Efficient computation of spherical harmonic transform using parallel architecture of CUDA

Weiyu Huang*, Zubair Khalid, Rodney A. Kennedy

*Corresponding author for this work

    Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

    1 Citation (Scopus)

    Abstract

    Spherical harmonics serve as basis functions on the unit sphere and spherical harmonic transform is required in analysis and processing of signals in the spectral domain. We investigate the possibility of parallel computation of spherical harmonic transform using Compute Unified Device Architecture (CUDA) with no communication between parallel kernels. We identify the parallel components in the widely used spherical harmonic transform method proposed by Driscoll and Healy. We provide the implementation details and compare the computational complexity with the sequential algorithm. For a given bandlimited signal with maximum spherical harmonics degree L, using the O(L) number of parallel processing kernels, we present that the spherical harmonic coefficients can be calculated in O(Llog 2L) time as compared to O(L 2log 2L). For corroboration, we provide the simulation results using CUDA which indicate the reduction in computational complexity.

    Original languageEnglish
    Title of host publication5th International Conference on Signal Processing and Communication Systems, ICSPCS'2011 - Proceedings
    DOIs
    Publication statusPublished - 2011
    Event5th International Conference on Signal Processing and Telecommunication Systems, ICSPCS'2011 - Honolulu, HI, United States
    Duration: 12 Dec 201114 Dec 2011

    Publication series

    Name5th International Conference on Signal Processing and Communication Systems, ICSPCS'2011 - Proceedings

    Conference

    Conference5th International Conference on Signal Processing and Telecommunication Systems, ICSPCS'2011
    Country/TerritoryUnited States
    CityHonolulu, HI
    Period12/12/1114/12/11

    Fingerprint

    Dive into the research topics of 'Efficient computation of spherical harmonic transform using parallel architecture of CUDA'. Together they form a unique fingerprint.

    Cite this