net.idrnd.voicesdk.media

Class SpeechEndpointDetector

java.lang.Object

net.idrnd.voicesdk.common.VoiceSdkNativePeer

net.idrnd.voicesdk.media.SpeechEndpointDetector

All Implemented Interfaces:

java.lang.AutoCloseable

public class SpeechEndpointDetector
extends VoiceSdkNativePeer

Provides the functionality of speech end detection in audio stream.

Enables streaming scenario for end detection when audio data is processed by continuous buffers. Intended usage scenario is following:

• initialize SpeechEndpointDetector
• call 1 or more addSamples(byte[]) methods with incoming audio data
• call isSpeechEnded()
• if false, repeat
• if true, stop processing and VoiceSdkNativePeer.close()

End detection is parameterized with 2 thresholds: minimum speech length and maximum silence length. End detection is triggered when minimum speech length is accumulated and then uninterrupted silence longer that 'max silence length' occurs. This class is stateful and is not thread safe.

This class serves as gateway to native Voice SDK implementation and allocates resources on native heap. To release the allocated memory, AutoCloseable.close() method should be invoked when the instance is no longer needed.

Any method that delegates to native call may throw VoiceSdkEngineException

Field Summary

Fields inherited from class net.idrnd.voicesdk.common.VoiceSdkNativePeer

nativeId

Constructor Summary

Constructors

Constructor and Description
SpeechEndpointDetector(int minSpeechLengthMs, int maxSilenceLengthMs, int sampleRate) Initializes speech endpoint detector.

Method Summary

Modifier and Type	Method and Description
void	addSamples(byte[] bytes) Adds audio samples for processing in PCM16 format
void	addSamples(float[] floatSamples) Audio samples for processing encoded in normalized float format
void	addSamples(short[] pcm16Samples) Adds audio samples for processing in PCM16 format
protected void	addSamples1(byte[] bytes)
protected void	addSamples2(short[] bytes)
protected void	addSamples3(float[] bytes)
protected static long	init(int minSpeechLengthMs, int maxSilenceLengthMs, int sampleRate)
boolean	isSpeechEnded() Checks if speech end is detected after the previous addSamples(byte[]) call
protected void	release() This method should release resources on native layer (indirectly using nativeId)
void	reset() Resets detector, clearing all the accumulated statistics

Methods inherited from class net.idrnd.voicesdk.common.VoiceSdkNativePeer

close, equals, finalize, hashCode

Methods inherited from class java.lang.Object

clone, getClass, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

SpeechEndpointDetector

public SpeechEndpointDetector(int minSpeechLengthMs,
                              int maxSilenceLengthMs,
                              int sampleRate)

Initializes speech endpoint detector.

Parameters:

minSpeechLengthMs - the threshold for required accumulated speech duration

maxSilenceLengthMs - the threshold for the duration of continuous silence that triggers the end detection if the required amount of speech is accumulated

sampleRate - sample rate of incoming audio data stream

Throws:

VoiceSdkEngineException - wraps native exceptions

Method Detail

reset

public void reset()

Resets detector, clearing all the accumulated statistics

Throws:

VoiceSdkEngineException - wraps native exceptions

addSamples

public void addSamples(byte[] bytes)

Adds audio samples for processing in PCM16 format

Parameters:

bytes - Array of little-endian PCM16 audio bytes

Throws:

VoiceSdkEngineException - wraps native exceptions

addSamples

public void addSamples(short[] pcm16Samples)

Adds audio samples for processing in PCM16 format

Parameters:

pcm16Samples - Array of PCM16 audio samples

Throws:

VoiceSdkEngineException - wraps native exceptions

addSamples

public void addSamples(float[] floatSamples)

Audio samples for processing encoded in normalized float format

Parameters:

floatSamples - Array of float audio samples (in [-1, 1] range)

Throws:

VoiceSdkEngineException - wraps native exceptions

addSamples1

protected void addSamples1(byte[] bytes)

addSamples2

protected void addSamples2(short[] bytes)

addSamples3

protected void addSamples3(float[] bytes)

isSpeechEnded

public boolean isSpeechEnded()

Checks if speech end is detected after the previous addSamples(byte[]) call

Returns:

true if speech end is detected

init

protected static long init(int minSpeechLengthMs,
                           int maxSilenceLengthMs,
                           int sampleRate)

release

protected void release()

Description copied from class: VoiceSdkNativePeer

This method should release resources on native layer (indirectly using nativeId)

Specified by:

release in class VoiceSdkNativePeer