Publication: Implementation of mp3 decoder with parallel technique
Loading...
Date
2020-06-01
Authors
Esther, Ong Tze Shin
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis describes the mp3 decoding algorithm implemented in Matlab. The algorithm will read the mp3 file from the input, decode the mp3 bitstream into pulse code modulated (PCM) outputs using a standard mp3 decoding algorithm and written into wav file. One option to improve the performance is by using parallel processing. An implementation of mp3 decoding with parallel technique in Matlab is presented. It focuses on software-based, with different number of tasks. The Parallel Computing Toolbox in Matlab is among several available tool that offer this capability. Three decoding methods are used in the implementation including sequential decoding, two-task parallel decoding and three-task parallel decoding. Six songs in mp3 file format have been used to test the different approaches of audio decoding. Equality of two variances based on Bonett’s method, T-test analysis and PSNR test have been employed to check the validity of the PCM output obtained from the sequential and parallel decoding compared to the PCM output generated in Matlab. Results of the tests can be observed from p-value and PSNR value respectively. The time taken to decode the mp3 file by using the three methods mentioned above are compared and their performances are evaluated. From the experimental results, two-task parallel decoding has improved the decoding time by 13.97% with speedup of 1.163 while three-task parallel decoding has 5.06% improvement in the decoding time with speedup of 1.053. This is because as the number of tasks increases, so does the overhead which comes from the communication during the parallel decoding. It shows that two-task parallel decoding is optimum to perform the mp3 decoding.