To improve processing speed, you can use the Universal Speech Model to select only the key model components for use in scoring. Set the CompSelect
parameter to the number of components to select.
A lower value leads to faster processing; however, there is a trade-off between processing speed and performance, so you might want to increase the value of the CompSelect
parameter if performance is less than you expect.
To use all components, set CompSelect
to 0
.
For the speakerid
module, HPE recommends that you set this parameter to match the setting that you used during optimization.
For the audiotemplatedevel
module, you must use the same setting for all actions during the threshold estimation process. You cannot use scores that you generated with different settings.
Action: | AddTask |
Task: | SpkIdDevel
SpkIdDevelStream |
Type: | Integer |
Default: | 8 |
Example: | CompSelect=4
|
See Also: | CompSelect (configuration parameter) |
|