我正在使用opencv（用于对象识别）结合portaudio来播放基于视频输入的声音。基本上，我的目标是以不同的速率播放某个音高/频率的正弦波音调。它有效，但结果非常难以预测。有时audioplayback工作（程序运行缓慢，但它工作），其他时候没有音频播放。简而言之，这就是我的程序所做的事情：
启动网络摄像头Feed - ＆gt;获取网络摄像头图像 - ＆gt;选择图像中的区域 - ＆gt;返回视频Feed - ＆gt; while（框架存在） - ＆gt;跟踪对象位置 - ＆gt;初始化端口音频工具 - ＆gt;根据位置播放声音 - ＆gt;终止Portaudio工具
我似乎无法弄清楚为什么音频播放不一致。你们都有任何提示吗？我一直在阅读，我的想法是这是一个延迟问题，但我真的没有经验。当我在没有opencv的情况下使用portaudio时，不会出现延迟问题，所以我知道它与组合这两者有关。任何帮助表示赞赏。
while (frame)
{
cvCopyImage(frame, drawImg);
// process
track(frame);
// get result
CvRect r;
float confidence;
bool valid;
/* getRoi tells us if the region being tracked on the screen
* is the same region that we chose prior to entering this while loop
*/
getRoi(&r, &confidence, &valid);
// show
cvDrawRect(drawImg, cvPoint(r.x, r.y),
cvPoint(r.x + r.width - 1, r.y + r.height - 1),
valid ? cvScalar(0, 255, 0) : cvScalar(0, 255, 255),
2
);
writeLogo(drawImg,"USC-IRIS");
int xpos = r.x;
int ypos = r.y;
cvShowImage("Tracking", drawImg);
cout << "valid " << valid << endl;
cout << "conf val " << confidence << endl;
cout << "xpos, ypos " << xpos << ", " << ypos << endl;
//If the region on the screen is the region we chose
//then we should play specific sounds
if(valid){
sI->soundWrite(xpos, ypos);
float freq = sI->getFreq();
int amp = sI->getAmp();
float pulse = sI->getPulse();
switch(amp){
case 0:
//printf("Hear sound in both ears.\n");
data.targetBalance = .5;
break;
case 1:
//printf("Hear sound in left ear.\n");
data.targetBalance = 0;
break;
case 2:
//printf("Hear sound in right ear.\n");
data.targetBalance = 1;
break;
default:
//printf("Incorrect value for amp (left/right sound indicator)");
data.targetBalance = .5;
break;
}
err = Pa_Initialize(); //scan for available devices i.e. audio jack, headphones
if(err != paNoError) {
printf("init\n");
goto error;
}
//open the sound stream for processing
err = Pa_OpenDefaultStream( &stream, 0, 2, paFloat32, SAMPLE_RATE,
256, patestCallback, &data ); //open the sound stream for processing
if( err != paNoError ) {
printf("open\n");
goto error;
}
//start the stream (i.e. play sound) if no errors
err = Pa_StartStream(stream);
if(err != paNoError) {
printf("start\n");
goto error;
}
//check which ear(s) the sound should be played to
//hold that tone for a certain amount of time (pulse*200 millisec)
Pa_Sleep(pulse*200);
cout << "pulse: " << pulse << endl << "freq: " << freq << endl;
cout << "amp: " << amp << endl;
//stop the stream (i.e. stop playing sound)
err = Pa_StopStream(stream);
if(err != paNoError) {
printf("stop\n");
goto error;
}
err = Pa_CloseStream( stream );
if( err != paNoError ) {
printf("close\n");
goto error;
}
err = Pa_Terminate();
if( err != paNoError ) {
printf("term\n");
goto error;
}
}
int key = cvWaitKey(1);
// write
if (output_txt)
fprintf(output_txt, "%d %d %d %d\n", r.x, r.y, r.width, r.height);
if (output_avi)
cvWriteFrame(output_avi, drawImg);
// next
if (key == 'q'||key=='Q')
break;
frame = cvQueryFrame(capture);
}
答案 0 :(得分：2)
似乎不一致的音频播放是由于我上面的问题中未显示的另一部分代码造成的。这个错误的代码如下。我相信错误与此函数中的第一个if语句和最后一个forloop有关。我认为变量framesToCalc没有正确计算。因此，第一个for循环没有将任何数据放入outputBuffer / out变量。然后，最后我将剩余的未使用的缓冲区空间归零。因此，由于缓冲区归零而没有声音。我的解决方案是删除第一个if else和最后一个forloop。另外，我做了第一个for循环，从i = 0到framesPerBuffer。现在它完美无缺。
static int patestCallback(const void *inputBuffer, void *outputBuffer, unsigned long framesPerBuffer, const PaStreamCallbackTimeInfo *timeInfo, PaStreamCallbackFlags statusFlags, void *userData){
paTestData *data = (paTestData*)userData;
SAMPLE_t *out = (SAMPLE_t *)outputBuffer;
int i;
int framesToCalc;
int finished = 0;
(void) inputBuffer;
int left_phase = data->left_phase;
int right_phase = data->right_phase;
if( data->framesToGo < framesPerBuffer )
{
framesToCalc = data->framesToGo;
data->framesToGo = 0;
finished = 1;
}
else
{
framesToCalc = framesPerBuffer;
data->framesToGo -= framesPerBuffer;
}
for( i=0; i<framesToCalc; i++ )
{
if( data->currentBalance < data->targetBalance )
{
data->currentBalance += BALANCE_DELTA;
}
else if( data->currentBalance > data->targetBalance )
{
data->currentBalance -= BALANCE_DELTA;
}
left_phase += (LEFT_FREQ / SAMPLE_RATE);
right_phase += (RIGHT_FREQ / SAMPLE_RATE);
if( fabs(data->currentBalance - .5) < .001){
//left_phase += (double)(LEFT_FREQ / SAMPLE_RATE);
if( left_phase > 1.0) left_phase -= 1.0;
*out++ = DOUBLE_TO_SAMPLE( AMPLITUDE * sin( (left_phase * M_PI * 2. )));
//right_phase += (double)(RIGHT_FREQ / SAMPLE_RATE);
if( right_phase > 1.0) right_phase -= 1.0;
*out++ = DOUBLE_TO_SAMPLE( AMPLITUDE * sin( (right_phase * M_PI * 2. )));
}else{
//left_phase += (double)(LEFT_FREQ / SAMPLE_RATE);
if( left_phase > 1.0) left_phase -= 1.0;
*out++ = DOUBLE_TO_SAMPLE( AMPLITUDE * sin( (left_phase * M_PI * 2. ))*(1.0 - data->currentBalance));
//right_phase += (double)(RIGHT_FREQ / SAMPLE_RATE);
if( right_phase > 1.0) right_phase -= 1.0;
*out++ = DOUBLE_TO_SAMPLE( AMPLITUDE * sin( (right_phase * M_PI * 2. ))*data->currentBalance);
}
}
// zero remainder of final buffer
for( ; i<(int)framesPerBuffer; i++ )
{
*out++ = SAMPLE_ZERO; //left
*out++ = SAMPLE_ZERO; //right
}
data->left_phase = left_phase;
data->right_phase = right_phase;
return finished;
}
答案 1 :(得分：1)
为什么要在帧循环中初始化和终止PortAudio？我建议在程序开始时初始化一次，并在结束时终止一次。