语音语音合成科大讯飞和Tizen-TTS语音合成引擎

废话就不多说了，开始。。。

     最近在做一个文本转语音TTS(Text to Speech)的第三方软件封装，应用的是海内语音技术龙头安徽科大讯飞公司提供的离线引擎AiSound5.0，重要用于汽车导航用途。科大讯飞还提供了AiTalk用于语音识别，AiWrite用于手写识别服务等。另外还有针对6种平台的SDK和开辟示例。

    一、科大讯飞语音平台

科大讯飞目前有不少产品应用在移动终端上了，比如说用在手机上的讯飞语点，可方便语音拨打电话和发送短信，查气候、股票等信息。

在上面讯飞语音云这个网站可以找到科大讯飞针对6种平台的SDK，http://open.voicecloud.cn/download.php，包含Android、Iphone、Windows、Linux、Java、Flash这些平台的语音合成、语音识别和语音听写的开辟文档和SDK下载，不过应用之前须要请求AppID(每一个语音应用程序须要一个Appid来独一标识，您须要通过注册帐号来获得自己应用的Appid，未注册的语音应用将无法正常获取语音服务。)，通过考核后就可以下载响应的开辟文档和SDK并用于自己的软件系统中了。

 二 、科大讯飞语音合成、识别-在线文档和应用程序示例网址

从这里可以查看在线的开辟文档http://open.voicecloud.cn/developer.php?column=aW50ZV9zZGs%3D

Android平台的开辟文档和应用程序示例：http://open.voicecloud.cn/developer.php?category=YW5kcm9pZA%3D%3D&column=ZG9jdW1lbnQ%3D&type=d2lkZ2V0#a38

Windows和Linux平台的开辟文档和程序示例：http://open.voicecloud.cn/developer.php?category=b3RoZXI%3D&column=ZG9jdW1lbnQ%3D&type=YXBp

   三、科大讯飞语音产品的移动应用

这里有一个科大讯飞的在线语音合成系统演示程序：ViViVoice 2.1在线演示系统

http://www.iflytek.com/TtsDemo/viviVoiceShow.aspx

1、讯飞语点的应用，官网下载网址如下：http://yudian.voicecloud.cn/yudian.htm，支持Android和Iphone两大平台。

2、讯飞输入法应用：http://ime.voicecloud.cn/index.html

3、讯飞语音输入法应用：http://kouxun.voicecloud.cn/index.html

4、开辟者社区：http://open.voicecloud.cn/index.php

   四、官网的Windows平台-语音合成和识别示例代码

网址如下：http://open.voicecloud.cn/developer.php?category=b3RoZXI%3D&column=c2FtcGxl&type=YXBp

以Windows下的开辟为例，给出了语音合成、语音识别和语音听写的编程示例。

1、语音合成开辟例程

#include <stdio.h>
 #include <string.h>
 #include "qtts.h"
 #define END_SYNTH( reason ) \
 { \
 ret = QTTSSessionEnd( session_id, #reason ); \
 if( 0 != ret ) \
 { \
 printf("QTTSSessionEnd failed, error code is %d", ret ); \
 } \
 \
 ret = QTTSFini(); \
 if( 0 != ret ) \
 { \
 printf("QTTSFini failed, error code is %d", ret ); \
 } \
 }
 int main()
 {
 const char* configs = NULL;
 const char* session_id = NULL;
 const char* synth_params = NULL;
 const char* synth_text = NULL;
 unsigned int text_len = 0;
 const char* synth_speech = NULL;
 unsigned int synth_speech_len = 0;
 FILE* f_speech = NULL;
 int synth_status = 0;
 int ret = 0;
 printf( "===================================================================\n"
 " Mobile Speech Platform 2.0 Client SDK Demo for TTS \n"
 "===================================================================\n" );

 /* 初始化 */
configs = "server_url=dev.voicecloud.cn/index.htm, timeout=10000, coding_libs=speex.dll";
 ret = QTTSInit( configs );
 if( 0 != ret )
 {
 printf( "QTTSInit failed, error code is %d", ret );
 return -1;
 }
 /* 开始一路会话，应用会话模式 */
synth_params = "ssm=1, auf=audio/L16;rate=16000, aue=speex-wb;7, ent=intp65";
 session_id = QTTSSessionBegin( synth_params, &ret );
 if( 0 != ret )
 {
 printf( "QTTSSessionBegin failed, error code is %d", ret );
 return -1;
 }
 /* 写入合成文本 */
synth_text = "讯飞语音云为您提供了最新最好的语音技术休会，"
 "我们在互联网上开放科大讯飞最新研发的各种语音技术，"
 "包含世界领先的语音合成技术、语音识别技术、声纹识别技术等。";
text_len = strlen( synth_text );
 ret = QTTSTextPut( session_id, synth_text, text_len, NULL );
 if( 0 != ret )
 {
 printf( "QTTSTextPut failed, error code is %d", ret );
 END_SYNTH( QTTSTextPut failed! );
 return -1;
 }
 /* 获取合成音频 */
f_speech = fopen( "synth_speech.pcm", "wb" );
 if( NULL == f_speech )
 {
 printf( "Can not open file \"synth_speech.pcm\"" );
 END_SYNTH( open file );
 return -1;
 }
 while( TTS_FLAG_DATA_END != synth_status )
 {
 synth_speech = QTTSAudioGet( session_id, &synth_speech_len, &synth_status, &ret );
 if( 0 != ret )
 {
 printf( "QTTSAudioGet failed, error code is: %d", ret );
 break;
 }
 printf( "QTTSAudioGet ok, speech length = %d\n", synth_speech_len );
 if( NULL != synth_speech && 0 != synth_speech_len )
 {
 fwrite( synth_speech, 1, synth_speech_len, f_speech );
 }
 }
 fclose( f_speech );
 /* 结束会话，释放资源 */
ret = QTTSSessionEnd( session_id, "normal end" );
 if( NULL == f_speech )
 {
 printf( "QTTSSessionEnd failed, error code is %d", ret );
 }
 session_id = NULL;
 ret = QTTSFini();
 if( 0 != ret )
 {
 printf( "QTTSFini failed, error code is %d", ret );
 }
 return 0;
 }

2、语音识别开辟例程

    每日一道理
流逝的日子像一片片凋零的枯叶与花瓣，渐去渐远的是青春的纯情与浪漫。不记得曾有多少雨飘在胸前风响在耳畔，只知道沧桑早已漫进了我的心爬上了我的脸。当一个人与追求同行，便坎坷是伴，磨难也是伴。

#include <stdio.h>
 #include <string.h>
 #include <Windows.h>
 #include "qisr.h"
 #define END_RECOG( reason ) \
 { \ ret = QISRSessionEnd( session_id, #reason ); \
 if( 0 != ret ) \
 { \ printf("QISRSessionEnd failed, error code is %d", ret ); \
 } \
 \
 ret = QISRFini(); \
 if( 0 != ret ) \
 { \
 printf("QISRFini failed, error code is %d", ret ); \
 } \
 }
 #define BLOCK_LEN 5 * 1024
 int main()
 {
 const char* configs = NULL;
 const char* session_id = NULL;
 const char* recog_grammar = NULL;
 const char* recog_params = NULL;
 char recog_audio[ BLOCK_LEN ];
 FILE* f_speech = NULL;
 int audio_status = 0;
 int ep_status = 0;
 int rec_status = 0;
 int rslt_status = 0;
 const char* rec_result = NULL;
 unsigned int audio_len = 0;
 int ret = 0;
 printf( "===================================================================\n"
 " Mobile Speech Platform 2.0 Client SDK Demo for IAT \n"
 "===================================================================\n" );
 /* 初始化 */
configs = "server_url=dev.voicecloud.cn/index.htm, coding_libs=speex.dll, vad_enable=true";
 ret = QISRInit( configs );
 if( 0 != ret )
 {
 printf( "QISRInit failed, error code is %d\n", ret );
 return -1;
 }
 /* 开始一路会话，应用会话模式，应用引擎内置的语法停止识别 */
recog_grammar = "builtin:grammar/../search/location.abnf?language=zh-cn";
 recog_params = "ssm=1, aue=speex-wb;7, auf=audio/L16;rate=16000, “
 ”ent=map, vad_speech_tail=900";
 session_id = QISRSessionBegin( recog_grammar, recog_params, &ret );
 if( 0 != ret )
 {
 printf( "QISRSessionBegin failed, error code is %d\n", ret );
 return -1;
 }
 /* 打开用来停止识别的语音文件，用户可以采取其他的获取音频的方式比如实时采集音频 */
f_speech = fopen( "sxk_16k.pcm", "rb" );
 if( NULL == f_speech )
 {
 printf( "Can not open file \"sxk_16k.pcm\"\n" );
 END_RECOG( open file );
 return -1;
 }
 /* 发送音频数据，获取语音听写结果 */
while( ISR_AUDIO_SAMPLE_LAST != audio_status )
 {
 audio_len = fread( recog_audio, 1, BLOCK_LEN, f_speech );
 audio_status = ( audio_len == BLOCK_LEN ) ?
 ISR_AUDIO_SAMPLE_CONTINUE :ISR_AUDIO_SAMPLE_LAST;
 ret = QISRAudioWrite( session_id, recog_audio, audio_len
 , audio_status, &ep_status, &rslt_status );
 if( 0 != ret )
 {
 printf( "QISRSessionBegin failed, error code is %d\n", ret );
 rslt_status = ISR_REC_STATUS_SPEECH_COMPLETE;
 break;
 }
 printf( "write audio data ok! len=%d, status=%d\n", audio_len, audio_status );
 /* 已经有结果缓存在MSC中了，可以获取了 */
if( ISR_REC_STATUS_SUCCESS == rslt_status )
 {
 rec_result = QISRGetResult( session_id, &rslt_status, 5000, &ret );
 if( 0 != ret )
 {
 printf( "QISRGetResult failed, error code is %d\n", ret );
 rslt_status = ISR_REC_STATUS_SPEECH_COMPLETE;
 break;
 }
 if( NULL != rec_result )
 { printf( "got a result: %s\n", rec_result );
 }
 /* 全部结果已经取完了 */
if( ISR_REC_STATUS_SPEECH_COMPLETE == rslt_status )
 {
 printf( "the result has been got completely!\n" );
 break;
 }
 }
 /* 检测到音频后端点，结束音频发送 */
if( ISR_EP_AFTER_SPEECH == ep_status )
 { printf( "end point of speech has been detected!\n" );
 break;
 }
 Sleep( 160 );
 }
 fclose( f_speech );
 /* 获取余下的识别结果 */
while( ISR_REC_STATUS_SPEECH_COMPLETE != rslt_status )
 {
 rec_result = QISRGetResult( session_id, &rslt_status, 5000, &ret );
 if( 0 != ret )
 {
 printf( "QISRGetResult failed, error code is: %d\n", ret );
 break;
 }
 if( NULL != rec_result )
 {
 printf( "got a result: %s\n", rec_result );
 }
 /* sleep一下很有必要，防止MSC端无缓存的识别结果时浪费CPU资源 */
Sleep( 200 );
 }
 /* 结束会话，释放资源 */
ret = QISRSessionEnd( session_id, "normal end" );
 if( NULL == f_speech )
 {
 printf( "QISRSessionEnd failed, error code is %d\n", ret );
 }
 session_id = NULL;
 ret = QISRFini();
 if( 0 != ret )
 {
 printf( "QISRFini failed, error code is %d\n", ret );
 }
 return 0;
 }

3、语音听写开辟例程

#include <stdio.h>
 #include <string.h>
 #include <Windows.h>
 #include "qisr.h"
 #define END_RECOG( reason ) \
 { \
 ret = QISRSessionEnd( session_id, #reason ); \
 if( 0 != ret ) \
 { \
 printf("QISRSessionEnd failed, error code is %d", ret ); \
 } \
 \
 ret = QISRFini(); \
 if( 0 != ret ) \
 { \
 printf("QISRFini failed, error code is %d", ret ); \
 } \
 }
 #define BLOCK_LEN 5 * 1024
 int main()
 {
 const char* configs = NULL;
 const char* session_id = NULL;
 const char* recog_params = NULL;
 char recog_audio[ BLOCK_LEN ];
 FILE* f_speech = NULL;
 int audio_status = 0;
 int ep_status = 0;
 int rec_status = 0;
 int rslt_status = 0;
 const char* rec_result = NULL;
 unsigned int audio_len = 0;
 int ret = 0;
 printf( "===================================================================\n"
 " Mobile Speech Platform 2.0 Client SDK Demo for IAT \n"
 "===================================================================\n" );
 /* 初始化 */
configs = "server_url=dev.voicecloud.cn/index.htm, coding_libs=speex.dll, vad_enable=true";
 ret = QISRInit( configs );
 if( 0 != ret )
 {
 printf( "QISRInit failed, error code is %d\n", ret );
 return -1;
 }
 /* 开始一路会话 */
recog_params = "ssm=1, sub=iat, aue=speex-wb;7, auf=audio/L16;rate=16000, “
 “ent=sms16k, rst=plain, vad_speech_tail=1500";
 session_id = QISRSessionBegin( NULL, recog_params, &ret );
 if( 0 != ret )
 {
 printf( "QISRSessionBegin failed, error code is %d\n", ret );
 return -1;
 }
 /* 打开用来停止识别的语音文件，用户可以采取其他的获取音频的方式比如实时采集音频 */
f_speech = fopen( "IAT_16KPCM_10s_0.pcm", "rb" );
 if( NULL == f_speech )
 {
 printf( "Can not open file \"IAT_16KPCM_10s_0.pcm\"\n" );
 END_RECOG( open file );
 return -1;
 }
 /* 发送音频数据，获取语音听写结果 */
while( ISR_AUDIO_SAMPLE_LAST != audio_status )
 {
 audio_len = fread( recog_audio, 1, BLOCK_LEN, f_speech );
 audio_status = ( audio_len == BLOCK_LEN ) ?
 ISR_AUDIO_SAMPLE_CONTINUE :ISR_AUDIO_SAMPLE_LAST;
 ret = QISRAudioWrite( session_id, recog_audio, audio_len, audio_status, &ep_status, &rslt_status );
 if( 0 != ret )
 {
 printf( "QISRSessionBegin failed, error code is %d\n", ret );
 rslt_status = ISR_REC_STATUS_SPEECH_COMPLETE;
 break;
 }
 printf( "write audio data ok! len=%d, status=%d\n", audio_len, audio_status );
 /* 已经有结果缓存在MSC中了，可以获取了 */
if( ISR_REC_STATUS_SUCCESS == rslt_status )
 { rec_result = QISRGetResult( session_id, &rslt_status, 5000, &ret );
 if( 0 != ret )
 {
 printf( "QISRGetResult failed, error code is %d\n", ret );
 rslt_status = ISR_REC_STATUS_SPEECH_COMPLETE;
 break;
 }
 if( NULL != rec_result )
 {
 printf( "got a result: %s\n", rec_result );
 }
 /* 全部结果已经取完了 */
if( ISR_REC_STATUS_SPEECH_COMPLETE == rslt_status )
 {
 printf( "the result has been got completely!\n" );
 break;
 }
 }
 /* 检测到音频后端点，结束音频发送 */
if( ISR_EP_AFTER_SPEECH == ep_status )
 {
 printf( "end point of speech has been detected!\n" );
 break;
 }
 Sleep( 160 );
 }
 fclose( f_speech );
 /* 获取余下的识别结果 */
while( ISR_REC_STATUS_SPEECH_COMPLETE != rslt_status )
 {
 rec_result = QISRGetResult( session_id, &rslt_status, 5000, &ret );
 if( 0 != ret )
 {
 printf( "QISRGetResult failed, error code is: %d\n", ret );
 break;
 }
 if( NULL != rec_result )
 {
 printf( "got a result: %s\n", rec_result );
 }
 /* sleep一下很有必要，防止MSC端无缓存的识别结果时浪费CPU资源 */
Sleep( 200 );
 }
 /* 结束会话，释放资源 */
ret = QISRSessionEnd( session_id, "normal end" );
 if( NULL == f_speech )
 {
 printf( "QISRSessionEnd failed, error code is %d\n", ret );
 }
 session_id = NULL;
 ret = QISRFini();
 if( 0 != ret )
 {
 printf( "QISRFini failed, error code is %d\n", ret );
 }
 return 0;
 }

   五、Tizen(泰泽)提供的TTS和STT语音合成、语音识别接口

泰泽是三星和英特尔合作开辟的一款操作系统。

1、Tizen SDK官网：https://developer.tizen.org/downloads/tizen-sdk

2、Tizen项目：https://review.tizen.org/git/

3、tts-api:https://review.tizen.org/git/?p=framework/api/tts-api.git;a=summary

其对应的tts.h头文件api接口在线网址为：https://review.tizen.org/git/?p=framework/api/tts-api.git;a=blob;f=include/tts.h;h=636470d923555a30d164cc09ceff841b72187e20;hb=98aee0bab00a418af162a0314ef931f8fd620892

4、stt-api：https://review.tizen.org/git/?p=framework/api/stt-api.git;a=summary

文章结束给大家分享下程序员的一些笑话语录：

问路
有一个驾驶热气球的人发现他迷路了。他降低了飞行的高度，并认出了地面 上的一个人。他继续下降高度并对着那个人大叫，“打扰一下，你能告诉我我 在哪吗？”
下面那个人说：“是的。你在热气球里啊，盘旋在 30 英尺的空中”。
热气球上的人说：“你一定是在 IT 部门做技术工作”。
“没错”，地面上的人说到，“你是怎么知道的？”
“呵呵”，热气球上的人说，“你告诉我的每件事在技术上都是对的，但对都没 有用”。
地面上的人说，“你一定是管理层的人”。
“没错”，热气球上的人说，“可是你是怎么知道的？”
“呵呵”，地面上的那人说到，“你不知道你在哪里，你也不知道你要去哪，你 总希望我能帮你。你现在和我们刚见面时还在原来那个地方，但现在却是我 错了”。

---------------------------------
原创文章 By
语音和语音合成
---------------------------------