International Journal of Computer Science and Telecommunications [Volume 4, Issue 8, August 2013] 18 Journal Homepage: www.ijcst.org Akingbade Kayode Francis 1 and Michael O. Kolawole 2 1,2 Department of Electrical & Electronics Engineering, The Federal University of Technology, Akure, Ondo State, Nigeria 1 kfakingbade@futa.edu.ng, 2 mokolawole@futa.edu.ng Abstract– This paper presents a simple method that dealt with independent component analysis and Finite Impulse Response (FIR) filter in separating convolutive mixtures. The original source was retrieved from the set of filtered versions of each mixed signals using independent component analysis method as well as filtering mixture of voices (audio) recorded in a noisy environment. Index Terms– Independent Component Analysis, Adaptive Filtering, Convolutive Mixture and Audio Signal I. INTRODUCTION IGITAL Signal Processing (DSP) is concerned with the theoretical and practical aspects of representing information bearing signals in digital form and with using computers or special purpose digital hardware either to extract information or to transform the signals in useful ways [1]. At achieving this, digital filters remain as the backbone for digital signal processing. The available types of digital filters are: Infinite Impulse Response (IIR) filters; and Finite Impulse Response (FIR) filters. An improvement on these resulted into the use of the digital filters at making adaptive filters where Finite Impulse Response filters remains the most used in this application because of its stability and short length of convergence [2]. They can be implemented using either of the two available types of digital filters i.e. the Infinite Impulse Response (IIR) filter or the Finite Impulse Response (FIR). However, the FIR filter is preferred for the implementation of the adaptive filters because they are more stable than the IIR filters and their convergence is achieved faster than that of the IIR filters [3]. Some of the adaptive filter performs its task using correlation principle mainly cross correlation. It is this method of adaptive filter filtering using cross correlation method to achieve signal separation coupled with using Least Mean Square adaptive algorithm that is employed in this paper to separate to mixed digital audio signals. II. PROPOSED ICA MODEL The proposed ICA model is shown in Fig. 1 and the optimization criterion is in general taken in the least squares family in order to work with linear operations [4]. By applying the adaptive filter coefficients, the general LMS is capable of removing noise or obtaining a desired signal [5]. Where β i and h (Fig. 1) are respectively the Adaptive filter weight and the Adaptive algorithm. Let S be time t indexed, k-dimensional independent signals from the linearly mixed observable variables. The ICA model is written as: ) ( ˆ ) ( ) ( ) ( ˆ t N t N t S t S o    (1) ) ( ˆ ) ( t S A t U o  (2) ) ( ˆ ) ( ) ( ˆ ) ( t V t N t S t X o    (3) ) ( ) ( t BX t Y  (4) Fig. 1: Adaptive ICA process model D Separation of Mixed Audio Signals Using Independent Component Analysis ISSN 2047-3338 S(t) N(t) Ŝ o (t) A U ICA X(t) h β i Ñ i (t) Σ     ε(t) Y(t) Σ B