Open Access Open Access  Restricted Access Subscription Access
Cover Image

Design of Hardware Mixing Model for ICA Algorithms Using Wireless MIMO System

Zahoor Uddin, Farooq Alam, Muhammad Altaf, Muhammad Uzair, Wilayat Khan


Independent component analysis (ICA) is a signal processing technique used for un-mixing of the recorded mixed signals. Un-mixing is based on the fact that the ICA algorithms must know the mixing model in advance i.e., how the signals are mixed. In the literature, different mixing models are presented for the ICA algorithms in digital domain, while the mixing phenomena normally occur in analog domain. The existing mixing models include linear mixing, non-linear mixing, instantaneous mixing, convolutive mixing and over complete mixing. Various algorithms are presented for these mixing models because the algorithm basically explores the mixing model and blindly un-mix the mixed data based on this information. We consider the instantaneous mixing model in this paper and propose a hardware design of the mixing model for ICA algorithms. The proposed model represents unity mixing, scaled mixing and ill-conditioned mixing. These mixing models are basically different conditions of the instantaneous mixing model. It is hoped that this model will serve as a test bench for evaluating the performance of the ICA algorithms. We simulate the proposed model using 16-QAM signals for all the three mixing conditions. The FastICA and the OBAICA algorithms are used to evaluate the proposed mixing models for un-mixing. Results obtained are such that the unity mixing provides excellent performance of the algorithms, while the scaled mixing provides good performance and the ill-conditioned mixing gives worse performance of the algorithms. Moreover, to the best of our knowledge, we are the first to present the hardware mixing model for ICA algorithms.

Full Text:



  • There are currently no refbacks.