|Title:||Optimization of the Gaussian mixture model evaluation on GPU|
|Other Titles:||Optimalizovaný výpočet směsí normálních rozložení na GPU|
Psutka, Josef V.
|Citation:||VANĚK, Jan; TRMAL, Jan; PSUTKA, Josef V.; PSUTKA, Josef. Optimization of the Gaussian mixture model evaluation on GPU. In: Proceedings of ICSPL 2011: 12th Annual Conference of the International Speech Communication Association 2011, Firenze, Italy, 27-31 August 2011. [Baixas]: ISCA, 2011, p. 1748-1751. ISSN 1990-9772.|
|Keywords:||CUDA;GPU;OpenCL;směsi Gaussovských modelů|
|Keywords in different language:||CUDA;GPU;OpenCL;Gaussian mixture models|
|Abstract in different language:||In this paper we present a highly optimized implementation of Gaussian mixture acoustic model evaluation algorithm. Evaluation of these likelihoods is one of the most computationally intensive parts of automatics speech recognizers but it can be well-parallelized and offloaded to GPU devices. Our approach offers significant speed-up compared to the recently published approaches, since it exploits the GPU architecture better. All the recent implementations were programmed either in CUDA or OpenCL GPU programming frameworks. We present results for both; CUDA as well as OpenCL. Results suggest that even very large acoustic models can be utilized in real-time speech recognition engines on computers and laptops equipped with a low-end GPU. Optimization of acoustic likelihoods computation on GPU enables to use the remaining GPU resources for offloading of other computeintensive parts of LVCSR decoder. Other possible use of the freed GPU resources is to evaluate several acoustic models at the same time and use fusion techniques or model selection techniques to improve the quality of resulting conditional likelihoods under diverse conditions.|
|Rights:||© Jan Vaněk - Jan Trmal - Josef V. Psutka - Josef Psutka|
|Appears in Collections:||Články / Articles (KIV)|
Články / Articles (KKY)
Files in This Item:
|VanekJ_2011_Optimizationofthe.pdf||Plný text||226,25 kB||Adobe PDF||View/Open|
Please use this identifier to cite or link to this item:
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.