TODO 重要提示:这里是用读取本地文件的形式模拟实时获取语音流并发送的,因为read很快,所以这里需要sleep// TODO 如果是真正的实时获取语音,则无需sleep, 如果是8k采样率语音,第二个参数改为8000// 8000采样率情况下,3200byte字节建议 sleep 200ms,16000采样率情况下,3200byte字节建议 sleep 100msint deltaSleep = getSleepDelta(len, sampleRate);Thread.sleep(deltaSleep);}//通知服务端语音数据发送完毕,等待服务端处理完成now = System.currentTimeMillis();// TODO 计算实际延迟: stop返回之后一般即是识别结果返回时间logger.info("ASR wait for complete");recognizer.stop();logger.info("ASR stop latency : " + (System.currentTimeMillis() - now) + " ms");fis.close();} catch (Exception e) {System.err.println(e.getMessage());} finally {//关闭连接if (null != recognizer) {recognizer.close();}}}public void shutdown() {client.shutdown();}// "e6hRW********ho" "659*************42ddf0f8016" "wss://nls-gateway.cn-shanghai.aliyuncs.com/ws/v1"public static void main(String[] args) throws Exception {String appKey = "你的appkey,在asr应用列表获取";String token = "你的token,上一步生成的,也支持在asr后台获取临时的";String url = ""; // 默认即可,默认值:wss://nls-gateway.cn-shanghai.aliyuncs.com/ws/v1if (args.length == 2) {appKey = args[0];token = args[1];} else if (args.length == 3) {appKey = args[0];token = args[1];url = args[2];} else {System.err.println("run error, need params(url is optional): " + "<app-key> <token> [url]");System.exit(-1);}SpeechRecognizerDemo demo = new SpeechRecognizerDemo(appKey, token, url);// TODO 重要提示: 这里用一个本地文件来模拟发送实时流数据,实际使用时,用户可以从某处实时采集或接收语音流并发送到ASR服务端demo.process("./nls-sample-16k.wav", 16000);//demo.process("./nls-sample.opus", 16000);demo.shutdown();}
}
运行:
nls-sample-16k.wav 。
2:对接azure asr
2.1:pom
<dependency><groupId>com.microsoft.cognitiveservices.speech</groupId><artifactId>client-sdk</artifactId><version>1.40.0</version>
</dependency>
2.2:在线asr
package com.dahuyou.azure.asr.A;import com.microsoft.cognitiveservices.speech.CancellationReason;
import com.microsoft.cognitiveservices.speech.ResultReason;
import com.microsoft.cognitiveservices.speech.SpeechConfig;
import com.microsoft.cognitiveservices.speech.SpeechRecognizer;
import com.microsoft.cognitiveservices.speech.audio.AudioConfig;
import com.microsoft.cognitiveservices.speech.audio.PushAudioInputStream;import java.io.File;
import java.io.FileInputStream;
import java.io.InputStream;public class AzureSpeechRecognition { public static void main(String[] args) { try { // 替换为你的订阅密钥和区域 String speechSubscriptionKey = "你的订阅密钥";String region = "你的区域";SpeechConfig speechConfig = SpeechConfig.fromSubscription(speechSubscriptionKey, region);// 设置中文speechConfig.setSpeechRecognitionLanguage("zh-CN");
// PushAudioInputStream pushAudioInputStream = new PushAudioInputStream();PushAudioInputStream pushAudioInputStream = PushAudioInputStream.create();// 使用默认麦克风
// AudioConfig audioConfig = AudioConfig.fromDefaultMicrophoneInput();// Recognized: 北京的天气。
// AudioConfig audioConfig = AudioConfig.fromWavFileInput("D:\\xiaofuge_sourcecode\\interview-master\\aliasr\\nls-sample-16k.wav");
// AudioConfig audioConfig = AudioConfig.fromWavFileInput("D:\\test\\ttsmaker-file-2024-9-19-17-35-30.wav");AudioConfig audioConfig = AudioConfig.fromStreamInput(pushAudioInputStream);// 假设你有一个方法可以从网络接收音频流
// InputStream audioStream = receiveAudioStreamFromNetwork();
//
// // 准备AudioConfig(这里需要你自己实现转换逻辑)
// AudioConfig audioConfig = prepareAudioConfig(audioStream);SpeechRecognizer recognizer = new SpeechRecognizer(speechConfig, audioConfig); // 订阅事件 recognizer.recognized.addEventListener((s, e) -> { if (e.getResult().getReason() == ResultReason.RecognizedSpeech) {System.out.println("Recognized: " + e.getResult().getText()); } });recognizer.recognizing.addEventListener((s, e) -> {if (e.getResult().getReason() == ResultReason.RecognizingSpeech) {System.out.println("RecognizingSpeech: " + e.getResult().getText());}});recognizer.canceled.addEventListener((s, e) -> { System.out.println("Canceled " + e.getReason()); if (e.getReason() == CancellationReason.Error) {System.out.println("Error details: " + e.getErrorDetails()); } }); // 开始识别 recognizer.startContinuousRecognitionAsync().get();String filepath = "d:\\test\\ttsmaker-file-2024-9-19-18-51-21.wav";File file = new File(filepath);FileInputStream fis = new FileInputStream(file);byte[] b = new byte[3200];int len;while ((len = fis.read(b)) > 0) {
// recognizer.send(b, len);byte[] usedByte = new byte[len];if (len < 3200) {System.arraycopy(b, 0, usedByte, 0, len);} else {usedByte = b;}System.out.println(" usedByte send data pack length: " + usedByte.length);// pushAudioInputStream.write(b);pushAudioInputStream.write(usedByte);// TODO 重要提示:这里是用读取本地文件的形式模拟实时获取语音流并发送的,因为read很快,所以这里需要sleep// TODO 如果是真正的实时获取语音,则无需sleep, 如果是8k采样率语音,第二个参数改为8000// 8000采样率情况下,3200byte字节建议 sleep 200ms,16000采样率情况下,3200byte字节建议 sleep 100ms
// int deltaSleep = getSleepDelta(len, sampleRate);int deltaSleep = 200;Thread.sleep(deltaSleep);usedByte = null;}pushAudioInputStream.close();// 保持程序运行,等待用户输入或其他方式停止 System.in.read(); // 停止识别 recognizer.stopContinuousRecognitionAsync().get(); } catch (Exception ex) { ex.printStackTrace(); } }// // 假设你有一个方法来接收网络上的音频流(这里用伪代码表示)
// static InputStream receiveAudioStreamFromNetwork() {
// // 使用HTTP、WebSocket等接收音频流
// // 这里返回一个InputStream,但实际上你可能需要更复杂的处理
// return new InputStream() {
// // 实现InputStream的read等方法来从网络读取数据
// };
// }// // 将InputStream转换为Azure Speech SDK可以处理的格式(这里简化为直接返回)
在实际中,你可能需要将其写入WAV文件或使用内存中的流
// static AudioConfig prepareAudioConfig(InputStream inputStream) {
// // 注意:Azure Speech SDK的Java版本通常不直接从InputStream读取
// // 你可能需要将inputStream写入到WAV文件,并使用AudioConfig.fromWavFileInput
// // 但这里我们假设有一个方法可以直接处理
// // return AudioConfig.fromCustomStream(inputStream); // 这是一个假设的方法
// return null; // 实际上你需要实现这个转换
// }}
运行:
RecognizingSpeech: 你好啊我usedByte send data pack length: 3200usedByte send data pack length: 3200usedByte send data pack length: 3200
RecognizingSpeech: 你好啊我是usedByte send data pack length: 3200usedByte send data pack length: 3200usedByte send data pack length: 3200usedByte send data pack length: 3200
RecognizingSpeech: 你好啊我是张三usedByte send data pack length: 2894
Recognized: 你好啊,我是张三。
Recognized:
Canceled EndOfStream
ttsmaker-file-2024-9-19-18-51-21.wav 。
写在后面
参考文章列表
Java SDK 。
azure 。
在线配音工具 。