edu.cmu.sphinx.frontend.endpoint

Class SpeechClassifier

java.lang.Object

edu.cmu.sphinx.util.props.ConfigurableAdapter

edu.cmu.sphinx.frontend.BaseDataProcessor

edu.cmu.sphinx.frontend.endpoint.AbstractVoiceActivityDetector

edu.cmu.sphinx.frontend.endpoint.SpeechClassifier

All Implemented Interfaces:

DataProcessor, Configurable

public class SpeechClassifier
extends AbstractVoiceActivityDetector

Implements a level tracking endpointer invented by Bent Schmidt Nielsen.

This endpointer is composed of two main steps.

1. classification of audio into speech and non-speech
2. inserting SPEECH_START and SPEECH_END signals around speech and removing non-speech regions

The first step, classification of audio into speech and non-speech, uses Bent Schmidt Nielsen's algorithm. Each time audio comes in, the average signal level and the background noise level are updated, using the signal level of the current audio. If the average signal level is greater than the background noise level by a certain threshold value (configurable), then the current audio is marked as speech. Otherwise, it is marked as non-speech.

The second step of this endpointer is documented in the class SpeechMarker

See Also:

SpeechMarker

Field Summary

Fields

Modifier and Type	Field and Description
protected double	adjustment
protected double	averageNumber
protected double	background background signal level.
protected long	backgroundFrames
protected float	frameLengthSec
protected boolean	isSpeech
protected double	level average signal level.
protected double	minSignal minimum valid signal level.
static java.lang.String	PROP_ADJUSTMENT The property specifying the adjustment.
static java.lang.String	PROP_FRAME_LENGTH_MS The property specifying the endpointing frame length in milliseconds.
static java.lang.String	PROP_MIN_SIGNAL The property specifying the minimum signal level used to update the background signal level.
static java.lang.String	PROP_THRESHOLD The property specifying the threshold.
protected long	speechFrames
protected double	threshold
protected double	totalBackgroundLevel
protected double	totalSpeechLevel

Fields inherited from class edu.cmu.sphinx.util.props.ConfigurableAdapter

logger

Constructor Summary

Constructors

Constructor and Description
SpeechClassifier()
SpeechClassifier(int frameLengthMs, double adjustment, double threshold, double minSignal)

Method Summary

Modifier and Type	Method and Description
protected SpeechClassifiedData	classify(DoubleData audio) Classifies the given audio frame as speech or not, and updates the endpointing parameters.
Data	getData() Returns the next Data object.
boolean	getNoisy() Return the estimation if input data was noisy enough to break recognition.
double	getSNR() Retrieves accumulated signal to noise ratio in dbScale
void	initialize() Initializes this LevelTracker endpointer and DataProcessor predecessor.
boolean	isSpeech() Method that returns if current returned frame contains speech.
static double	logRootMeanSquare(double[] samples) Returns the logarithm base 10 of the root mean square of the given samples.
void	newProperties(PropertySheet ps) This method is called when this configurable component needs to be reconfigured.
protected void	reset() Resets this LevelTracker to a starting state.

Methods inherited from class edu.cmu.sphinx.frontend.BaseDataProcessor

getPredecessor, setPredecessor

Methods inherited from class edu.cmu.sphinx.util.props.ConfigurableAdapter

getName, initLogger, toString

Methods inherited from class java.lang.Object

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

Field Detail

PROP_FRAME_LENGTH_MS

@S4Integer(defaultValue=10)
public static final java.lang.String PROP_FRAME_LENGTH_MS

The property specifying the endpointing frame length in milliseconds.

See Also:

Constant Field Values

PROP_MIN_SIGNAL

@S4Double(defaultValue=0.0)
public static final java.lang.String PROP_MIN_SIGNAL

The property specifying the minimum signal level used to update the background signal level.

See Also:

Constant Field Values

PROP_THRESHOLD

@S4Double(defaultValue=10.0)
public static final java.lang.String PROP_THRESHOLD

The property specifying the threshold. If the current signal level is greater than the background level by this threshold, then the current signal is marked as speech. Therefore, a lower threshold will make the endpointer more sensitive, that is, mark more audio as speech. A higher threshold will make the endpointer less sensitive, that is, mark less audio as speech.

See Also:

Constant Field Values

PROP_ADJUSTMENT

@S4Double(defaultValue=0.003)
public static final java.lang.String PROP_ADJUSTMENT

The property specifying the adjustment.

See Also:

Constant Field Values

averageNumber

protected final double averageNumber

See Also:

Constant Field Values

adjustment

protected double adjustment

level

protected double level

average signal level.

background

protected double background

background signal level.

minSignal

protected double minSignal

minimum valid signal level.

threshold

protected double threshold

frameLengthSec

protected float frameLengthSec

isSpeech

protected boolean isSpeech

speechFrames

protected long speechFrames

backgroundFrames

protected long backgroundFrames

totalBackgroundLevel

protected double totalBackgroundLevel

totalSpeechLevel

protected double totalSpeechLevel

Constructor Detail

SpeechClassifier

public SpeechClassifier(int frameLengthMs,
                        double adjustment,
                        double threshold,
                        double minSignal)

SpeechClassifier

public SpeechClassifier()

Method Detail

newProperties

public void newProperties(PropertySheet ps)
                   throws PropertyException

Description copied from interface: Configurable

This method is called when this configurable component needs to be reconfigured.

Specified by:

newProperties in interface Configurable

Overrides:

newProperties in class ConfigurableAdapter

Parameters:

ps - a property sheet holding the new data

Throws:

PropertyException - if there is a problem with the properties.

initialize

public void initialize()

Initializes this LevelTracker endpointer and DataProcessor predecessor.

Specified by:

initialize in interface DataProcessor

Overrides:

initialize in class BaseDataProcessor

reset

protected void reset()

Resets this LevelTracker to a starting state.

logRootMeanSquare

public static double logRootMeanSquare(double[] samples)

Returns the logarithm base 10 of the root mean square of the given samples.

Parameters:

samples - the samples

Returns:

the calculated log root mean square in log 10

classify

protected SpeechClassifiedData classify(DoubleData audio)

Classifies the given audio frame as speech or not, and updates the endpointing parameters.

Parameters:

audio - the audio frame

Returns:

Data with classification flag

getData

public Data getData()
             throws DataProcessingException

Returns the next Data object.

Specified by:

getData in interface DataProcessor

Specified by:

getData in class BaseDataProcessor

Returns:

the next Data object, or null if none available

Throws:

DataProcessingException - if a data processing error occurs

isSpeech

public boolean isSpeech()

Method that returns if current returned frame contains speech. It could be used by noise filter for example to adjust noise spectrum estimation.

Specified by:

isSpeech in class AbstractVoiceActivityDetector

Returns:

if current frame is speech

getSNR

public double getSNR()

Retrieves accumulated signal to noise ratio in dbScale

Returns:

signal to noise ratio

getNoisy

public boolean getNoisy()

Return the estimation if input data was noisy enough to break recognition. The audio is counted noisy if signal to noise ratio is less then 20dB.

Returns:

estimation of data being noisy