Parallel Implementation Of Huwang Model Algorithm For Casa

Computational Auditory Scene Analysis (CASA) uses computer technology to simulate the process and auditory physiological function of the human auditory psychological, and finally realizes a computer having an ability to process sounds like the human ear through isolating and interpreting. CASA has become a new edge subject in academic reasearch. Previous researches already did computer-based sound process, automatic speech recognition.However, making the the product of automatic speech recognition going to real life is still a problem now, many researchers have noted that there is a great difference between the HMM statistical model used by the main framework of computer speech recognition system and the human auditory system. This gap has arised great interests in study of the human auditory system, which drives the progress of Computational Auditory Scene Analysis subject.HuWang model is a classical algorithm of Computational Auditory Scene Analysis.The main idea is to separate low frequency and high frequency audio region, The entire model system has several stages.Although the Aalgorithm is feasible, it has high time complexity, thus the speed is too slow, to be used in practice, This needs to accelerate the speed of the algorithm.This thesis chooses HuWang model for single-channel voice processing, and use the CUDA to do parallel implementation.This thesis mainly introduces HuWang model algorithm acceleration technology for Computational Auditory Scene Analysis using, GPUs parallel computing method.Finally, the performance of the parallized algorithm and testing results are given.