端到端语音语言大模型Android SDK

implementation fileTree(dir: 'libs', include: ['*.jar', '*.aar'])
implementation 'com.baidu.speech.asr_tts:bdspeech:1.0.0.3a98445'

// ... 其他参数已忽略
map.put(SpeechConstant.APP_ID, APP_ID); // 添加appId
map.put(SpeechConstant.APP_API_KEY, APP_KEY); // 添加apiKey
SpeechEventManager.startAsr(this, new JSONObject(map), this);

private void initialEnv() {
        if (mSampleDirPath == null) {
            String sdcardPath = this.getFilesDir().getPath();
            mSampleDirPath = sdcardPath + "/" + SAMPLE_DIR_NAME;
        }
        makeDir(mSampleDirPath);

        pathVadRes = mSampleDirPath + File.separator + VAD_MODEL;
        copyFromAssets(this, true, VAD_MODEL, pathVadRes);

        pathAecRes = mSampleDirPath + File.separator + SpeechConstant.AEC_ALGO_MODEL;
        copyFromAssets(this, true, SpeechConstant.AEC_ALGO_MODEL, pathAecRes);
        ConfigUtil.setAecModelFile(pathAecRes);
    }

/**
 * 请求权限
 *
 * @return 需要请求的权限数
 */
protected int requestPermissions() {
    String[] permissionArray = checkPermissions();
    if (permissionArray != null && permissionArray.length > 0) {
        ActivityCompat.requestPermissions(this, permissionArray, CODE_REQUEST_PERMISSION);
        return permissionArray.length;
    }
    return 0;
}

protected String[] checkPermissions() {
    String[] noGrantedArr = {};
    List<String> noGrantedList = new ArrayList<>();
    for (String item : sPermissions) {
        int hasWriteStoragePermission = ContextCompat.checkSelfPermission(getApplication(), item);
        if (hasWriteStoragePermission != PackageManager.PERMISSION_GRANTED) {   // 未拥有权限
            noGrantedList.add(item);
        }
    }
    return noGrantedList.toArray(noGrantedArr);

@Override
 public void onEvent(String name, String params, byte[] data, int offset, int length) {
     String msg = "name: " + name + ", params: " + params;
     switch (name) {
         case SpeechConstant.CALLBACK_EVENT_ASR_READY: //准备就绪，可以说话，⼀般在收到此事件后通过UI通知⽤户可以说话了
         case SpeechConstant.CALLBACK_EVENT_ASR_BEGIN: //检测到开始说话
             appendToView(mTvLog, msg);
             break;
         case SpeechConstant.CALLBACK_EVENT_ASR_PARTIAL: //中间结果返回
             logD(TAG, "SpeechConstant.CALLBACK_EVENT_ASR_PARTIAL" +params);
             JSONObject partialJson = JsonUtil.toJSON(params);
             String resultType = partialJson.optString("result_type");
             String bestResult = "";
             String showAsr = "中间识别结果";
             if ("partial_result".equals(resultType)) {
                 bestResult = partialJson.optString("best_result");
             } else if ("final_result".equals(resultType)) {
                 bestResult = partialJson.optString("best_result");
                 logD(TAG, "------" + params + ", " + mLastFinalSn);
                 JSONObject json = JsonUtil.toJSON(params);
                 mLastFinalSn = json.optString("sn");
                 showAsr = "最终识别结果";
             } else if ("third_result".equals(resultType)) {
                 Log.d(TAG, "third_result: " + resultType);
             }
             if (!TextUtils.isEmpty(bestResult)) {
                 String resultInfo = showAsr + ": " + bestResult;
                 printToView(mTvResult, resultInfo);
             }
             break;
         case SpeechConstant.CALLBACK_EVENT_ASR_EXIT: //退出
             appendToView(mTvLog, msg);
             handleAsrExit();
             break;
         case SpeechConstant.CALLBACK_EVENT_ASR_END: //说话结束等待识别结果
         case SpeechConstant.CALLBACK_EVENT_ASR_FINISH: //识别完成
         case SpeechConstant.CALLBACK_EVENT_ASR_CANCEL: //识别已取消
         case SpeechConstant.CALLBACK_EVENT_ASR_PAUSE:
         case SpeechConstant.CALLBACK_EVENT_ASR_RESUME:
             break;
         case SpeechConstant.CALLBACK_ASR_TTS_RESULT: //TTS 数据返回
             logD(TAG, "CALLBACK_ASR_TTS_RESULT" + params);
             JSONObject ttsJson = JsonUtil.toJSON(params);
             String origin_result =ttsJson.optString("origin_result");
             JSONObject tts_origin_result =JsonUtil.toJSON(origin_result);
             String text = tts_origin_result.optString("tex");
             int idx = tts_origin_result.optInt("idx", 0);
             break;
         case SpeechConstant.CALLBACK_EVENT_TTS_FIRST_PLAYED: // tts 开始播放
             appendToView(mTvLog, msg);
             Log.d(TAG, "CALLBACK_EVENT_TTS_FIRST_PLAYED: " + msg);
             break;
         case SpeechConstant.CALLBACK_EVENT_TTS_END_PLAYED: // tts 播放完成
             Log.d(TAG, "CALLBACK_EVENT_TTS_END_PLAYED: " + msg);
             break;
         case SpeechConstant.CALLBACK_EVENT_UPLOAD_FINISH: //上传完成
         case SpeechConstant.CALLBACK_EVENT_ASR_AUDIO:
         case SpeechConstant.CALLBACK_EVENT_ASR_VOLUME: //⾳量回调
             break;
         }
     }

@Override
public void onSynthesizeResponse(SynthesizerResponse synthesizerResponse) {
   SynthesizerResponse.SynthesizeType synthesizeType = synthesizerResponse.getSynthesizeType();
    String sn = synthesizerResponse.getSn();
    String utteranceId = synthesizerResponse.getUtteranceId();
    String instanceId = synthesizerResponse.getInstanceId();
    switch (synthesizeType) {
        case PLAY_START:
        case PLAY_FINISH:
            LoggerProxy.d(TAG, "onSynthesizeResult type =  " + synthesizeType.name());
            break;
        case PLAY_PROGRESS:
            SynthesizerResponse.SynthesizerData synthesizerData =
                    synthesizerResponse.getSynthesizerData();
            int progress = synthesizerData.getAudioProgress();
            int engineType = synthesizerData.getEngineType();
            int sampleRate = synthesizerData.getAudioSampleRate();
            byte[] audioData = synthesizerData.getAudioData();
            LoggerProxy.d(TAG, "onSynthesizeResult type =  " + synthesizeType.name()
                    + " sn = " + sn
                    + " , instanceId = " + instanceId
                    + " , progress = " + progress
                    + " , engineType = " + engineType + " , sampleRate = " + sampleRate);
            break;
    }

}

SpeechEventManager.SDKConfig.Builder builder = new 
SpeechEventManager.SDKConfig.Builder();
HashMap<String, String> asrParamMap = new HashMap<>();
asrParamMap.put(SpeechConstant.PID, "8533");
asrParamMap.put(SpeechConstant.DEVICE_ID, mCuid);
asrParamMap.put(SpeechConstant.LOG_LEVEL,
String.valueOf(LogUtil.VERBOSE));

HashMap<String, String> externalMap = new HashMap<>();
externalMap.put(SpeechConstant.LOCATION, "Beijing");
externalMap.put(SpeechConstant.APP_VERSION, "1.0.0");
externalMap.put(SpeechConstant.DEVICE_ID, mCuid);

SpeechEventManager.SDKConfig sdkConfig = builder
         .setAsrParamMap(asrParamMap)
         .setExternalMap(externalMap)
         .build();
int ret = SpeechEventManager.initSDK(this, sdkConfig);
Log.d(TAG, "initSDK: ret = " + ret);

HashMap<String, Object> asrParams = new HashMap<>();

// 环境信息设置：
// 识别环境 - pid. 邀测环境PID统一为4144779，仅供体验使用。生产环境使用请联系技术支持获取专属PID
asrParams.put(SpeechConstant.PID, 1860); 
// 识别环境 - url
asrParams.put(SpeechConstant.URL, "https://vse.baidu.com/v2"); 

// 必选参数设置
// 设置VAD/TriggerMode/双工，请参考**开发场景**根据您的业务需求设置。下面示例为双工识别场景。
// 开启VAD
asrParams.put(SpeechConstant.VAD, SpeechConstant.VAD_DOAVAD);
// 是否启动双工
asrParams.put(SpeechConstant.ASR_ENABLE_MUTIPLY_SPEECH, 1);
asrParams.put(SpeechConstant.TRIGGER_MODE, triggerMode); // trigger 1: 点按识别； 2: 唤醒后识别，4：长按识别

// 必选：设置音频压缩类型，  
asrParams.put(SpeechConstant.ASR_AUDIO_COMPRESSION_TYPE, SpeechConstant.OPUS); 
// 是否返回识别中间结果
asrParams.put(SpeechConstant.ENABLE_EARLY_RETURN, true);
asrParams.put(SpeechConstant.ASR_VAD_RES_PATH, pathVadRes); // vad资源路径，上⽂拷⻉资源的路径
asrParams.put(SpeechConstant.NET_TYPE, 1); // ⽹络类型；1：有turbonet的http；3：⽆turbonet的http
asrParams.put(SpeechConstant.LOG_LEVEL, LogUtil.VERBOSE); // 设置⽇志等级，Debug 使⽤
asrParams.put(SpeechConstant.PAM,
// 第一个参数为DEVICE_ID, 第二个参数为BDVS_ID.  
JsonUtil.genAsrPam("75CEB7CF757E522BAEA2E94BCBC676E5|VUGJPCGTI",
"68b509484f249601628e7c1b547a003c", null)); // bdvs 协议参数
SpeechEventManager.startAsr(this, new JSONObject(asrParams), this);

SpeechEventManager.stopASR();

SpeechEventManager.exitASR();

Map<String, Object> params = new HashMap<>();
LogUtil.d(TAG, "pauseAsr lastSn: " + lastSn); // 该 sn 之后的识别都会被取消
params.put(SpeechConstant.KEY_LAST_FINAL_RESULT_SN, lastSn);
SpeechEventManager.pauseASR(new JSONObject(params));

SpeechEventManager.resumeASR();

// .... 其他参数已省略
map.put(SpeechConstant.VAD, SpeechConstant.VAD_DOAVAD);
// 双工识别
map.put(SpeechConstant.ASR_ENABLE_MUTIPLY_SPEECH, 1);
// TRIGGER_MODE:2, 唤醒后识别
map.put(SpeechConstant.TRIGGER_MODE, 2);
JSONObject jsonObject;
jsonObject = new JSONObject(map);
SpeechEventManager.startAsr(this.getApplicationContext(), jsonObject, this);

// .... 其他参数已省略
map.put(SpeechConstant.VAD, SpeechConstant.VAD_DOAVAD);
// TRIGGER_MODE:1 短按识别
map.put(SpeechConstant.TRIGGER_MODE, 1)
JSONObject jsonObject;
jsonObject = new JSONObject(map);
SpeechEventManager.startAsr(this.getApplicationContext(), jsonObject, this);

// .... 其他参数已省略
// 在按下时startAsr
asrParams.put(SpeechConstant.VAD, SpeechConstant.VAD_TOUCH);
// TRIGGER_MODE:4 长按识别模式
asrParams.put(SpeechConstant.TRIGGER_MODE, 4)
JSONObject jsonObject;
jsonObject = new JSONObject(asrParams);
SpeechEventManager.startAsr(this.getApplicationContext(), jsonObject, this);

// 在松开时stopAsr
SpeechEventManager.stopAsr();