如何在vc 6 0/2003/2005/2008中调用ATK

最新推荐文章于 2024-06-05 15:23:05 发布

我漂亮吗

最新推荐文章于 2024-06-05 15:23:05 发布

阅读量148

点赞数

CC 4.0 BY-SA版权

本文链接：https://blog.youkuaiyun.com/fdgugfv/article/details/83756845

本文详细介绍了一个基于HTK的实时API应用实例，通过列举代码片段展示了如何处理和识别音视频文件，包括列表化.wav和.vec文件，加载样本集，以及进行信号源向量的匹配和比较。文章还涉及了音量标志的设置及其对识别过程的影响。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

分享一下我老师大神的人工智能教程！零基础，通俗易懂！http://blog.youkuaiyun.com/jiangjunshow

也欢迎大家转载本篇文章。分享知识，造福人民，实现我们中华民族伟大复兴！

Application Example using the ATK Real-Time API

下面是一个关于HTK的例子

#include "stdafx.h"
#include "dllSudx.h"
typedef vector<STRING> TStringArray;
struct ASampleSet
{
char name[256];
TSoundTag tags[301];
};
typedef vector<ASAMPLESET> TSampleSetList;
TStringArray sampleList;
TSampleSetList sampleSetList;
int vector_count = 0;
int ListWavFiles(char* dir, TStringArray& fileList)
{
struct _finddata_t c_file;
long hFile;
char tmp[1024];
int retVal = 0;
sprintf(tmp, "%s//%s", dir, "*.wav");
/* Find first .c file in current directory */
if( (hFile = _findfirst( tmp, &c_file )) == -1L )
return retVal;
else
{
sprintf(tmp, "%s//%s", dir, c_file.name);
fileList.push_back(tmp);
retVal++;
while( _findnext( hFile, &c_file ) == 0 )
{
sprintf(tmp, "%s//%s", dir, c_file.name);
fileList.push_back(tmp);
retVal++;
}
_findclose( hFile );
}
return retVal;
}
int ListVecFiles(char* dir, TStringArray& fileList)
{
struct _finddata_t c_file;
long hFile;
char tmp[1024];
int retVal = 0;
sprintf(tmp, "%s//%s", dir, "*.vec");
/* Find first .c file in current directory */
if( (hFile = _findfirst( tmp, &c_file )) == -1L )
return retVal;
else
{
sprintf(tmp, "%s//%s", dir, c_file.name);
fileList.push_back(tmp);
retVal++;
while( _findnext( hFile, &c_file ) == 0 )
{
sprintf(tmp, "%s//%s", dir, c_file.name);
fileList.push_back(tmp);
retVal++;
}
_findclose( hFile );
}
return retVal;
}
int LoadVecFiles(TStringArray& fileList, TSampleSetList& sampleSetList)
{
int retVal = 0;
for(int i=0; i<FILELIST.SIZE(); <声音文件名 printf(?使用方法：test printf(?参数错误/n?); { *if(argc!="4)" bVolumeFlag) bool , vecDirectory char* *VoiceName, VRecognize16BitWav(char int ; g_TestVectorFileName="test.vec" string } FALSE; : TRUE ? matchCount="0;" return matchCount++; i+="3;" pMatchResult[matchCount].dwMatchOffset="i;" pMatchResult[matchCount].ucScore="score;" &score)) bMode, pSampleVec, if(sudxCompareEx(pSrcVec+i, score; i++) matchCount<100; && i<nSrcLen-tailLen i="0;" for(int tailLen="64;" else 0; if(nSampleLen<300) if(nSampleLen<64) if(bMode="=TRUE)" tailLen; 输出匹配结果（得分、偏移量） pMatchResult) MATCHRESULT* 输出匹配结果个数 pnMaxMatch, int* 模式。="TRUE：快速模式；=FALSE：准确模式" BOOL 样本向量的个数 nSampleLen, 样本向量的首地址指针 TSoundTag* 信号源向量的长度（以TSoundTag为计量单位） nSrcLen, 待识别信号源向量的首地址指针 pSrcVec, sudxMatch(TSoundTag* 注意事项：pnMaxMatch的存储空间由应用程序分配，最大为100条记录。函数返回：1：调用成功。0：调用失败 MATCHRESULT; dwMatchOffset; DWORD ucScore; char unsigned tagMATCHRESULT struct typedef retVal; ); fflush(stdout fileList[i].c_str()); [%s]/n?, file open not printf(?can fclose(fp); !/n?, OK set[%s] printf(?sample retVal++; sampleSetList.push_back(aset); strcpy(aset.name, short!/n?, too if(size!="80)" fp); 80, sizeof(TSoundTag), size="fread(aset.tags," aset; ASampleSet if(fp) ?rb?); fp="fopen(fileList[i].c_str()," FILE*> <样本目录> <音量标志>/n");
printf("/t声音文件名是要被识别的声音文件，样本目录中包含已经预先处理好的样本/n"
"/t 音量标志: 表示识别时是否采用音量相关的方式,为0表示识别时不管音量大小，为1表示识别时需要考虑音量大小/n"
"Press any key continue...");
getch();
return -1;
}*/
//int volFlag = atol(argv[3]);
if(bVolumeFlag)
{
sudxSetCompVolumeFlag(TRUE);
printf("采用与音量相关的方式比较/n");
}
else
{
sudxSetCompVolumeFlag(FALSE);
printf("采用与音量无关的方式比较/n");
}
TStringArray sampleList;
int i = ListVecFiles(vecDirectory, sampleList);
if(i<=0)
{
printf("No sample files in dir [%s]/n", vecDirectory);
return -2;
}
TSampleSetList sampleSetList;
int vector_count = LoadVecFiles(sampleList, sampleSetList);
if(vector_count<=0)
{
printf("没有找到样本集合文件/n");
return -3;
}
printf("Now start.../n");
const int bufSampleSize = 49306; ////充值成功,你的账户余额为的 8k采样,8位长度编码的语音文件大小
unsigned char * p8bitBuf = new unsigned char[ bufSampleSize];
short *pFileBuf = new short[bufSampleSize*2]; //每次处理2M个16bit pcm样本，也就是4M byte数据。
// char outFile[1000];
// sprintf(outFile, "%s//%s.vec", "c:", VoiceName);
// FILE* fout = fopen(outFile, "w+b"); //打开或创建向量文件
FILE *fp = fopen(VoiceName, "rb");
fseek(fp, 0L, SEEK_END);
long size = ftell(fp);
fseek(fp, 44L, SEEK_SET); //跳过wav头
int old_clock = clock();
// long segSize = fread(p8bitBuf, sizeof(char), bufSampleSize, fp);
//alaw2linear( p8bitBuf , bufSampleSize , pFileBuf );
long segSize = fread(pFileBuf, sizeof(short), bufSampleSize, fp);
TSoundTag* pOut;
int outSize, tailSize;
short* pTail=NULL;
int last_result[17];
int iMax = sudxGetMaxInputSampleSize();
int error = sudxCalcBuffer( pFileBuf, segSize, &pOut, &outSize, &pTail, &tailSize, last_result);
if ( error > 0 )
{
if ( outSize > (64 + 1 ))
{
old_clock = clock();
int compare_turn = outSize - 64 - 1;
for(i=0; i<COMPARE_TURN; { int i++) for(int [%s]/n?, file if(tailSize tailSize="compare_turn+64-300-i;" sampleSetList[j].name); (i+1)*128, %d, pos sample at Match printf(?Quick ) FALSE) &(sampleSetList[j].tags[1]), sudxCompare(pOut+1+i, if( j++) j<vector_count; j="0;">300)
{
if(sudxCompare(pOut+1+i, sampleSetList[j].tags+1, TRUE))
printf("精确匹配位置 %d, file[%s]/n", (i+1)*128, sampleSetList[j].name);
else
printf("!!Slow Not Match at sample pos %d, file[%s]/n", (i+1)*128, sampleSetList[j].name);
}
}
else
{
printf("Not Match at sample pos %d, file [%s]/n", (i+1)*128, sampleSetList[j].name);
}
}
}
int time_passed = clock()-old_clock;
printf("计算结束, 耗时%d ms/n", time_passed);
printf("比较速度为每秒钟可查找%f秒音乐/n", double(compare_turn)*128.0/8000.0/(double(time_passed)/1000.0));
}
else
{
printf("输入的声音过短,不能比较/n");
return -2;
}
}
else
{
return -1;
}
return 0;
}
int VRecoInit(char * vecDirectory )
{
int i = ListVecFiles(vecDirectory, sampleList);
if(i<=0)
{
printf("No sample files in dir [%s]/n", vecDirectory);
return -2;
}
vector_count = LoadVecFiles(sampleList, sampleSetList);
if(vector_count<=0)
{
printf("没有找到样本集合文件/n");
return -3;
}
return 0;
}
///
int VRecognize(char *VoiceName, char* vecDirectory , bool bVolumeFlag, int * iMatchCount)
{
/*if(argc!=4)
{
printf("参数错误/n");
printf("使用方法：test <声音文件名> <样本目录> <音量标志>/n");
printf("/t声音文件名是要被识别的声音文件，样本目录中包含已经预先处理好的样本/n"
"/t 音量标志: 表示识别时是否采用音量相关的方式,为0表示识别时不管音量大小，为1表示识别时需要考虑音量大小/n"
"Press any key continue...");
getch();
return -1;
}*/
//int volFlag = atol(argv[3]);
if(bVolumeFlag)
{
sudxSetCompVolumeFlag(TRUE);
printf("采用与音量相关的方式比较/n");
}
else
{
sudxSetCompVolumeFlag(FALSE);
printf("采用与音量无关的方式比较/n");
}
char gcaMsg[1000];
sprintf(gcaMsg , "Now start...%s/n", VoiceName);
printf( gcaMsg );
fflush(stdout);
const int bufSampleSize = 2048*1024; ////充值成功,你的账户余额为的 8k采样,8位长度编码的语音文件大小
unsigned char * p8bitBuf = new unsigned char[ bufSampleSize];
short *pFileBuf = new short[bufSampleSize]; //每次处理2M个16bit pcm样本，也就是4M byte数据。
// char outFile[1000];
// sprintf(outFile, "%s//%s.vec", "c:", VoiceName);
// FILE* fout = fopen(outFile, "w+b"); //打开或创建向量文件
try
{
sprintf(gcaMsg , "bf fopen...%s/n", VoiceName);
printf( gcaMsg );
fflush(stdout);
FILE *fp = fopen(VoiceName, "rb");
fseek(fp, 0L, SEEK_END);
long size = ftell(fp);
fseek(fp, 44L, SEEK_SET); //跳过wav头
sprintf(gcaMsg , "af fopen...%s/n", VoiceName);
printf( gcaMsg );
fflush(stdout);
int old_clock = clock();
long segSize = fread(p8bitBuf, sizeof(char), bufSampleSize, fp);
alaw2linear( p8bitBuf , bufSampleSize , pFileBuf );
sprintf(gcaMsg , "Af alaw2linear...%s/n", VoiceName);
printf( gcaMsg );
fflush(stdout);
// long segSize = fread(pFileBuf, sizeof(short), bufSampleSize, fp);
TSoundTag* pOut;
int outSize, tailSize;
short* pTail=NULL;
int last_result[17];
//int iMax = sudxGetMaxInputSampleSize();
int error = sudxCalcBuffer( pFileBuf, segSize, &pOut, &outSize, &pTail, &tailSize, last_result);
if ( error > 0 )
{
sprintf(gcaMsg , "Af sudxCalcBuffer...%s/n", VoiceName);
printf( gcaMsg );
fflush(stdout);
if ( outSize > (64 + 1 ))
{
old_clock = clock();
int compare_turn = outSize - 64 - 1;
for(int i=0; i<COMPARE_TURN; { int i++) for(int [%s]/n?, file if(tailSize tailSize="compare_turn+64-300-i;" sampleSetList[j].name); (i+1)*128, %d, pos sample at Match printf(?Quick ) FALSE) &(sampleSetList[j].tags[1]), sudxCompare(pOut+1+i, if( j++) j<vector_count; j="0;" iMatchCount)++; (*>300)
{
if(sudxCompare(pOut+1+i, sampleSetList[j].tags+1, TRUE))
{
(* iMatchCount) ++;
printf("精确匹配位置 %d, file[%s]/n", (i+1)*128, sampleSetList[j].name);
}
else
{
// printf("!!Slow Not Match at sample pos %d, file[%s]/n", (i+1)*128, sampleSetList[j].name);
}
&nb
给我老师的人工智能教程打call！http://blog.youkuaiyun.com/jiangjunshow