org.openimaj.audio.analysis

Class FourierTransform

java.lang.Object

org.openimaj.audio.Audio

org.openimaj.audio.AudioStream

org.openimaj.audio.processor.AudioProcessor

org.openimaj.audio.analysis.FourierTransform

All Implemented Interfaces:

Iterable<SampleChunk>

public class FourierTransform
extends AudioProcessor

Perform an FFT on an audio signal. An FFT will be calculated for every channel in the audio separately. Use getLastFFT() to get the last generated frequency domain calculation.

The class also includes an inverse transform function that takes a frequency domain array (such as that delivered by getLastFFT()) and returns a SampleChunk. The format of the output sample chunk is determined by the given audio format.

Author:

David Dupplaw (dpd@ecs.soton.ac.uk)

Created:

28 Oct 2011

Field Summary

Fields inherited from class org.openimaj.audio.Audio

format

Constructor Summary

Constructors

Constructor and Description
FourierTransform() Default constructor for ad-hoc processing.
FourierTransform(AudioStream as) Constructor for chaining.

Method Summary

Modifier and Type	Method and Description
float[][]	getLastFFT() Get the last processed FFT frequency data.
float[][]	getMagnitudes() Returns the magnitudes of the last FFT data.
float[][]	getNormalisedMagnitudes(float scalar) Scales the real and imaginary parts by the scalar prior to calculating the (square) magnitude for normalising the outputs.
float[][]	getPowerMagnitudes() Returns the power magnitudes of the last FFT data.
float[][]	getReals() Returns just the real numbers from the last FFT.
float	getScalingFactor() Get the scaling factor in use.
static SampleChunk	inverseTransform(AudioFormat format, float[][] transformedData) Given some transformed audio data, will convert it back into a sample chunk.
boolean	isPadToNextPowerOf2() Returns whether the input will be padded to be the length of the next higher power of 2.
SampleBuffer	process(SampleBuffer sb) Process the given sample buffer
SampleChunk	process(SampleChunk sample) Function that takes a sample chunk and processes the chunk.
void	setPadToNextPowerOf2(boolean padToNextPowerOf2) Set whether to pad the input to the next power of 2.
void	setScalingFactor(float scalingFactor) Set the scaling factor to use.

Methods inherited from class org.openimaj.audio.processor.AudioProcessor

getLength, getUnderlyingStream, nextSampleChunk, process, processingComplete, reset, setUnderlyingStream

Methods inherited from class org.openimaj.audio.AudioStream

iterator, seek

Methods inherited from class org.openimaj.audio.Audio

getFormat, setFormat

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface java.lang.Iterable

forEach, spliterator

Constructor Detail

FourierTransform

public FourierTransform()

Default constructor for ad-hoc processing.

FourierTransform

public FourierTransform(AudioStream as)

Constructor for chaining.

Parameters:

as - The stream to chain to

Method Detail

process

public SampleChunk process(SampleChunk sample)

Function that takes a sample chunk and processes the chunk. It should also return a sample chunk containing the processed data. If wished, the chunk may be side-affected and the input chunk returned. It should not be assumed that the input chunk will be side-affected, but it must be noted that it is possible that it could be. This process function may also return null. If null is returned it means that the rest of the audio stream is not required to be processed by this processing function. Whether the rest of the sample chunks are copied or ignored is up to the caller.

Specified by:

process in class AudioProcessor

Parameters:

sample - The sample chunk to process.

Returns:

A sample chunk containing processed data.

See Also:

AudioProcessor.process(org.openimaj.audio.SampleChunk)

process

public SampleBuffer process(SampleBuffer sb)

Process the given sample buffer

Parameters:

sb - The sample buffer

Returns:

The sample buffer

inverseTransform

public static SampleChunk inverseTransform(AudioFormat format,
                                           float[][] transformedData)

Given some transformed audio data, will convert it back into a sample chunk. The number of channels given audio format must match the data that is provided in the transformedData array.

Parameters:

format - The required format for the output

transformedData - The frequency domain data

Returns:

A SampleChunk

getLastFFT

public float[][] getLastFFT()

Get the last processed FFT frequency data.

Returns:

The fft of the last processed window

getMagnitudes

public float[][] getMagnitudes()

Returns the magnitudes of the last FFT data. The length of the returned array of magnitudes will be half the length of the FFT data (up to the Nyquist frequency).

Returns:

The magnitudes of the last FFT data.

getPowerMagnitudes

public float[][] getPowerMagnitudes()

Returns the power magnitudes of the last FFT data. The length of the returned array of magnitudes will be half the length of the FFT data (up to the Nyquist frequency). The power is calculated using:

10log10( real^2 + imaginary^2 )

Returns:

The magnitudes of the last FFT data.

getNormalisedMagnitudes

public float[][] getNormalisedMagnitudes(float scalar)

Scales the real and imaginary parts by the scalar prior to calculating the (square) magnitude for normalising the outputs. Returns only those values up to the Nyquist frequency.

Parameters:

scalar - The scalar

Returns:

Normalised magnitudes.

getReals

public float[][] getReals()

Returns just the real numbers from the last FFT. The result will include the symmetrical part.

Returns:

The real numbers

getScalingFactor

public float getScalingFactor()

Get the scaling factor in use.

Returns:

The scaling factor.

setScalingFactor

public void setScalingFactor(float scalingFactor)

Set the scaling factor to use. This factor will be applied to signal data prior to performing the FFT. The default is, of course, 1.

Parameters:

scalingFactor - The scaling factor to use.

isPadToNextPowerOf2

public boolean isPadToNextPowerOf2()

Returns whether the input will be padded to be the length of the next higher power of 2.

Returns:

TRUE if the input will be padded, FALSE otherwise.

setPadToNextPowerOf2

public void setPadToNextPowerOf2(boolean padToNextPowerOf2)

Set whether to pad the input to the next power of 2.

Parameters:

padToNextPowerOf2 - TRUE to pad the input, FALSE otherwise