Deconvolution of the Basilar Membrane Spreading Function Using Recursive Adaptive Renormalization

• Main Authors: Yu-Lan Lin, 林煜嵐
• Other Authors: 簡福榮
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/9bz7qg

碩士 === 國立臺北科技大學 === 電機工程系所 === 96 === For the purpose of storing up and transmitting audio signals effectively, audio coding is an important issue. After encoding the audio signals, the size of the signals could be reduced, contrarily, the quality of the coded signals might be worse than the original ones, and that is why Psychoacoustic Model plays an important role in audio coders. It can reduce the waste of the bits by removing the noise not perceived by human ears. In the Psychoacoustic Model, we have to deconvolve the noise excitation patterns with spreading function to obtain the correct noise masking threshold in each critical band. In this thesis, we use six methods to carry out or substitute the deconvolution operation. They are Renormalization, Inverse Filtering, Iterative Procedure, Greedy Algorithm, Adaptive Renormalization Method, and Recursive Adaptive Renormalization Method. The audio database used in the experiments is divided into 9 categories, including 4 kinds of musical instruments, Symphony, Hip-Hop, rock, and songs sung by both males and females. The experimental results show that using Recursive Adaptive Renormalization method could improve the accuracy of the deconvolution about 87 %. Objective tests show that the scores of using Recursive Adaptive Renormalization method are a little lower than using Renormalization in males and females singing, symphony, and Hip-Hop. The differences between both of them are from 0.02 to 0.2. On the other hand, the audio quality of using Recursive Adaptive Renormalization Method has higher scores than Renormalization in categories of flute, guitar, piano, cello, and rock.