基于重采样,相位声码器及BP神经网络基音分类的变声器,数学,UI及信号处理算法基于PainterEngine开发.zip

#include "SoundLab_Main.h" #define SOUNDLAB_DEFAULT_FRAME_COLOR 255,96,192,255 static px_char SoundLab_ExportfilePath[MAX_PATH]; DWORD WINAPI SoundLab_Main_ExportFeatures(LPVOID ptr) { px_int i; SoundLab_Main *pMain=(SoundLab_Main *)ptr; SOUNDLAB_FEATURE_DATA Data; px_complex Frame[SOUNDLAB_WINDOW_WIDTH]; px_double Framed[SOUNDLAB_WINDOW_WIDTH]; px_double PitchShift; FILE *pf; px_int randSample; px_int tryCount=128; if (pMain->sourcePCMSize<SOUNDLAB_WINDOW_WIDTH*2*2) { goto _ERROR; } pMain->schedule=0; pMain->bDone=PX_FALSE; PX_memset(&Data,0,sizeof(Data)); PX_memcpy(&Data.magic,"sinc",4); PX_Object_FilterEditorMapData(pMain->FilterEditor,Data.filter,200); PX_Object_FilterEditorMapData(pMain->FixEditor,Data.fix,200); PitchShift=(PX_Object_SliderBarGetValue(pMain->SliderBar_PitchShift)-50.0)/100; if(PitchShift==0) { PitchShift=1; } else if (PitchShift>0) { PitchShift=1+PitchShift/5*(2.75)*10; } else { PitchShift=-PitchShift/5*(2.75)*10; PitchShift=1/(1+PitchShift); } Data.pitchshift=PitchShift; //MFCC for (i=0;i<SOUNDLAB_FEATURE_DATA_MFCC_SIZE;i++) { px_int j; px_int sum,pitchE; px_short *PCM; tryCount=128; while (PX_TRUE) { randSample=(px_int)PX_randRange(0,pMain->sourcePCMSize-SOUNDLAB_WINDOW_WIDTH*2*2); randSample=(randSample/4)*4; PCM=(px_short *)(pMain->sourcePCM+randSample); if (tryCount==0&&pMain->schedule==0) { PX_strset(pMain->lastError,"Invalid Data."); goto _ERROR; } sum=0; for (j=0;j<SOUNDLAB_WINDOW_WIDTH;j++) { Frame[j].re=(px_double)PCM[j*2]; Frame[j].im=0; Framed[j]=(px_double)PCM[j*2]; sum+=PX_ABS(PCM[j*2]); } sum/=SOUNDLAB_WINDOW_WIDTH; if (sum<pMain->threshold) { tryCount--; continue; } pitchE=PX_PitchEstimation(Frame,SOUNDLAB_WINDOW_WIDTH,44100); if (pitchE<60&&pitchE>800) { tryCount--; continue; } PX_MFCCParse(&pMain->mfcc,Framed,&Data.mfcc[i]); pMain->schedule=i*100.0/SOUNDLAB_FEATURE_DATA_MFCC_SIZE; break; } } pf=fopen(SoundLab_ExportfilePath,"wb"); if (!pf) { PX_strset(pMain->lastError,"Could not write file."); goto _ERROR; } fwrite(&Data,1,sizeof(Data),pf); fclose(pf); pMain->lastError[0]=0; PX_MessageBoxAlertOk(&pMain->messagebox,(const px_char *)L"Export Succeeded."); pMain->bDone=PX_TRUE; return 0; _ERROR: PX_MessageBoxAlertOk(&pMain->messagebox,(const px_char *)L"Export Failed"); pMain->bDone=PX_TRUE; return 0; } static px_char TuningRuntime[1024*1024*8]; DWORD WINAPI SoundLab_Main_ApplyTuning(LPVOID ptr) { px_double zero0=0; px_int offset,i; px_double pitchshift,timeshift; SoundLab_Main *pMain=(SoundLab_Main *)ptr; PX_Tuning Tuning1; PX_Tuning Tuning2; px_short *pDataU16; px_double Frame1[SOUNDLAB_WINDOW_WIDTH*8]; px_double Frame2[SOUNDLAB_WINDOW_WIDTH*8]; px_double Frame[SOUNDLAB_WINDOW_WIDTH*8]; px_double filter[SOUNDLAB_WINDOW_WIDTH]; px_double fix[SOUNDLAB_WINDOW_WIDTH]; px_int FrameSize; px_memory solveData; offset=0; pMain->schedule=0; if (pMain->sourcePCMSize) { px_memorypool mp; mp=MP_Create(TuningRuntime,sizeof(TuningRuntime)); PX_MemoryInit(&pMain->Ins->runtime.mp_game,&solveData); pitchshift=(PX_Object_SliderBarGetValue(pMain->SliderBar_PitchShift)-50.0)/100; if(pitchshift==0) { pitchshift=1; } else if (pitchshift>0) { pitchshift=1+pitchshift/5*(2.75)*10; } else { pitchshift=-pitchshift/5*(2.75)*10; pitchshift=1/(1+pitchshift); } timeshift=(PX_Object_SliderBarGetValue(pMain->SliderBar_TimeScale)-50.0)/100; if(timeshift==0) { timeshift=1; } else if (timeshift>0) { timeshift=1+timeshift/5*(2.75)*10; } else { timeshift=-timeshift/5*(2.75)*10; timeshift=1/(1+timeshift); } //filter for (i=0;i<PX_COUNTOF(filter);i++) { filter[i]=0; fix[i]=0; } PX_Object_FilterEditorMapData(pMain->FilterEditor,filter,200); PX_Object_FilterEditorMapData(pMain->FixEditor,fix,200); PX_TuningInitialize(&mp,&Tuning1,pitchshift,timeshift,PX_NULL,filter,PX_NULL,PX_TUNING_WINDOW_SIZE_1024); PX_TuningInitialize(&mp,&Tuning2,pitchshift,timeshift,PX_NULL,filter,PX_NULL,PX_TUNING_WINDOW_SIZE_1024); for (i=0;i<512*3;i++) { PX_MemoryCat(&solveData,&zero0,sizeof(px_double)); } while (offset<pMain->sourcePCMSize-SOUNDLAB_WINDOW_WIDTH*2*2*2) { pDataU16=(px_short *)(pMain->sourcePCM+offset); pMain->schedule=offset*1.0/pMain->sourcePCMSize*100; //channel 1 for (i=0;i<SOUNDLAB_WINDOW_WIDTH;i++) { Frame[i]=pDataU16[i*2]/32768.0; } FrameSize=PX_TuningFilter(&Tuning1,Frame,SOUNDLAB_WINDOW_WIDTH,Frame1); //channel 2 for (i=0;i<SOUNDLAB_WINDOW_WIDTH;i++) { Frame[i]=pDataU16[i*2+1]/32768.0; } PX_TuningFilter(&Tuning2,Frame,SOUNDLAB_WINDOW_WIDTH,Frame2); for (i=0;i<FrameSize;i++) { if (PX_ABS(Frame1[i])<1) { Frame1[i]*=32768.0; } else { if (Frame1[i]>0) { Frame1[i]=32768.0; } else { Frame1[i]=-32767.0; } } if (PX_ABS(Frame2[i])<1) { Frame2[i]*=32768.0; } else { if (Frame2[i]>0) { Frame2[i]=32768; } else { Frame2[i]=-32767; } } } //blend for (i=0;i<FrameSize;i++) { px_short val; val=(px_short)Frame1[i]; PX_MemoryCat(&solveData,&val,sizeof(px_short)); val=(px_short)Frame2[i]; PX_MemoryCat(&solveData,&val,sizeof(px_short)); } offset+=SOUNDLAB_WINDOW_WIDTH*2*2; } PX_SoundStaticDataFree(&pMain->SoundData); pMain->SoundData.buffer=solveData.buffer; pMain->SoundData.mp=solveData.mp; pMain->SoundData.size=solveData.usedsize; pMain->SoundData.channel=PX_SOUND_CHANNEL_DOUBLE; } pMain->bDone=PX_TRUE; return 0; } DWORD WINAPI SoundLab_Main_InitializeSoundData(LPVOID ptr) { px_int i,w,x,y; px_int max=1; px_int min=-1; px_double pmax=0; px_complex frame[SOUNDLAB_WINDOW_WIDTH]; px_double window[SOUNDLAB_WINDOW_WIDTH]; SoundLab_Main *pMain=(SoundLab_Main *)ptr; if (pMain->SoundData.size) { px_short *PCM16=(px_short *)pMain->SoundData.buffer; for (i=0;i<pMain->SoundData.size/2;i+=2) { if (PCM16[i]>max) { max=PCM16[i]; } if (PCM16[i]<min) { min=PCM16[i]; } } pMain->wave_min=min; pMain->wave_max=max; //Build Spectrum if (pMain->tex_Spectrum.MP) { PX_TextureFree(&pMain->tex_Spectrum); } w=pMain->SoundData.size/2/2/SOUNDLAB_WINDOW_WIDTH*4;//2 bytes 2 channels 75% overlap pMain->process_memories=w*SOUNDLAB_WINDOW_WIDTH/2*4; pMain->lastError[0]=0; pMain->schedule=0; if(!PX_TextureCreate(&pMain->Ins->runtime.mp_game,&pMain->tex_Spectrum,w,SOUNDLAB_WINDOW_WIDTH/2)) { PX_strset(pMain->lastError,"Out Of Memory."); pMain->bDone=PX_TRUE; return 0; } PX_WindowFunction_hamming(window,SOUNDLAB_WINDOW_WIDTH); for (x=0;x<w;x++) { //Data Sampling for (i=0;i<SOUNDLAB_WINDOW_WIDTH;i++) { frame[i].re=PCM16[x*2*SOUNDLAB_WINDOW_WIDTH/4+i*2];//2bytes frame[i].im=0; frame[i].re*=window[i]; } PX_FFT(frame,frame,SOUNDLAB_WINDOW_WIDTH); for (i=0;i<SOUNDLAB_WINDOW_WIDTH;i++) { if (frame[i].re*frame[i].re+frame[i].im*frame[i].im>pmax) { pmax=frame[i].re*frame[i].re+frame[i].im*frame[i].im; } } pMain->schedule=1.0*x/w/2*100; } pMain->MaxPower=PX_sqrtd(pmax); pmax=PX_log(pmax+1); for (x=0;x<w;x++) { px_double unit; px_color pixel; //Data Sampling for (i=0;i<SOUNDLAB_WINDOW_WIDTH;i++) { frame[i].re=PCM16[x*2*SOUNDLAB_WINDOW_WIDTH/4+i*2];//2bytes frame[i].im=0; frame[i].re*=window[i]; } //FFT PX_FFT(frame,frame,SOUNDLAB_WINDOW_WIDTH); for (y=0;y<SOUNDLAB_WINDOW_WIDTH/2;y++) { unit=PX_log(frame[y].re*frame[y].re+frame[y].im*frame[y].im+1)/pmax; if (unit<pMain->PowerThreshold) { unit=0; } pixel._argb.a=(px_uchar)(64+191*unit); pixel._argb.g=(px_uchar)(255*unit); pixel._argb.r=(px_ucha