Parallel processing of statistical data using graphical processor for changeability evaluation of human voice parameters

Advanced computer analysis of human voice is still time and memory consuming process. One of the greatest problems is reduction of time calculations to obtain the proper results, even for relatively short samples of the voice. All signal processing operations require significant number of calculations, therefore we propose to exploit parallel processing capabilities of graphical processor. Due to requirements of graphics generation, graphical processing units have become specialized chipsets with multi-core architecture, capable of executing a single instruction set on multiple sets of data. This approach is desirable in data intensive calculations, therefore special libraries were made by graphic chips manufacturers, to enable an use of these chips in non-graphical applications. It is demonstrated in this paper that using graphical processors for signal processing, results can be obtained in a shorter time, feasible for processing of multiple full length samples.