SpectralPeakProcessor

java.lang.Object
- be.tarsos.dsp.SpectralPeakProcessor

All Implemented Interfaces:: AudioProcessor

public class SpectralPeakProcessor
extends java.lang.Object
implements AudioProcessor

This class implements a spectral peak follower as described in Sethares et al. 2009 - Spectral Tools for Dynamic Tonality and Audio Morphing - section "Analysis-Resynthessis". It calculates a noise floor and picks spectral peaks rising above a calculated noise floor with a certain factor. The noise floor is determined using a simple median filter.

Parts of the code is modified from the code accompanying "Spectral Tools for Dynamic Tonality and Audio Morphing".

To get the spectral peaks from an audio frame, call getPeakList

 
AudioDispatcher dispatcher = new AudioDispatcher(stream, fftsize, overlap);
dispatcher.addAudioProcessor(spectralPeakFollower);
dispatcher.addAudioProcessor(new AudioProcessor() {

  public void processingFinished() {
  }
  
  public boolean process(AudioEvent audioEvent) {
    float[] noiseFloor = SpectralPeakProcessor.calculateNoiseFloor(spectralPeakFollower.getMagnitudes(), medianFilterLength, noiseFloorFactor);
        List localMaxima = SpectralPeakProcessor.findLocalMaxima(spectralPeakFollower.getMagnitudes(), noiseFloor);
        List<> list = SpectralPeakProcessor.findPeaks(spectralPeakFollower.getMagnitudes(), spectralPeakFollower.getFrequencyEstimates(), localMaxima, numberOfPeaks);
    // do something with the list...
    return true;
  }
});
dispatcher.run();

Author:: Joren Six, William A. Sethares, Andrew J. Milne, Stefan Tiedje, Anthony Prechtl, James Plamondon

Nested Class Summary

Nested Classes
Modifier and Type Class and Description

static class SpectralPeakProcessor.SpectralPeak

Nested Classes
Modifier and Type	Class and Description
`static class`	`SpectralPeakProcessor.SpectralPeak`

Constructor Summary

Constructors
Constructor and Description

SpectralPeakProcessor(int bufferSize, int overlap, int sampleRate)

Constructors
Constructor and Description
`SpectralPeakProcessor(int bufferSize, int overlap, int sampleRate)`

Method Summary

All Methods Static Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`static float[]`	`calculateNoiseFloor(float[] magnitudes, int medianFilterLength, float noiseFloorFactor)` Calculate a noise floor for an array of magnitudes.
`static java.util.List<java.lang.Integer>`	`findLocalMaxima(float[] magnitudes, float[] noisefloor)` Finds the local magintude maxima and stores them in the given list.
`static java.util.List<SpectralPeakProcessor.SpectralPeak>`	`findPeaks(float[] magnitudes, float[] frequencyEstimates, java.util.List<java.lang.Integer> localMaximaIndexes, int numberOfPeaks, int minDistanceInCents)`
`float[]`	`getFrequencyEstimates()`
`float[]`	`getMagnitudes()`
`static float`	`median(double[] arr)`
`static double`	`median(float[] m)`
`static float`	`percentile(double[] arr, double p)` Returns the p-th percentile of values in an array.
`boolean`	`process(AudioEvent audioEvent)` Process the audio event.
`void`	`processingFinished()` Notify the AudioProcessor that no more data is available and processing has finished.

Methods inherited from class java.lang.Object
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

- Constructor Detail
  - SpectralPeakProcessor
```
public SpectralPeakProcessor(int bufferSize,
                             int overlap,
                             int sampleRate)
```
- Method Detail
  - process
```
public boolean process(AudioEvent audioEvent)
```
    Description copied from interface: AudioProcessor
    
    Process the audio event. Do the actual signal processing on an (optionally) overlapping buffer.
    
    Specified by:
    
    process in interface AudioProcessor
    
    Parameters:
    
    audioEvent - The audio event that contains audio data.
    
    Returns:
    
    False if the chain needs to stop here, true otherwise. This can be used to implement e.g. a silence detector.
  - processingFinished
```
public void processingFinished()
```
    Description copied from interface: AudioProcessor
    
    Notify the AudioProcessor that no more data is available and processing has finished. Can be used to deallocate resources or cleanup.
    
    Specified by:
    
    processingFinished in interface AudioProcessor
  - getMagnitudes
```
public float[] getMagnitudes()
```
    Returns:
    
    the magnitudes.
  - getFrequencyEstimates
```
public float[] getFrequencyEstimates()
```
    Returns:
    
    the precise frequency for each bin.
  - calculateNoiseFloor
```
public static float[] calculateNoiseFloor(float[] magnitudes,
                                          int medianFilterLength,
                                          float noiseFloorFactor)
```
    Calculate a noise floor for an array of magnitudes.
    
    Parameters:
    
    magnitudes - The magnitudes of the current frame.
    
    medianFilterLength - The length of the median filter used to determine the noise floor.
    
    noiseFloorFactor - The noise floor is multiplied with this factor to determine if the information is either noise or an interesting spectral peak.
    
    Returns:
    
    a float array representing the noise floor.
  - findLocalMaxima
```
public static java.util.List<java.lang.Integer> findLocalMaxima(float[] magnitudes,
                                                                float[] noisefloor)
```
    Finds the local magintude maxima and stores them in the given list.
    
    Parameters:
    
    magnitudes - The magnitudes.
    
    noisefloor - The noise floor.
    
    Returns:
    
    a list of local maxima.
  - findPeaks
```
public static java.util.List<SpectralPeakProcessor.SpectralPeak> findPeaks(float[] magnitudes,
                                                                           float[] frequencyEstimates,
                                                                           java.util.List<java.lang.Integer> localMaximaIndexes,
                                                                           int numberOfPeaks,
                                                                           int minDistanceInCents)
```
    Parameters:
    
    magnitudes - the magnitudes..
    
    frequencyEstimates - The frequency estimates for each bin.
    
    localMaximaIndexes - The indexes of the local maxima.
    
    numberOfPeaks - The requested number of peaks.
    
    minDistanceInCents - The minimum distance in cents between the peaks
    
    Returns:
    
    A list with spectral peaks.
  - median
```
public static final float median(double[] arr)
```
  - percentile
```
public static final float percentile(double[] arr,
                                     double p)
```
    Returns the p-th percentile of values in an array. You can use this function to establish a threshold of acceptance. For example, you can decide to examine candidates who score above the 90th percentile (0.9). The elements of the input array are modified (sorted) by this method.
    
    Parameters:
    
    arr - An array of sample data values that define relative standing. The contents of the input array are sorted by this method.
    
    p - The percentile value in the range 0..1, inclusive.
    
    Returns:
    
    The p-th percentile of values in an array. If p is not a multiple of 1/(n - 1), this method interpolates to determine the value at the p-th percentile.
  - median
```
public static double median(float[] m)
```

Class SpectralPeakProcessor

Nested Class Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Constructor Detail

SpectralPeakProcessor

Method Detail

process

processingFinished

getMagnitudes

getFrequencyEstimates

calculateNoiseFloor

findLocalMaxima

findPeaks

median

percentile

median