Accelerated Feature Extraction Tool
A fast feature extraction software tool for speech analysis and processing. It incorporates standard MFCC, PLP, and TRAPS features. The tool is a specially designed to process very large audio data sets. It uses GPU acceleration if compatible GPU available. CPU SSE intristic instruction set is used in cases where no compatible GPU present.
Source codes and binary executable files
It is an open-source software. Source codes and binary executable files for the most popular platforms are in repozitory at SourceForge.net
user guide
The user guide with exapmples of using of the software is in PFD format HERE.
VERIFICation and validation
The software wav verified by processing of a few tenths of hours of telephony and high-quality speech. The computed output feature-vectors were compared with a reference HYK software. The achieved results were comparable. The total elapsed time was 3 to 20 times shorter tahn the HTK reference. 10 hours of speech were processed in 6 minutes. The speed-up strongly depends on individual computer hardware configuration, a specially on hard disk speed and GPU performance.
Licence
It is an open-source software which can be used freely.
Confirmation of usage
Software is available freely to all laboratories, which are invovled in speech processing domain. It is also used at the Department of Cybernetics at Univerzity of West Bohemia in Pilsen.
Contact form
This software is protected by license. To download or get more information, please fill in the form below: