www.pudn.com > 语音语谱图与基频.rar > speechView.cpp
// speechView.cpp : implementation of the CSpeechView class
//
#include "stdafx.h"
#include "speech.h"
#include
#include "speechDoc.h"
#include "speechView.h"
//#include "fftw3.h"
#include
#include "fstream.h"
#ifdef _DEBUG
#define new DEBUG_NEW
#undef THIS_FILE
static char THIS_FILE[] = __FILE__;
#endif
#define FFTSIZE 1024
#define WINTIME 30
#define PI 3.14
#define STEPTIME 10
#define Samplerate 16000
/////////////////////////////////////////////////////////////////////////////
// CSpeechView
IMPLEMENT_DYNCREATE(CSpeechView, CView)
BEGIN_MESSAGE_MAP(CSpeechView, CView)
//{{AFX_MSG_MAP(CSpeechView)
ON_COMMAND(ID_FILE_OPEN, OnFileOpen)
//}}AFX_MSG_MAP
END_MESSAGE_MAP()
/////////////////////////////////////////////////////////////////////////////
// CSpeechView construction/destruction
CSpeechView::CSpeechView()
{
// TODO: add construction code here
m_pWaveBuffer = NULL;
m_flag=0;
}
CSpeechView::~CSpeechView()
{
delete m_pWaveBuffer;
}
BOOL CSpeechView::PreCreateWindow(CREATESTRUCT& cs)
{
// TODO: Modify the Window class or styles here by modifying
// the CREATESTRUCT cs
return CView::PreCreateWindow(cs);
}
/////////////////////////////////////////////////////////////////////////////
// CSpeechView drawing
void CSpeechView::OnDraw(CDC* pDC)
{
CSpeechDoc* pDoc = GetDocument();
if(m_flag)
{
//m_flag=0;
SHORT *psWaveData;
unsigned long dwSpeechDataNum;
//读出声音信号后，准备画波形图
psWaveData = (short*) m_pWaveBuffer;
//dwSpeechDataNum是声音文件的点数
//总的时间可以通过点数除以采样率的时间秒
dwSpeechDataNum = m_nWaveLength / sizeof(short);
int nMiddle;
CPoint pt;
unsigned long i;
CRect rect;
GetClientRect(&rect);
int t=rect.right-rect.left-30;
//坐标轴的绘制
int pic1;
int pic2;
int pic3;
pic3=rect.bottom-25;
pDC->MoveTo(0,pic3);
pDC->LineTo(rect.right,pic3);
pic2=pic3-rect.bottom/3;
pDC->MoveTo(0,pic2);
pDC->LineTo(rect.right,pic2);
pic1=pic2-rect.bottom/3;
pDC->MoveTo(0,pic1);
pDC->LineTo(rect.right,pic1);
pDC->MoveTo(30,0);
pDC->LineTo(30,rect.bottom);
nMiddle=pic1/2;
//起始点
pt.x=30;
pt.y=nMiddle;
// MoveToEx(hMemDC, pt.x, pt.y, NULL);
pDC->MoveTo(pt.x,pt.y);
short max=0;
for(i = 0; i < dwSpeechDataNum; i+=5)
if(psWaveData[i]>max)
max=psWaveData[i];
double prob=max/32767;
for (i = 0; i < dwSpeechDataNum; i+=5)
{
//此处的t调整应该根据屏幕坐标来定,时间轴
//最大不能超过窗口的宽
pt.x = i * t/ dwSpeechDataNum+30;
pt.y = nMiddle - psWaveData[i] * nMiddle / max;
pDC->LineTo(pt.x, pt.y);
}
/*
//画语谱图
CString filename;
double data[FFTSIZE];
int bottom=400;
CPoint point;
double maxR=0;
// double minR=15;
int j;
for(i=0;idata[j])
minR=data[j];
}
for(j=0;jMoveTo(point.x,point.y);
int k=255*((maxR-data[j])/(maxR-minR));
if(k>60)
{
k+=130;
if(k>255)
k=255;
}
else
{
k=k+100;
}
CPen pen(0,0,RGB(k,k,k));
CPen *oldpen=pDC->SelectObject(&pen);
pt.y=point.y;
pt.x=point.x+win_size*t/ dwSpeechDataNum;
pDC->LineTo(pt);
pDC->SelectObject(oldpen);
}
}
*/
}
ASSERT_VALID(pDoc);
// TODO: add draw code for native data here
}
/////////////////////////////////////////////////////////////////////////////
// CSpeechView diagnostics
#ifdef _DEBUG
void CSpeechView::AssertValid() const
{
CView::AssertValid();
}
void CSpeechView::Dump(CDumpContext& dc) const
{
CView::Dump(dc);
}
CSpeechDoc* CSpeechView::GetDocument() // non-debug version is inline
{
ASSERT(m_pDocument->IsKindOf(RUNTIME_CLASS(CSpeechDoc)));
return (CSpeechDoc*)m_pDocument;
}
#endif //_DEBUG
/////////////////////////////////////////////////////////////////////////////
// CSpeechView message handlers
void CSpeechView::OnFileOpen()
{
// TODO: Add your command handler code here
//打开语音文件
m_flag=0;
//free(m_pWaveBuffer);
CFileDialog oFileDlg(TRUE, "wav", "*.wav");
if (oFileDlg.DoModal() != IDOK) return;
m_filepath = oFileDlg.GetPathName();
m_name=oFileDlg.GetFileName();
//读取文件
ReadFile(m_filepath);
}
BOOL CSpeechView::ReadFile(const CString & File)
{
CString szTemp;
FILE* fpWave;
char format[9];
delete m_pWaveBuffer;
fpWave = fopen(File, "rb");
if (fpWave == NULL)
{
szTemp.Format("打开文件[%s]失败!", File);
::MessageBox(m_hWnd, szTemp, "ERROR", MB_ICONERROR | MB_OK);
return false;
}
fseek(fpWave, 8, SEEK_SET);
fread(format, sizeof(char), 8, fpWave);
format[8] = 0;
fseek(fpWave, 0, SEEK_END);
m_nWaveLength = ftell(fpWave);
if (m_nWaveLength == 0)
{
m_pWaveBuffer = NULL;
}
else
{
fseek(fpWave, 0, SEEK_SET);
//处理包含文件头的声音文件
if ( strcmp(format, "WAVEfmt ")==0 )
{
m_nWaveLength -= 44;
fseek(fpWave, 44, SEEK_SET);
}
m_pWaveBuffer = static_cast(new char[m_nWaveLength]);
if (m_pWaveBuffer == NULL)
{
szTemp.Format("在读取%s文件之前，分配内存失败", File);
::MessageBox(m_hWnd, szTemp, "ERROR", MB_ICONERROR | MB_OK);
return false;
}
else if (m_nWaveLength != fread(m_pWaveBuffer, sizeof(char), m_nWaveLength, fpWave) )
{
szTemp.Format("读取%s文件时出错", File);
::MessageBox(m_hWnd, szTemp, "ERROR", MB_ICONERROR | MB_OK);
return false;
}
}
fclose(fpWave);
//进行基频提取处理
// fundamental_frequency((short *)m_pWaveBuffer);
m_flag=1;
Invalidate();
return true;
}
void CSpeechView::fundamental_frequency(short *data)
{
// int base_frequency;
int win_size = (int) (Samplerate * WINTIME / 1000);
int step_size = (int) (Samplerate * STEPTIME / 1000);
int i;
int j=m_name.Find('.');
m_name=".\\"+m_name.Left(j);
CreateDirectory((LPCTSTR)m_name,NULL);
Numzheng=(m_nWaveLength/2-win_size)/step_size;
for (i = 0; i >list;
data[i]=atof(list);
}
infile.close();
}
int CSpeechView::isgood(double *power)
{
int i;
int max[FFTSIZE]; /* index of local maximum point */
int max_n = 0; /* number of local maximum point */
int interval[FFTSIZE]; /* distance of two adjacent maximum point */
double average; /* average of the distances */
double variance; /* variance of the distances */
double sum = 0.0; /* temporary sum */
for (i = 9; i <= 1014; i++) {
if ((power[i]>power[i-1]) &&
(power[i]>power[i-2]) &&
(power[i]>power[i-3]) &&
(power[i]>power[i-4]) &&
(power[i]>power[i-5]) &&
(power[i]>power[i-6]) &&
(power[i]>power[i-7]) &&
(power[i]>power[i-8]) &&
(power[i]>power[i-9]) &&
(power[i]>power[i+1]) &&
(power[i]>power[i+2]) &&
(power[i]>power[i+3]) &&
(power[i]>power[i+4]) &&
(power[i]>power[i+5]) &&
(power[i]>power[i+6]) &&
(power[i]>power[i+7]) &&
(power[i]>power[i+8]) &&
(power[i]>power[i+9])) {
max[max_n] = i;
max_n++;
}
}
for (i=0; i < max_n-1; i++) {
interval[i] = max[i+1] - max[i];
sum += interval[i];
}
average = sum / (max_n-1);
sum = 0;
/* standard deviation */
for (i = 0; i < max_n-1; i++)
sum += (interval[i]-average)*(interval[i]-average);
variance = sum / (max_n-1);
if (variance < 25) {
printf("max_n = %d variance = %f ", max_n, variance);
return 1;
} else
return 0;
}