FeatureExtractionAggregator

Overview

Package

Class

Use

Tree

Deprecated

Index

Help

PREV CLASS NEXT CLASS

FRAMES NO FRAMES

SUMMARY: NESTED | FIELD | CONSTR | METHOD

DETAIL: FIELD | CONSTR | METHOD

marf.FeatureExtraction
Class FeatureExtractionAggregator

java.lang.Object
  marf.Storage.StorageManager
      marf.FeatureExtraction.FeatureExtraction
          marf.FeatureExtraction.FeatureExtractionAggregator

All Implemented Interfaces:: java.io.Serializable, java.lang.Cloneable, IFeatureExtraction, IStorageManager

public class FeatureExtractionAggregator
extends FeatureExtraction
extends FeatureExtraction

This class by itself does not do any feature extraction, but instead allows concatenation of the results of several actual feature extractors to be combined in a single result. This should give the classification modules more discriminatory power (e.g. when combining the results of FFT and F0 together).

FeatureExtractionAggregator itself still implements the FeatureExtraction API in order to be used in the main pipeline of MARF.

The aggregator expects ModuleParams to be set to the enumeration constants of a module to be invoked followed by that module's enclosed instance ModuleParams. As of this implementation, that enclosed instance of ModuleParams isn't really used, so the main limitation of the aggregator is that all the aggregated feature extractors act with their default settings. This will happen until the pipeline is re-designed a bit to include this capability.

The aggregator clones the incoming preprocessed sample for each feature extractor and runs each module in a separate thread. A the end, the results of each thread are collected in the same order as specified by the initial ModuleParams and returned as a concatenated feature vector. Some meta-information is available if needed.

$Id: FeatureExtractionAggregator.java,v 1.9 2007/12/18 03:45:40 mokhov Exp $

Since:: 0.3.0.5
Version:: $Revision: 1.9 $
Author:: Serguei Mokhov
See Also:: FFT, F0, ModuleParams, FeatureExtraction, Serialized Form

Nested Class Summary
`class`	`FeatureExtractionAggregator.FeatureExtractionThread` A feature extraction thread simply calls its concrete encapsulated feature extraction module and collects any results or errors from its run.

Field Summary
`static java.lang.String`	`ERR_AGGR_MODULES_FAILED` Error indicating the fact that one or more aggregated modules failed.
`static java.lang.String`	`ERR_MALFORMED_PARAMS` Error message for no malformed ModuleParams data structure.
`static java.lang.String`	`ERR_NO_FEATURES` A rare error when all feature extractors ran, but there were a total of zero features extracted.
`static java.lang.String`	`ERR_NO_MODULES` Error message for no modules found.
`protected ExpandedThreadGroup`	`oFeatureExtractors` A collection of the feature extraction threads.

Fields inherited from class marf.FeatureExtraction.FeatureExtraction
`adFeatures, oFeatureSet, oPreprocessing`

Fields inherited from class marf.Storage.StorageManager
`bDumpOnNotFound, iCurrentDumpMode, oObjectToSerialize, strFilename`

Fields inherited from interface marf.FeatureExtraction.IFeatureExtraction
`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
`FeatureExtractionAggregator(IPreprocessing poPreprocessing)` Implementation of the feature extraction interface.

Method Summary
`boolean`	`extractFeatures()` Implementation of the feature extraction interface.
`boolean`	`extractFeatures(double[] padSampleData)` Implementation of the feature extraction interface.
`protected boolean`	`extractFeaturesImplementation(double[] padSampleData)` Performs the actual business logic of the aggregator per described algorithm.
`static java.lang.String`	`getMARFSourceCodeRevision()` Returns source code revision information.

Methods inherited from class marf.FeatureExtraction.FeatureExtraction
`backSynchronizeObject, clone, getFeaturesArray, getPreprocessing, setPreprocessing`

Methods inherited from class marf.Storage.StorageManager
`dump, dumpBinary, dumpCSV, dumpGzipBinary, dumpHTML, dumpSQL, dumpXML, enableDumpOnNotFound, equals, getDefaultExtension, getDefaultExtension, getDumpMode, getFilename, getObjectToSerialize, hashCode, restore, restoreBinary, restoreCSV, restoreGzipBinary, restoreHTML, restoreSQL, restoreXML, setDumpMode, setFilename, toString`

Methods inherited from class java.lang.Object
`finalize, getClass, notify, notifyAll, wait, wait, wait`

Field Detail

ERR_NO_MODULES

public static final java.lang.String ERR_NO_MODULES

Error message for no modules found.

See Also:: Constant Field Values

ERR_MALFORMED_PARAMS

public static final java.lang.String ERR_MALFORMED_PARAMS

Error message for no malformed ModuleParams data structure.

See Also:: Constant Field Values

ERR_AGGR_MODULES_FAILED

public static final java.lang.String ERR_AGGR_MODULES_FAILED

Error indicating the fact that one or more aggregated modules failed. Should be suffixed with the detailed error information.

See Also:: Constant Field Values

ERR_NO_FEATURES

public static final java.lang.String ERR_NO_FEATURES

A rare error when all feature extractors ran, but there were a total of zero features extracted.

See Also:: Constant Field Values

oFeatureExtractors

protected transient ExpandedThreadGroup oFeatureExtractors

A collection of the feature extraction threads.

Constructor Detail