VoCAS – Test voice recognition systems under realistic and reproducible conditions

Voice recognition systems are already part of numerous applications including... 

Voice recognition systems are already part of numerous applications including voice-operated IoT devices like smart speaker, multimedia systems in vehicles, mobile phones, and tablets. The software VoCAS (Voice Control Analysis System) evaluates these devices realistically and objectively in complex scenarios. VoCAS takes into account crucial factors like acoustic environment, background noise, different talkers as well as different languages or accents, which significantly influence the performance of voice recognition systems.

By means of reproducible test sequences adapted to individual applications, manufacturers receive important information to optimize their voice recognition systems.

More Less

Voice recognition systems are already part of numerous applications including voice-operated IoT devices like smart speaker, multimedia systems in vehicles, mobile phones, and tablets. The software VoCAS (Voice Control Analysis System) evaluates these devices realistically and... 

Voice recognition systems are already part of numerous applications including voice-operated IoT devices like smart speaker, multimedia systems in vehicles, mobile phones, and tablets. The software VoCAS (Voice Control Analysis System) evaluates these devices realistically and objectively in complex scenarios. VoCAS takes into account crucial factors like acoustic environment, background noise, different talkers as well as different languages or accents, which significantly influence the performance of voice recognition systems.

By means of reproducible test sequences adapted to individual applications, manufacturers receive important information to optimize their voice recognition systems.

More Less

Highlights

  • VoCAS is capable of testing voice recognition systems under realistic and reproducible conditions
  • The software enables complex test scenarios with multiple talkers, background noise, and the acoustic environment
  • VoCAS is perfectly suited for benchmarking of different voice recognition systems
Testing voice recognition systems in a lab with VoCAS
Testing voice recognition systems in cars with VoCAS

Applications

  • Reproducible tests of voice recognition systems in realistically simulated environments (complex scenarios with multiple talkers, background noise, acoustic environment)
  • Benchmarking of different voice recognition systems

Features

  • Users can create customized and reproducible test sequences in VoCAS, fitting to the voice recognition system to be evaluated
  • Test the measurement object by choosing various parameter sets like e.g. different talkers, languages, background noises, and reverberation
  • The easy-to-use interface allows to build automated test sequences for any voice recognition system
  • Via an integrated recorder, users are able to record own voice commands as well as import existing individual audio files
  • Expand the audio database required for testing voice control systems individually by importing own voice commands into the data management system
  • Users are able to cut, filter and adjust audio files to defined speech levels
  • Tags allow fast and easy file management of a large number of voice commands
  • Retrieve and assign all voice commands easily by tagging
  • Record new voice commands and create own sentences for individual tests with the integrated recorder
  • Python scripts enable users to customize and automate test sequences for the voice recognition system
  • The background noise simulation system (3PASS flex, 3PASS lab, HAE-BGN, HAE-car) can be controlled via VoCAS to select the appropriate realistic background noise scenario including room reverberation (3PASS reverb) for the application under test (e.g. vehicle, cafeteria, platform)
  • Test results are easy to interpret due to clear representation of results (percentage values, colorful accentuation)
  • The hardware platform labCORE is controllable via VoCAS

 

 


Downloads

Relevant Products

Modular & Scalable
labCORE
(7700)
Directional
HRT I
(6498)
Directional
HRR I
(6498)
Artificial Heads
HMS II.3
(1230)
Artificial Heads
HMS II.3-LN
(1230.3)
Artificial Heads
HMS II.6
(1389)
Background Noise Simulation
3PASS lab
(6690)
Background Noise Simulation
3PASS flex
(6695)
Background Noise Simulation
3PASS reverb
(6696)
Background Noise Simulation
labBGN
(6486)
Background Noise Simulation
MSA I and MSA II
(6487, 6487.2)