An ICA Algorithm for Separation of Convolutive Mixture of Periodic Signals

EasyChair Preprint 346

Abstract

This paper describes Independent Component Analysis (ICA) based fixed-point algorithm for the blind separation of convolutive mixture of periodical signals. The proposed algorithm extracts independent periodical sources from their mixtures in frequency domain, where they are represented by their sets of harmonics. The individual harmonics are separated by referring to narrow frequency segments of the mixed signals, which include two harmonics each at most. The algorithm offers a simple solution to the permutation problem common to source separation using ICA.

Keyphrases: Independent Component Analysis, blind signal separation, convolutive mixture

@booklet{EasyChair:346,
  author    = {Doron Benzvi and Adam Shafir},
  title     = {An ICA Algorithm for Separation of Convolutive Mixture of Periodic Signals},
  doi       = {10.29007/8qk6},
  howpublished = {EasyChair Preprint 346},
  year      = {EasyChair, 2018}}