如何在vc++6.0/2003/2005/2008中调用ATK

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);
}
}
}
}
}
int time_passed = clock()-old_clock;
printf("计算结束, 耗时%d ms Match Count =%d/n", time_passed,(* iMatchCount) );
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));
fflush( stdout );
}
else
{
printf("输入的声音过短,不能比较/n");
return -2;
}
}
else
{
sprintf(gcaMsg , "Now Fail ...%s/n return -1", "CalcBuffer");
printf( gcaMsg );
fflush(stdout);
return -1;
}
}
catch ( ... )
{
sprintf(gcaMsg , "Now Fail ...%s/n return -3", "Catch");
printf( gcaMsg );
fflush(stdout);
return -3;
}
return 0;
}
char * asds_version="!HVER!ASDS: 1.6.0 [SJY 01/06/07]";
#include "AMonitor.h"
#include "AIO.h"
// ---------------- Globals To Define the Recognition System -----------------
class QA;
// Information Channel - connects to AIO output
static ABuffer *inChan; // input channel from AIO
// Active components (threads)
// NB: AIO will start-up its own ASource,ASyn,ACode & ARec
static AIO *aio; // the AIO subsystem
static AMonitor *amon; // system monitor
// Global resources - these will be passed to AIO on creation
ARMan *rman; // resource manager for dict, grammars and HMMSet
AHmms *hset; // HMM set is global since it never changes
ADict *dict; // ditto dictionary (though it can be edited if desired)
AGram *ggrm; // global grammar, always active
QA *qty, *top; // question/answer instances
// ------------------ Simple Question/Answer Class ----------------------
// the status of each slot determines the dialogue flow. each slot
// starts as unknown. When an answer is provided with high confidence
// (>minconf), the slot is immediately grounded. Otherwise the slot
// status moves to unconfirmed ready for explicit checking.
typedef enum { unknown, unconfirmed, grounded, cancelled } SlotStatus;
// the system listens to the user in QA.listen, this implements a simple
// state machine where the states are:
typedef enum { aActive,aInProcess,aTimeout,aDone,aTerminated} ASRStatus;
typedef enum { sActive, sDone, sInt } SynStatus;
static Boolean terminated = FALSE;
static const float minconf = 0.5;
class QA {
public:
QA(const string& aname, // name of qa object
const string& aprompt, // query prompt
const string& gramfile, // name of grammar file
const string& ahelp); // help message
ASRStatus Listen(string prompt, string rgroup, SynStatus& ss);
// output prompt and listen using given res group and update slot value
void Ask(); // ask question, record answer and set status
void Check(); // check answer and set status
void GetSlot(); // get a value for the slot by asking and checking
string GetValue(); // strip tag and return actual value
void Show(); // show current slot status
void Reset();
SlotStatus status; // slot status
string value; // slot value
float curconf; // current input
string curtag;
string curwords;
private:
string name; // name of slot (also name of semantic tag)
string prompt; // question
string help; // help string
ResourceGroup *ask;
ResourceGroup *chk;
};
// construct a QA object with given prompt and grammar
QA::QA(const string& aname, const string& aprompt,
const string& gramfile, const string& ahelp)
{
// save the prompts
prompt = aprompt; help = ahelp; name = aname;
// create grammar specific to this question
AGram *g1 = new AGram(name,gramfile);
rman->StoreGram(g1);
ask = rman->NewGroup(name+"-ask");
ask->AddHMMs(hset); // Add the global resources
ask->AddDict(dict);
ask->AddGram(ggrm); // include global grammar in parallel with
ask->AddGram(g1); // qa specific grammar
// create copy of ask but prepend "No" to front of qa grammar
// and add confirm in parallel
chk = rman->NewGroup(name+"-chk");
chk->AddHMMs(hset); // Add the global resources
chk->AddDict(dict);
chk->AddGram(ggrm); // include global grammar
AGram *g2 = new AGram(*g1); // copy g1
rman->StoreGram(g2);
g2->OpenEdit(); // Modify copy of g1 by prepending
GramSubN *s = g2->main; // "NO" to front with skip to end
GramNode *ent = s->NewNullNode(99);
GramNode *no = s->NewWordNode("NO","no");
GramNode *yes = s->NewCallNode("confirm","yes"); // Add confirm in //
s->AddLink(ent,no); s->AddLink(ent,yes); s->AddLink(yes,s->exit);
s->AddLink(no,s->entry); s->AddLink(no,s->exit);
s->entry = ent;
g2->CloseEdit();
chk->AddGram(g2); // Add the new "correction" grammar
// initial status is unknown
status = unknown; value = "";
}
// Reset slot values for a new dialogue
void QA::Reset()
{
status = unknown; value = "";
}
// For debugging only
void QA::Show()
{
string s = "???";
switch (status){
case unknown: s = "unknown"; break;
case unconfirmed: s = "unconfirmed"; break;
case grounded: s = "grounded"; break;
case cancelled: s = "cancelled"; break;
}
printf("Slot: %s: status=%s value=%s [cur=%s(%s) %.1f]/n",
name.c_str(),s.c_str(),value.c_str(),curtag.c_str(),curwords.c_str(),curconf);
}
// listen to rec packets and construct slot value
// return system status when finished listening
ASRStatus QA::Listen(string prompt, string rgroup, SynStatus& ss)
{
APacket p;
string s="";
float conf=0.0;
int numwords=0;
ASRStatus as;
curtag=""; curwords="";
// set the resource group to use and start recognising
string grp = "setgroup("+rgroup+")";
string prmpt = "ask(/""+prompt+"/")";
printf("%s --> %s/n",prmpt.c_str(),grp.c_str());
aio->SendMessage(grp);
aio->SendMessage(prmpt);
// listen to response of form "tag(w1 w2 ...)"
// Following implements a simple state machine ss= syn state as=asr state
ss = sActive; as = aActive;
do {
AStringData *sd;
ACommandData *cd;
APhraseData *pd;
string cmd;
p = inChan->GetPacket();
switch(p.GetKind()) {
case StringPacket:
sd = (AStringData *)p.GetData();
if (sd->data.find("TERMINATED") != string::npos) {
as = aTerminated;
}
break;
case CommandPacket:
cd = (ACommandData *)p.GetData();
cmd = cd->GetCommand();
if (cmd=="synFinished")
ss = sDone;
else if (cmd=="synInterrupted")
ss = sInt;
else if (cmd=="asrTimeOut")
as = aTimeout;
break;
case PhrasePacket:
pd = (APhraseData *)p.GetData();
if (as == aActive && pd->ptype == OpenTag_PT){
assert(s == "" && numwords == 0);
as = aInProcess;
}else if (as == aInProcess && pd->ptype == Word_PT) {
if (s=="") s = pd->word; else s = s + " " + pd->word;
// printf("{%s[%.2f]}",pd->word.c_str(),pd->confidence);
conf += pd->confidence; ++numwords;
}else if (as == aInProcess &&
(pd->ptype == CloseTag_PT || pd->ptype == End_PT)){
curtag = pd->tag;
curwords = s;
as = aDone;
}
break;
}
} while (as <= aInProcess);
curconf = numwords==0?0.0:conf/numwords;
printf("/n Response: %s(%s)[%.1f]/n",curtag.c_str(),curwords.c_str(),curconf);
return as;
}
// ask for information
void QA::Ask()
{
Boolean ok = FALSE;
ASRStatus as; SynStatus ss;
do {
as=Listen(prompt,ask->gname,ss);
if (as==aDone){
if (curtag == "command") {
if (curwords == "HELP") {
aio->SendMessage("tell(/""+help+"/")");
}else if (curwords == "CANCEL"){
aio->SendMessage("tell(/"Cancelled/")");
ok = TRUE; status = cancelled;
}
}
if (curtag == name){
if (curconf > minconf){
status = grounded;
}else{
status = unconfirmed;
}
value = curtag+"("+curwords+")";
ok = TRUE;
}
}else if (as==aTimeout) {
printf("Timeout/n");
aio->SendMessage("tell(/"Sorry, I didnt hear that!/")");
}else if (as==aTerminated) {
terminated = TRUE;
}
} while (!ok && !terminated);
}
// confirm current slot value
void QA::Check()
{
Boolean ok = FALSE;
ASRStatus as; SynStatus ss;
string prompt;
do {
prompt = "Are sure you want "+GetValue()+" ?";
as = Listen(prompt,chk->gname,ss);
if (as==aDone){
if (curtag == "command") {
if (curwords == "HELP") {
aio->SendMessage("tell(/"I am trying to confirm your last input./")");
aio->SendMessage("tell(/"Please say Yes or No. If No, you can also give the correct value./")");
}else if (curwords == "CANCEL"){
ok = TRUE; status = cancelled;
}
}
if (curtag == "yes"){
ok = TRUE; status = grounded;
}else if (curtag == name){
if (curconf > minconf){
status = grounded;
}else{
status = unconfirmed;
}
value = curtag+"("+curwords+")";
ok = TRUE;
}else {
ok = TRUE; status = unknown; // user must have said No without correction
}
}else if (as==aTimeout) {
aio->SendMessage("Sorry, I didnt hear that!");
}else if (as==aTerminated) {
terminated = TRUE;
}
} while (!ok && !terminated);
}
void QA::GetSlot()
{
do {
Ask();
while (status == unconfirmed) Check();
}while (status != grounded && status != cancelled && !terminated);
}
string QA::GetValue()
{
int n = value.find("(");
if (n==string::npos) return "";
++n;
int m = value.length() - n -1;
string val = value.substr(n,m);
for (int i = 0; i<M; rman- AGram(?GGram?,?global.net?); ggrm="new" ADict(?ADict?); dict="new" AHmms(?HmmSet?); hset="new" resources fixed global create ARMan; rman="new" manager resource a ABuffer(?inChan?); inChan="new" AIO from channel return the { BuildRecogniser() void } getchar(); Continue/n/n?); to Return Press printf(?='==========================================/n/n");' microphone./n/n?); by up picked not printf(?is output/n?); speech that ensure headphones printf(?Use microphone./n?); open an uses demo this - printf(?/nWARNING System/n?); Dialog Spoken Asynchronous ATK printf(?/nASDS: HTK_Error(9999); throw HTK/n?); initialise cannot HRError(9999,?ASDS: (InitHTK(argc,argv,asds_version)<SUCCESS){ if *argv[]) char argc, Initialise(int --------------------------- Code Initialisation ---------------------- val; val[i]="tolower(ch);" ch="val[i];" int i++){>StoreHMMs(hset);
rman->StoreDict(dict);
rman->StoreGram(ggrm);
ResourceGroup *main = rman->NewGroup("main");
main->AddHMMs(hset); main->AddDict(dict); main->AddGram(ggrm);
// create the qa objects, these will create the resource groups
top = new QA("topping",
"What topping would you like?",
"topping.net",
"I have a variety of toppings/n - try your favourite combination");
qty = new QA("quantity",
"How many pizzas would you like?",
"howmany.net",
"I need to know how many pizzas to deliver");
// now we have resources, create aio
aio = new AIO("aio",inChan,rman);
aio->DefineFiller("ER");
aio->DefineFiller("SIL");
aio->DefineFiller("OH");
// finally create Monitor and attach it to AIO
amon = new AMonitor;
aio->AttachMonitor(amon);
}
void StartRecogniser()
{
// Start up each component thread
amon->Start();
aio->Start();
}
void ShutDown()
{
printf("Shutting down/n");
// wait till components shutdown
aio->SendMessage("closedown()");
// then kill monitor
amon->Terminate();
HJoinMonitor();
}
// ------------------------- Application Code ---------------------------
void RunApplication()
{
// Start recogniser running
StartRecogniser();
do{
top->GetSlot();
if (!terminated && top->status == grounded) qty->GetSlot();
if (top->status == grounded && qty->status == grounded) {
string order = "Your order is "+qty->GetValue()+" "+top->GetValue();
if (qty->GetValue() == "one")
order += " pizza.";
else
order += " pizzas.";
printf("Order: %s/n",order.c_str());
aio->SendMessage("tell(/""+order+"/")");
}
top->Reset(); qty->Reset();
printf("/n/n=====================================/n/n");
}while(!terminated);
ShutDown();
}
// --------------------------- Main Program ----------------------------
int main(int argc, char *argv[])
{
try {
Initialise(argc,argv);
BuildRecogniser();
RunApplication();
}
catch (ATK_Error e){
int n = HRErrorCount();
printf("ATK Error %d/n",e.i);
for (int i=1; i<=n; i++)
printf(" %d. %s/n",i,HRErrorGetMess(i));
}
catch (HTK_Error e){
int n = HRErrorCount();
printf("HTK Error %d/n",e.i);
for (int i=1; i<=n; i++)
printf(" %d. %s/n",i,HRErrorGetMess(i));
}
return 0;
}

posted on 2008-12-27 00:08 Yincheng 阅读(367) 评论(0) 编辑收藏举报

刷新页面返回顶部

公告