VDK 5.4.0 - 30th October 2023

Summary

We are pleased to introduce VDK 5.4.0 which is a new technology release: The Speech Enhancer.

Speech enhancement on real world music

This technology will empower your voice in an audio file or audio recording. It is meant to be used coupled to the Speech recognition technology to improve its accuracy in noisy environments.

You can find more details about other minor changes described below, under relevant components.

Versions

Studio

Component	Version
Studio	5.5.2 UPDATED
Dictionary manager	2.1.1 UPDATED
Free-speech	5.1.1 UPDATED
Grammar editor	5.1.1 UPDATED
Phonetic editor	5.1.1 UPDATED
Project editor	2.2.1 UPDATED
Simple assistant maker	5.1.1 UPDATED
TTS manager	5.1.1 UPDATED
Voice biometrics	4.4.1 UPDATED
Nlu Editor	2.1.1 UPDATED
Speech Enhancer	1.0.2 NEW
Vsdk-daemon	1.12.2 UPDATED

Samples

Component	Version
Chained grammars	5.1.1 UPDATED
Dynamic grammar	5.1.1 UPDATED
Simple application	5.1.0 UPDATED
Tts	5.1.0 UPDATED
Voice biometrics	5.1.1 UPDATED
Voice Commands Language Understanding	2.0.1 UPDATED
Speech Enhancement	1.0.0 NEW

VSDK (C++)

Component	Version
Vsdk	9.1.0 UPDATED
Vsdk-vasr	5.2.2 UPDATED
Vsdk-csdk	5.1.1 UPDATED
Vsdk-tnl	3.2.0 UPDATED
Vsdk-baratinoo	4.2.0 UPDATED
Vsdk-vtapi	3.2.1 UPDATED
Vsdk-idvoice	4.3.0 UPDATED
Vsdk-tssv	3.3.0 UPDATED
Vsdk-vnlu	2.2.0 UPDATED
Vsdk-audio-portaudio	3.5.1 UPDATED
Vsdk-s2c	1.1.1 NEW

VSDK (JAVA)

Component	Version
Vsdk	5.3.0 UPDATED
Vsdk-csdk	5.1.1 UPDATED
Vsdk-tnl	2.5.0
Vsdk-baratinoo	4.0.0
Vsdk-vtapi	3.1.0 UPDATED
Vsdk-idvoice	2.5.0 UPDATED
Vsdk-tssv	2.6.0 UPDATED
Vsdk-vasr	2.1.4 UPDATED
Vsdk-vnlu	1.0.2 UPDATED
Vsdk-s2c	1.0.0 NEW

Details

Studio

Bug-fixes:

Some styling was missing for spin boxes
The support link was broken in the About VDK Studio view
At startup, translations were missing on restored UI elements

Project Editor

Bug-fixes:

Removing a technology subcard could in some cases lead to a crash when not alone in the technology

Vsdk-daemon

Features:

Add new routes for the Speech Enhancement technology

Samples (C++)

Chained grammars

Improvements:

Now use a utility library vsdk-samples-utils to wrap the custom Event loop system

Dynamic grammar

Improvements:

Now use a utility library vsdk-samples-utils to wrap the custom Event loop system

Simple application

Bug-fixes:

When using Vsdk-vtapi as the TTS SDK, it could happen that the synthesized audio gets truncated

Improvements:

Now use a utility library vsdk-samples-utils to wrap the custom Event loop system

Tts

Improvements:

Now use a utility library vsdk-samples-utils to wrap the custom Event loop system

Voice Biometrics

Improvements:

Now use a utility library vsdk-samples-utils to wrap the custom Event loop system

Voice Commands Language Understanding

Improvements:

Now use a utility library vsdk-samples-utils to wrap the custom Event loop system

VSDK (C++)

Vsdk

Removals:

Old and deprecated AFE technology from the API

Features:

New Speech Enhancement technology API

Vsdk-vasr

Bug-fixes:

Do not remove all models from the recognizer when it has recognize an utterance but remove only the concerned one
Send the RECOGNIZER_STOPPED event only when the user has stopped sending audio data
Do not reset the uptime when receiving a final result
For dynamic models, the default phonetic alphabet was L&HP instead of Kirshenbaum

Vsdk-csdk

Bug-fixes:

Prevent a crash when Recognizer::stop() fails
Reinstalling models would sometimes silently do nothing

Removals:

Deprecated exception macros

Vsdk-tnl

Bug-fixes:

Avoid a potential deadlock situation

Vsdk-vnlu

Bug-fixes:

The tokenizer was too aggressive and could split entity names

Vsdk-audio-portaudio

Bug-fixes:

Frames were confused with samples in some algorithms

VSDK (Java)

Vsdk

Features:

Offer a way to check for audio quality using voice biometrics
New Speech Enhancement technology API

Vsdk-idvoice

Features:

New audio quality verification method

Vsdk-tssv

Features:

New audio quality verification method