marf.Preprocessing.FFTFilter
Class FFTFilter

java.lang.Object
  marf.Storage.StorageManager
      marf.Preprocessing.Preprocessing
          marf.Preprocessing.FFTFilter.FFTFilter

All Implemented Interfaces:

java.io.Serializable, IPreprocessing, IStorageManager

Direct Known Subclasses:

BandpassFilter, HighFrequencyBoost, HighPassFilter, LowPassFilter

public abstract class FFTFilter
extends Preprocessing

FFTFilter class implements filtering using the FFT algorithm.

Derivatives must set frequency response based on the type of filter they are.

$Id: FFTFilter.java,v 1.22 2005/06/16 19:58:47 mokhov Exp $

Since:

0.0.1

Version:

$Revision: 1.22 $

Author:

Stephen Sinclair, Serguei Mokhov

See Also:

Serialized Form

Field Summary

protected double[]	adFreqResponse Frequency repsonse to be multiplied by the incoming value.
static int	DEFAULT_FREQUENCY_RESPONSE_SIZE Default size of the frequency response vector, 128.

Fields inherited from class marf.Preprocessing.Preprocessing

oSample

Fields inherited from class marf.Storage.StorageManager

bDumpOnNotFound, iCurrentDumpMode, oObjectToSerialize, strFilename

Fields inherited from interface marf.Preprocessing.IPreprocessing

MARF_INTERFACE_CODE_REVISION

Fields inherited from interface marf.Storage.IStorageManager

DUMP_BINARY, DUMP_CSV_TEXT, DUMP_GZIP_BINARY, DUMP_HTML, DUMP_SQL, DUMP_XML, MARF_INTERFACE_CODE_REVISION, STORAGE_FILE_EXTENSIONS

Constructor Summary

FFTFilter(Preprocessing poPreprocessing)
Pipelined constructor.

FFTFilter(Sample poSample)
FFTFilter Constructor.

Method Summary

boolean	filter(double[] sample, double[] filtered) Perform a filter by the following algorithm: (1) sample -> window -> FFT -> buffer (2) buffer * frequency response (3) buffer -> IFFT -> window -> sample.
abstract void	genereateResponseCoefficients() Creates frequency response coefficients and sets applies them to the frequency response vector.
static java.lang.String	getMARFSourceCodeRevision() Returns source code revision information.
boolean	preprocess() FFTFilter implementation of preprocess().
void	setFrequencyResponse(double[] padPesponse) Sets frequency response.

Methods inherited from class marf.Preprocessing.Preprocessing

backSynchronizeObject, cropAudio, getSample, normalize, normalize, normalize, removeNoise, removeSilence

Methods inherited from class marf.Storage.StorageManager

dump, dumpBinary, dumpCSV, dumpGzipBinary, dumpHTML, dumpSQL, dumpXML, enableDumpOnNotFound, getDefaultExtension, getDefaultExtension, getDumpMode, getFilename, restore, restoreBinary, restoreCSV, restoreGzipBinary, restoreHTML, restoreSQL, restoreXML, setDumpMode, setFilename

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_FREQUENCY_RESPONSE_SIZE

public static final transient int DEFAULT_FREQUENCY_RESPONSE_SIZE

Default size of the frequency response vector, 128.

See Also:

Constant Field Values

adFreqResponse

protected transient double[] adFreqResponse

Frequency repsonse to be multiplied by the incoming value.

Constructor Detail

FFTFilter

public FFTFilter(Preprocessing poPreprocessing)
          throws PreprocessingException

Pipelined constructor.

Parameters:

poPreprocessing - followup preprocessing module

Throws:

PreprocessingException

Since:

0.3.0

FFTFilter

public FFTFilter(Sample poSample)
          throws PreprocessingException

FFTFilter Constructor.

Parameters:

poSample - incoming sample

Throws:

PreprocessingException

Method Detail

preprocess

public boolean preprocess()
                   throws PreprocessingException

FFTFilter implementation of preprocess().

It does call removeNoise() and removeSilence() if they were explicitly requested by an app before applying filtering.

NOTE: it alters inner Sample by resetting its data array to the new filtered values.

Returns:

true if there was something filtered out

Throws:

PreprocessingException - if the frequency response is null

See Also:

IPreprocessing.removeNoise(), IPreprocessing.removeSilence()

setFrequencyResponse

public final void setFrequencyResponse(double[] padPesponse)

Sets frequency response. Derivatives must call this method before any preprocessing occurs. TODO: use marf.util.Arrays.copy()

Parameters:

padPesponse - desired frequency response coefficients

filter

public final boolean filter(double[] sample,
                            double[] filtered)
                     throws PreprocessingException

Perform a filter by the following algorithm: (1) sample -> window -> FFT -> buffer
(2) buffer * frequency response
(3) buffer -> IFFT -> window -> sample. Window used is square root of Hamming window, because the sum at half-window overlap is a constant, which avoids amplitude distortion from noise. Also, start sampling at -responseSize/2, in order to avoid amplitude distortion of the first half of the first window.

Parameters:

sample - incoming sample analog data

filtered - will contain data after the filter was applied. "filtered" must be at least as long as "sample".

Returns:

true if some filtering actually happened

Throws:

PreprocessingException - if the filtered and sample data arrays are not of the same size, the frequency response was not set, or there was an underlying FeatureExctractionException while executing the underlying FFT algorithm.

genereateResponseCoefficients

public abstract void genereateResponseCoefficients()

Creates frequency response coefficients and sets applies them to the frequency response vector. Must be overridden by individual filters.

Since:

0.3.0

getMARFSourceCodeRevision

public static java.lang.String getMARFSourceCodeRevision()

Returns source code revision information.

Returns:

revision string

Since:

0.3.0