 |
| VoCAS allows the quality evaluation of Automatic Speech Recognition (ASR) systems in reproducible conditions. It offers the following functions: |
 |
|
| Integration into HEAD acoustics product ecosystem: |
| - |
|
MFE VI.1 incl. mouth equalization, playback, monitoring |
| - |
|
Remote control of HAE-BGN/-car or 3PASS |
|
|
|
| Preparation of audio file databases: |
| - |
|
Record files with integrated recorder |
| - |
|
Easy import of existing files with import wizard |
| - |
|
Add meta-data to audio sources by setting values for different attributes |
|
|
|
| Reproducible execution of test sequences |
|
|
|
| Result acquisition, presentation and export |
|
|
|
| Flexible definition of test sequences: |
| - |
|
Template definition |
| - |
|
Variations via parameter sets (e.g. different background noise) |
|
|
|
|
| Test sequence parameter sets can determine the following values: |
|
|
| - |
|
Level at MRP |
| - |
|
Lombard effect (e.g. ITU-T P.1100) |
|
|
|
| Background noise settings |
| - |
|
Noise signals (e.g. from 3PASS) |
| - |
|
Noise levels |
|
|
|
| Different audio sources by changing meta data values |
| - |
|
e.g. “Eiffel Tower, Paris” or “Airport Cologne” for the attribute “Address” |
|
|
|
|
|
|
Typical test sequence for a navigation system. The test sequence template is modifiable by parameter sets (e.g. language, speaker, address). Realistic acoustic scenarios can also be chosen (e.g. background noise). A possible variant of the sequence template with specific parameter values is shown here. VoCAS leads the user through the complete test sequence while keeping track of all possible variants and allowing reproducible repetitions. |
|
|
|
|
|
|
|
VoCAS - Product Description
