io.spokestack.spokestack.webrtc

Class VoiceActivityDetector

java.lang.Object

io.spokestack.spokestack.webrtc.VoiceActivityDetector

All Implemented Interfaces:

SpeechProcessor, AutoCloseable

public class VoiceActivityDetector
extends Object
implements SpeechProcessor

Voice Activity Detection (VAD) pipeline component

VoiceActivityDetector is a speech pipeline component that implements Voice Activity Detection (VAD) using the webrtc native component. The detector processes each frame and sets the speech context to speech/nonspeech based on the results of the VAD algorithm. The VAD implementation is based on the webrtc VAD in the Chromium browser. It supports 16-bit PCM samples.

This pipeline component supports the following configuration properties:

• sample-rate (int): audio sample rate, in Hz (supports 8000/16000/32000/48000Hz)
• frame-width (int): audio frame width, in ms (supports 10/20/30ms)
• vad-mode (string): detector mode, one of the following:
  • quality: highest recall
  • low-bitrate: higher recall
  • aggressive: higher precision
  • very-aggressive: highest precision
• vad-rise-delay (int): rising-edge detection run length, in ms; this value determines how many positive samples must be received to flip the detector to positive
• vad-fall-delay (int): falling-edge detection run length, in ms; this value determines how many negative samples must be received to flip the detector to negative

The detector uses a simple consecutive value filter to eliminate noisy transitions.

Field Summary

Fields

Modifier and Type	Field and Description
static String	DEFAULT_MODE default voice detection mode (high precision).

Constructor Summary

Constructors

Constructor and Description
VoiceActivityDetector(SpeechConfig config) constructs a new trigger instance.

Method Summary

Modifier and Type	Method and Description
void	close() destroys the unmanaged VAD instance.
void	process(SpeechContext context, ByteBuffer frame) processes a frame of audio.
void	reset() resets all state internal to the stage.

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_MODE

public static final String DEFAULT_MODE

default voice detection mode (high precision).

See Also:

Constant Field Values

Constructor Detail

VoiceActivityDetector

public VoiceActivityDetector(SpeechConfig config)

constructs a new trigger instance.

Parameters:

config - the pipeline configuration instance

Method Detail

close

public void close()

destroys the unmanaged VAD instance.

Specified by:

close in interface AutoCloseable

process

public void process(SpeechContext context,
                    ByteBuffer frame)

processes a frame of audio.

Specified by:

process in interface SpeechProcessor

Parameters:

context - the current speech context

frame - the audio frame to detect

reset

public void reset()

Description copied from interface: SpeechProcessor

resets all state internal to the stage.

Specified by:

reset in interface SpeechProcessor