Speech recognition systems are an essential component of voice-controlled IoT devices...
Speech recognition systems are an essential component of voice-controlled IoT devices such as smart speakers, smartphones, tablets, and multimedia systems in vehicles. The VoCAS (Voice Control Analysis System) software realistically and objectively evaluates the speech recognition quality of all these devices in complex scenarios. VoCAS considers all factors that significantly influence the quality of speech recognition systems: the acoustic environment, background noise, different talkers, and different languages or accents.
With the help of reproducible test sequences adapted to the respective applications, manufacturers thus receive essential information for optimizing their speech recognition systems.
Speech recognition systems are an essential component of voice-controlled IoT devices such as smart speakers, smartphones, tablets, and multimedia systems in vehicles. The VoCAS (Voice Control Analysis System) software realistically and objectively evaluates the speech recognition...
Speech recognition systems are an essential component of voice-controlled IoT devices such as smart speakers, smartphones, tablets, and multimedia systems in vehicles. The VoCAS (Voice Control Analysis System) software realistically and objectively evaluates the speech recognition quality of all these devices in complex scenarios. VoCAS considers all factors that significantly influence the quality of speech recognition systems: the acoustic environment, background noise, different talkers, and different languages or accents.
With the help of reproducible test sequences adapted to the respective applications, manufacturers thus receive essential information for optimizing their speech recognition systems.
Highlights
- VoCAS tests speech recognition systems under realistic and reproducible conditions.
- The software enables the automation of complex test scenarios with up to four simultaneous talkers, background noise, reverberation, and echo.
- VoCAS is ideally suited for benchmarking different speech recognition systems.
Features
- In VoCAS, you can create customized and reproducible test sequences that suit the speech recognition system you want to evaluate.
- Set numerous parameters for evaluating the test object, for example, up to four different talkers, various languages, background noise, and reverberation.
- Create automated test sequences for any speech recognition system via the easy-to-use interface.
- You can import your existing audio files or record voice commands via an integrated recorder and create your own sentences for individual tests.
- Individually expand the audio database needed for testing speech control systems by importing your own voice commands into the data management system.
- Cut and filter audio files and adjust them to defined speech levels.
- Tags allow quick and easy management of a large number of voice commands.
- Use Python scripts to customize test sequences for the speech recognition system and automate processes.
- From VoCAS, you control the background noise simulation system (3PASS flex, 3PASS lab, HAE-BGN, HAE-car) and select the ideal background noise scenario, including spatial reverberation (3PASS reverb) for the respective application (vehicle, cafeteria, train platform…).
- The test results are easy to interpret thanks to the clear presentation of the results (percentages, color highlighting).
- You control the hardware platform labCORE directly with VoCAS.
