/*
* _______ _____ _____ _____
* |__ __| | __ \ / ____| __ \
* | | __ _ _ __ ___ ___ ___| | | | (___ | |__) |
* | |/ _` | '__/ __|/ _ \/ __| | | |\___ \| ___/
* | | (_| | | \__ \ (_) \__ \ |__| |____) | |
* |_|\__,_|_| |___/\___/|___/_____/|_____/|_|
*
* -------------------------------------------------------------
*
* TarsosDSP is developed by Joren Six at IPEM, University Ghent
*
* -------------------------------------------------------------
*
* Info: http://0110.be/tag/TarsosDSP
* Github: https://github.com/JorenSix/TarsosDSP
* Releases: http://0110.be/releases/TarsosDSP/
*
* TarsosDSP includes modified source code by various authors,
* for credits and info, see README.
*
*/
package be.tarsos.dsp;
import java.nio.ByteOrder;
import java.util.List;
import java.util.concurrent.CopyOnWriteArrayList;
import java.util.logging.Logger;
import be.tarsos.dsp.io.TarsosDSPAudioFormat;
/**
* This class plays a file and sends float arrays to registered AudioProcessor
* implementors. This class can be used to feed FFT's, pitch detectors, audio players, ...
* Using a (blocking) audio player it is even possible to synchronize execution of
* AudioProcessors and sound. This behavior can be used for visualization.
* @author Joren Six
*/
public class AudioGenerator implements Runnable {
/**
* Log messages.
*/
private static final Logger LOG = Logger.getLogger(AudioGenerator.class.getName());
/**
* This buffer is reused again and again to store audio data using the float
* data type.
*/
private float[] audioFloatBuffer;
/**
* A list of registered audio processors. The audio processors are
* responsible for actually doing the digital signal processing
*/
private final List<AudioProcessor> audioProcessors;
private final TarsosDSPAudioFormat format;
/**
* The floatOverlap: the number of elements that are copied in the buffer
* from the previous buffer. Overlap should be smaller (strict) than the
* buffer size and can be zero. Defined in number of samples.
*/
private int floatOverlap, floatStepSize;
private int samplesProcessed;
/**
* The audio event that is send through the processing chain.
*/
private AudioEvent audioEvent;
/**
* If true the dispatcher stops dispatching audio.
*/
private boolean stopped;
/**
* Create a new generator.
* @param audioBufferSize
* The size of the buffer defines how much samples are processed
* in one step. Common values are 1024,2048.
* @param bufferOverlap
* How much consecutive buffers overlap (in samples). Half of the
* AudioBufferSize is common (512, 1024) for an FFT.
*/
public AudioGenerator(final int audioBufferSize, final int bufferOverlap){
this(audioBufferSize,bufferOverlap,44100);
}
public AudioGenerator(final int audioBufferSize, final int bufferOverlap,final int samplerate){
audioProcessors = new CopyOnWriteArrayList<AudioProcessor>();
format = getTargetAudioFormat(samplerate);
setStepSizeAndOverlap(audioBufferSize, bufferOverlap);
audioEvent = new AudioEvent(format);
audioEvent.setFloatBuffer(audioFloatBuffer);
stopped = false;
samplesProcessed = 0;
}
/**
* Constructs the target audio format. The audio format is one channel
* signed PCM of a given sample rate.
*
* @param targetSampleRate
* The sample rate to convert to.
* @return The audio format after conversion.
*/
private TarsosDSPAudioFormat getTargetAudioFormat(int targetSampleRate) {
TarsosDSPAudioFormat audioFormat = new TarsosDSPAudioFormat(TarsosDSPAudioFormat.Encoding.PCM_SIGNED,
targetSampleRate,
2 * 8,
1,
2 * 1,
targetSampleRate,
ByteOrder.BIG_ENDIAN.equals(ByteOrder.nativeOrder()));
return audioFormat;
}
/**
* Set a new step size and overlap size. Both in number of samples. Watch
* out with this method: it should be called after a batch of samples is
* processed, not during.
*
* @param audioBufferSize
* The size of the buffer defines how much samples are processed
* in one step. Common values are 1024,2048.
* @param bufferOverlap
* How much consecutive buffers overlap (in samples). Half of the
* AudioBufferSize is common (512, 1024) for an FFT.
*/
public void setStepSizeAndOverlap(final int audioBufferSize, final int bufferOverlap){
audioFloatBuffer = new float[audioBufferSize];
floatOverlap = bufferOverlap;
floatStepSize = audioFloatBuffer.length - floatOverlap;
}
/**
* Adds an AudioProcessor to the chain of processors.
*
* @param audioProcessor
* The AudioProcessor to add.
*/
public void addAudioProcessor(final AudioProcessor audioProcessor) {
audioProcessors.add(audioProcessor);
LOG.fine("Added an audioprocessor to the list of processors: " + audioProcessor.toString());
}
/**
* Removes an AudioProcessor to the chain of processors and calls processingFinished.
*
* @param audioProcessor
* The AudioProcessor to remove.
*/
public void removeAudioProcessor(final AudioProcessor audioProcessor) {
audioProcessors.remove(audioProcessor);
audioProcessor.processingFinished();
LOG.fine("Remove an audioprocessor to the list of processors: " + audioProcessor.toString());
}
public void run() {
//Read the first (and in some cases last) audio block.
generateNextAudioBlock();
// As long as the stream has not ended
while (!stopped) {
//Makes sure the right buffers are processed, they can be changed by audio processors.
for (final AudioProcessor processor : audioProcessors) {
if(!processor.process(audioEvent)){
//skip to the next audio processors if false is returned.
break;
}
}
if(!stopped){
audioEvent.setBytesProcessed(samplesProcessed * format.getFrameSize());
// Read, convert and process consecutive overlapping buffers.
// Slide the buffer.
generateNextAudioBlock();
}
}
// Notify all processors that no more data is available.
// when stop() is called processingFinished is called explicitly, no need to do this again.
// The explicit call is to prevent timing issues.
if(!stopped){
stop();
}
}
/**
* Stops dispatching audio data.
*/
public void stop() {
stopped = true;
for (final AudioProcessor processor : audioProcessors) {
processor.processingFinished();
}
}
/**
* Reads the next audio block. It tries to read the number of bytes defined
* by the audio buffer size minus the overlap. If the expected number of
* bytes could not be read either the end of the stream is reached or
* something went wrong.
*
* The behavior for the first and last buffer is defined by their corresponding the zero pad settings. The method also handles the case if
* the first buffer is also the last.
*
*/
private void generateNextAudioBlock() {
assert floatOverlap < audioFloatBuffer.length;
//Shift the audio information using array copy since it is probably faster than manually shifting it.
// No need to do this on the first buffer
if(audioFloatBuffer.length == floatOverlap + floatStepSize ){
System.arraycopy(audioFloatBuffer, floatStepSize, audioFloatBuffer,0 ,floatOverlap);
}
samplesProcessed += floatStepSize;
}
public void resetTime(){
samplesProcessed=0;
}
public TarsosDSPAudioFormat getFormat(){
return format;
}
/**
*
* @return The currently processed number of seconds.
*/
public float secondsProcessed(){
return samplesProcessed / format.getSampleRate() / format.getChannels() ;
}
}