Vector Cross Product and Coordinate Rotation Based nD Hybrid FastICA

Bhardwaj, Swati and Raghuraman, Shashank and Adapa, Bhagyaraja and Acharyya, Amit (2018) Vector Cross Product and Coordinate Rotation Based nD Hybrid FastICA. Journal of Low Power Electronics, 14 (2). pp. 351-364. ISSN 1546-1998

Full text not available from this repository.

Abstract

This paper proposes an n-dimensional (nD) Hybrid FastICA methodology introducing the concept of Vector Cross Product into the Coordinate Rotation Digital Computer (CORDIC) based FastICA to attain high computation speed. The use of Vector Cross Product into the nD FastICA eliminates the complete Iteration stage required for computation of the nth weight vector. This introduction of Vector Cross Product into the entirely sequential procedure of CORDIC-based FastICA provides significant gain in terms of computation time. The functionality of the proposed methodology is validated by applying it on random signals, audio signals and biomedical signals. Subsequently, the hardware architecture has also been implemented and validated. The average improvement in computation time obtained by employing this methodology for 4D to 7D FastICA with a typical case of 1024 samples was found to be a minimum of 77% for nth vector computation, which further increases with increase in the iterations required for nth vector computation and the frame-length.

[error in script]
IITH Creators:
IITH CreatorsORCiD
Acharyya, Amithttp://orcid.org/0000-0002-5636-0676
Item Type: Article
Uncontrolled Keywords: Blind source separation, Independent component analysis, Vector cross product
Subjects: Electrical Engineering
Divisions: Department of Electrical Engineering
Depositing User: Team Library
Date Deposited: 05 Oct 2018 07:15
Last Modified: 05 Oct 2018 07:15
URI: http://raiith.iith.ac.in/id/eprint/4467
Publisher URL: http://doi.org/10.1166/jolpe.2018.1551
OA policy: http://www.sherpa.ac.uk/romeo/issn/1546-1998/
Related URLs:

Actions (login required)

View Item View Item
Statistics for RAIITH ePrint 4467 Statistics for this ePrint Item