diff --git a/tts/tts-pad/libs/DUI-lite-SDK-for-Android-2.18.1.jar b/tts/tts-pad/libs/DUI-lite-SDK-for-Android-2.18.1.jar new file mode 100644 index 0000000000..ff88eb4f4c Binary files /dev/null and b/tts/tts-pad/libs/DUI-lite-SDK-for-Android-2.18.1.jar differ diff --git a/tts/tts-pad/src/main/assets/v2.1.31_aitts_sent_dict_local.db b/tts/tts-pad/src/main/assets/v2.1.31_aitts_sent_dict_local.db new file mode 100644 index 0000000000..991cee42a9 Binary files /dev/null and b/tts/tts-pad/src/main/assets/v2.1.31_aitts_sent_dict_local.db differ diff --git a/tts/tts-pad/src/main/assets/v2.1.31_local_front.bin b/tts/tts-pad/src/main/assets/v2.1.31_local_front.bin new file mode 100644 index 0000000000..e3ed833aaa Binary files /dev/null and b/tts/tts-pad/src/main/assets/v2.1.31_local_front.bin differ diff --git a/tts/tts-pad/src/main/assets/zhilingf_common_back_ce_local.v2.1.0.bin b/tts/tts-pad/src/main/assets/zhilingf_common_back_ce_local.v2.1.0.bin new file mode 100644 index 0000000000..5841f2cf48 Binary files /dev/null and b/tts/tts-pad/src/main/assets/zhilingf_common_back_ce_local.v2.1.0.bin differ diff --git a/tts/tts-pad/src/main/java/com/mogo/tts/pad/PadTTS.java b/tts/tts-pad/src/main/java/com/mogo/tts/pad/PadTTS.java index acec3188fe..8e68db3244 100644 --- a/tts/tts-pad/src/main/java/com/mogo/tts/pad/PadTTS.java +++ b/tts/tts-pad/src/main/java/com/mogo/tts/pad/PadTTS.java @@ -3,17 +3,27 @@ package com.mogo.tts.pad; import android.content.Context; import android.content.Intent; import android.os.Build; +import android.os.Environment; import android.provider.Settings; import android.text.TextUtils; +import com.aispeech.AIEchoConfig; +import com.aispeech.AIError; +import com.aispeech.DUILiteConfig; +import com.aispeech.DUILiteSDK; +import com.aispeech.common.AIConstant; +import com.aispeech.export.config.AILocalTTSConfig; +import com.aispeech.export.engines2.AILocalTTSEngine; +import com.aispeech.export.intent.AILocalTTSIntent; +import com.aispeech.export.listeners.AILocalTTSListener; import com.alibaba.android.arouter.facade.annotation.Route; +import com.mogo.cloud.commons.BuildConfig; import com.mogo.tts.base.IMogoTTS; import com.mogo.tts.base.IMogoTTSCallback; import com.mogo.tts.base.MogoTTSConstants; import com.mogo.tts.base.PreemptType; import com.mogo.utils.AppUtils; import com.mogo.utils.logger.Logger; -import com.mogo.utils.permissions.PermissionsDialogUtils; import com.zhidao.auto.platform.voice.VoiceClient; import com.zhidao.voicesdk.MogoVoiceManager; import com.zhidao.voicesdk.MogoVoiceManagerImpl; @@ -40,8 +50,8 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { private static final String TAG = "PadTTS"; private String mLastQAndASpeakText; - private boolean mHasFlush = false; - private boolean mInitReady = false; + private boolean mHasFlush = true; + private boolean mInitReady = true; private Context mContext; public synchronized void release() { @@ -57,6 +67,10 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { } mQAndAMap.clear(); mVoiceClient.release(); + if (mEngine != null) { + mEngine.destroy(); + } + mHasAuth = false; mSpeakVoiceMap.clear(); mCacheUnWakeupCommands.clear(); mContext = null; @@ -73,12 +87,117 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { private Map< String, String[] > mCacheUnWakeupCommands = new ConcurrentHashMap<>(); + private static final String TTS_BACK_RES_ZHILING = "zhilingf_common_back_ce_local.v2.1.0.bin"; + private int audioRecorderType = DUILiteConfig.TYPE_COMMON_MIC; + //echo module res + private static final String ECHO_RES = "echo/sspe_aec_ch2_mic1_ref1_asr_v2.0.0.95.bin"; + //local tts module res + private static final String TTS_DICT_RES = "v2.1.31_aitts_sent_dict_local.db"; + private static final String TTS_FRONT_RES = "v2.1.31_local_front.bin"; + private AILocalTTSEngine mEngine; + private AILocalTTSIntent mAILocalTTSIntent; + private String[] mBackResBinArray = new String[]{TTS_BACK_RES_ZHILING}; + // 单独的语音播放 + private boolean mHasAuth; + private void initFlushStatus() { if ( !mHasFlush ) { mHasFlush = isVoiceServiceReady( mContext ); } } + private void initTtsConfig() { + Logger.d(TAG, "initTtsConfig"); + // 产品认证需设置 apiKey, productId, productKey, productSecret + DUILiteConfig config = new DUILiteConfig( + "1009e96c95da655dcad8d6eb6145bbdd", + "278586132", + "1fe5930844b488a8d32d9ef7717be7dc", + "f601ecc407986b548ac8ab2a9144162e"); + config.setAuthTimeout(5000); //设置授权连接超时时长，默认5000ms + config.setExtraParameter("DEVICE_NAME", "fea815f374af8"); + config.setExtraParameter("DEVICE_ID", "fea815f374af8"); + config.setDeviceProfileDirPath("/sdcard/speech"); // 自定义设置授权文件的保存路径,需要确保该路径事先存在 +// config.setThreadAffinity(3); //绑定第3个核，降低CPU占用 +// config.setOfflineProfileName("auth.txt");//设置assetes目录下的离线授权文件 + //设置SDK录音模式 + // 单麦 单麦Echo 双麦 线性4麦 环形4麦 环形6麦，可选的 + config.setAudioRecorderType(audioRecorderType); + if (config.getAudioRecorderType() == DUILiteConfig.TYPE_COMMON_ECHO) { + AIEchoConfig aiEchoConfig = new AIEchoConfig(); + aiEchoConfig.setAecResource(ECHO_RES); // 设置echo的AEC资源文件 + aiEchoConfig.setChannels(2); //音频总的通道数 + aiEchoConfig.setMicNumber(1); //真实mic数 + // 默认为1,即左通道为rec录音音频,右通道为play参考音频（播放音频） + // 若设置为2,通道会互换，即左通道为play参考音频（播放音频）,右通道为rec录音音频 + aiEchoConfig.setRecChannel(1); + aiEchoConfig.setSavedDirPath("/sdcard/aispeech/aecPcmFile/");//设置保存的aec原始输入和aec之后的音频文件路径 + config.setEchoConfig(aiEchoConfig); + } + if (BuildConfig.DEBUG) { + config.openLog(); + // 设置log日志的级别 + DUILiteSDK.setDebugMode(3); + } + Logger.d(TAG, "DUILite SDK is isAuthorized ？ " + DUILiteSDK.isAuthorized(mContext)); + DUILiteSDK.init(mContext, + config, + new DUILiteSDK.InitListener() { + @Override + public void success() { + Logger.d(TAG, "授权成功!"); + mHasAuth = true; + initTtsEngine(); + } + + @Override + public void error(final String errorCode, final String errorInfo) { + Logger.d(TAG, "授权失败\n\nErrorCode：" + errorCode + "\n\nErrorInfo：" + errorInfo); + } + }); + } + + private void initTtsEngine() { + Logger.d(TAG, "initTtsEngine"); + if (mEngine != null) { + mEngine.destroy(); + } + AILocalTTSConfig config = new AILocalTTSConfig(); + // 设置assets目录下合成字典名 + config.setDictResource(TTS_DICT_RES);//普通话字典 + // 设置合成字典的外部路径，包含文件名(需要手动拷贝到指定目录) + // config.setDictResource("/sdcard/speech/tts/aitts_sent_dict_idx_middle_2.0.4_20180806.db"); + + // 非必需，用户自定义词典，用于修复离线合成问题，如多音字发音、停顿和数字字母符号读法错误等 + // 正常情況用不着，这个只有有发音问题需要紧急修复可以改这个资源。user_source.bin该资源由合成组提供的工具包生成。 + // config.setUserDictResource("user_source.bin"); + // 设置assets目录下前端合成资源名 + config.setFrontBinResource(TTS_FRONT_RES);//普通话前端资源 + //config.setFrontBinResource(SampleConstants.TTS_FRONT_RES_YUEYU);//粤语前端资源 + //config.setFrontBinResource(SampleConstants.TTS_FRONT_RES_SICHUAN);//四川话前端资源 + // 设置合成前端资源的外部路径，包含文件名(需要手动拷贝到指定目录) + // config.setFrontBinResource("/sdcard/speech/tts/v2.1.23_local_front.bin"); + // default is true + config.setUseCache(false); + //设置后端合成音色资源，如果只需设置一个，则array只需要传一个成员值就可以 + config.addSpeakerResource(mBackResBinArray); + // 设置合成音色的外部路径，包含文件名(需要手动拷贝到指定目录) + // config.addSpeakerResource("/sdcard/speech/tts/zhilingf_common_back_ce_local.v2.1.0.bin"); + + mEngine = AILocalTTSEngine.createInstance();//创建实例 + mEngine.init(config, new AILocalTTSListenerImpl());//初始化合成引擎 + + mAILocalTTSIntent = new AILocalTTSIntent(); + // 设置合成音语速，范围为0.5～2.0 + mAILocalTTSIntent.setSpeed(0.85f); + + mAILocalTTSIntent.setUseSSML(false); // 设置是否使用ssml合成语法，默认为false + mAILocalTTSIntent.setVolume(100); // 设置合成音频的音量，范围为1～500 + // 保存合成音频到指定路径，格式为wav + mAILocalTTSIntent.setSaveAudioFilePath(Environment.getExternalStorageDirectory() + "/tts/" + + System.currentTimeMillis() + ".wav"); + } + /** * 初始化 */ @@ -190,15 +309,16 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { * @param text */ public void speakTTSVoice( String text, IMogoTTSCallback callBack ) { - try { - initFlushStatus(); - if ( mHasFlush ) { - mSpeakVoiceMap.put( text, callBack ); - mVoiceClient.speakDefault( text ); - } - } catch ( Exception e ) { - e.printStackTrace(); - } + speakTTSVoice(text); +// try { +// initFlushStatus(); +// if ( mHasFlush ) { +// mSpeakVoiceMap.put( text, callBack ); +// mVoiceClient.speakDefault( text ); +// } +// } catch ( Exception e ) { +// e.printStackTrace(); +// } } /** @@ -207,13 +327,12 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { * @param text */ public void speakTTSVoice( String text ) { - try { - initFlushStatus(); - if ( mHasFlush ) { - mVoiceClient.speakDefault( text ); - } - } catch ( Exception e ) { - e.printStackTrace(); + Logger.d( TAG, "speakTTSVoice"); + if (mEngine != null && mHasAuth) { + // 合成并播放 + mEngine.speak(mAILocalTTSIntent, text, "1024"); + // 合成音频，不播放，同时输出实时pcm音频,音频回调在onSynthesizeDataArrived接口 + mEngine.synthesize(mAILocalTTSIntent, text, "1024"); } } @@ -245,7 +364,8 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { break; } } - mVoiceClient.speakTypeText( text, preemptType ); + speakTTSVoice(text); +// mVoiceClient.speakTypeText( text, preemptType ); } } catch ( Exception e ) { e.printStackTrace(); @@ -261,7 +381,8 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { initFlushStatus(); if ( mHasFlush ) { mQAndAMap.put( tts, callBack ); - mVoiceClient.speakTtsAndRegistCmd( tts ); + speakTTSVoice(tts); +// mVoiceClient.speakTtsAndRegistCmd( tts ); } } @@ -276,7 +397,8 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { initFlushStatus(); if ( mHasFlush ) { mQAndAMap.put( tts, callBack ); - mVoiceClient.speakTtsAndRegistCmd( tts, okCmds, cancelCmds ); + speakTTSVoice(tts); +// mVoiceClient.speakTtsAndRegistCmd( tts, okCmds, cancelCmds ); } } @@ -471,6 +593,7 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { mVoiceClient = new VoiceClient( mContext ); mVoiceClient.setCallBack( this ); initFlushStatus(); + initTtsConfig(); if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.M) { // 检查是否有悬浮窗权限 if (Settings.canDrawOverlays(context)) { @@ -499,4 +622,54 @@ class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsListener { Logger.d( TAG, "status = %d", status ); context.sendBroadcast( intent ); } + + private static class AILocalTTSListenerImpl implements AILocalTTSListener { + + @Override + public void onInit(int status) { + Logger.d(TAG, "初始化完成，返回值：" + status); + if (status == AIConstant.OPT_SUCCESS) { + Logger.d(TAG, "初始化成功!"); + } else { + Logger.d(TAG, "初始化失败!code:" + status); + } + } + + @Override + public void onError(String utteranceId, AIError error) { + Logger.d(TAG, "检测到错误:" + error.toString()); + } + + @Override + public void onSynthesizeStart(String utteranceId) { + Logger.d(TAG, "合成开始"); + } + + @Override + public void onSynthesizeDataArrived(String utteranceId, byte[] audioData) { + //Log.d(Tag, "合成pcm音频数据:" + audioData.length); + //正常合成结束后会收到size大小为0的audioData,即audioData.length == 0。应用层可以根据该标志停止播放 + //若合成过程中取消(stop或release)，则不会收到该结束标志 + } + + @Override + public void onSynthesizeFinish(String utteranceId) { + Logger.d(TAG, "合成结束"); + } + + @Override + public void onSpeechStart(String utteranceId) { + Logger.d(TAG, "开始播放"); + } + + @Override + public void onSpeechProgress(int currentTime, int totalTime, boolean isRefTextTTSFinished) { + Logger.d(TAG, "当前:" + currentTime + "ms, 总计:" + totalTime + "ms, 可信度:" + isRefTextTTSFinished); + } + + @Override + public void onSpeechFinish(String utteranceId) { + Logger.d(TAG, "播放完成"); + } + } } diff --git a/tts/tts-pad/src/main/jniLibs/arm64-v8a/libcntts.so b/tts/tts-pad/src/main/jniLibs/arm64-v8a/libcntts.so new file mode 100644 index 0000000000..36db975411 Binary files /dev/null and b/tts/tts-pad/src/main/jniLibs/arm64-v8a/libcntts.so differ diff --git a/tts/tts-pad/src/main/jniLibs/arm64-v8a/libduidns.so b/tts/tts-pad/src/main/jniLibs/arm64-v8a/libduidns.so new file mode 100644 index 0000000000..485f33df9d Binary files /dev/null and b/tts/tts-pad/src/main/jniLibs/arm64-v8a/libduidns.so differ diff --git a/tts/tts-pad/src/main/jniLibs/arm64-v8a/libduiutils.so b/tts/tts-pad/src/main/jniLibs/arm64-v8a/libduiutils.so new file mode 100644 index 0000000000..9fcd8446f8 Binary files /dev/null and b/tts/tts-pad/src/main/jniLibs/arm64-v8a/libduiutils.so differ diff --git a/tts/tts-pad/src/main/jniLibs/arm64-v8a/libliteca.so b/tts/tts-pad/src/main/jniLibs/arm64-v8a/libliteca.so new file mode 100644 index 0000000000..25f6ba8a15 Binary files /dev/null and b/tts/tts-pad/src/main/jniLibs/arm64-v8a/libliteca.so differ