2

I'm trying to make 3D sound with the SDL_Mixer lib by making a convolution of the audio stream with a HRTF (I'm using the MIT HRTF database respectively the mit_hrtf_lib based on the compact dataset).

I register an effect with the callback function on MIX_CHANNEL_POST. Then, in the callback function of my effect, I make the convolution of the audio stream with the HRTF.

But I got no 3D sound. The audiostream I want to play sounds like it's double or even triple overlayed by itself and has no 3D effect.

Here's what I'm doing:

First load wav, load HRTF and start audio:

if( Mix_OpenAudio( 44100, MIX_DEFAULT_FORMAT, 2, 16384 ) == -1 ) {
    fprintf(stderr, "Couldn't open SDL_mixer", SDL_GetError()); return 0; }

myChunk = Mix_LoadWAV( "bombsBitrate176400Hz.wav" ); //load wav

if( myChunk == NULL) { 
    fprintf(stderr, "Couldn't load wav file", Mix_GetError()); return 0; 
}

myLoadHRTF = new LoadHRTF;
myHRTFs = myLoadHRTF->getHRTF(90,0,44100);//load hrtf => works

Mix_RegisterEffect(MIX_CHANNEL_POST, HRTFEffect, NULL, &myHRTFs);

while(1){
    if (canPlay){//virtual source is set and azimuth is calculated => I can start audio
        playAudio();
        canPlay = false;
    }
}

playAudio():

void PlaySDLAudio::playAudio(){
int channel;
channel = Mix_PlayChannel(-1,myChunk,0); //play audio once}

The callback for my registered effect:

void HRTFEffect(int chan, void* stream, int len, void* udata){
Mix_Pause(-1);

short* Y;
short* X = (short*) stream;
LeftRightHRTF* myHRTFs = (LeftRightHRTF*) udata;
short* HL;
short* HR;

int m = len/2;//size= len/2 beause we use 2 bytes values (16 bit audio format => 2 byte), len is the size of stream in bytes
int n = myHRTFs->nTaps;

HL = new short[n+m];
HR = new short[n+m];
Y = new short[n+m-1];

for (int i = m/2; i < (m+n)/2; i++){// fill with 0 
    X[i*2] = 0;
    X[i*2+1] = 0;
}
for( int i = 0 ; i < n; i++ ){
    HL[i] = *myHRTFs->leftHRTF;
    myHRTFs->leftHRTF++;
    HR[i] = *myHRTFs->rightHRTF;
    myHRTFs->rightHRTF++;
}
for ( int i = n; i < n+m; i++ ){//fill rest with 0
    HL[i] = 0;
    HR[i] = 0;
}

for ( int i = 0; i < (n+m)/2-1; i++ ) { // convolution
    Y[i*2] = 0;
    Y[i*2+1] = 0;
    for ( int j = 0; j < m/2; j++ ) {
        if (i-j>=0){
            Y[i*2] = Y[i*2] + X[j*2] * HL[i-j];//stream for left speaker
            Y[i*2+1] = Y[i*2+1] + X[j*2+1] * HR[i-j];//stream for right speaker
        }
    }
}

stream = Y; //set result of convolution to stream

Mix_Resume(-1);

delete[] Y;
Y = 0;
delete[] HL;
HL = 0;
delete[] HR;
HR = 0;}

I guess that there's a problem by just saying stream = Y and because my convolution result Y is longer than the original stream (m+n-1, original stream m).

genpfault
  • 51,148
  • 11
  • 85
  • 139
user1800143
  • 21
  • 2
  • Why do this manually if you have [OpenAL](http://repo.or.cz/w/openal-soft.git/blob/HEAD:/hrtf.txt)? – genpfault Nov 05 '12 at 18:40
  • I didn't know that OpenAL has this functionality... And I never used OpenAL. Is there a quick tutorial to start with OpenAL and the usage with the build-in HRTFs? – user1800143 Nov 05 '12 at 19:00
  • If the 3D sound is running, I need to implement a Crosstalk Canceller. Is there a easy way to do this with OpenAL too respectively work with the audiostream to implement a crosstalk canceller? Nevertheless, I wonder what's the problem in my sourcecode, because I've spend some time in searching the mistake... It would be great if someone has an idea. – user1800143 Nov 05 '12 at 20:17

0 Answers0