c  2007 Nonlinear Phenomena in Complex Systems Independent Component Analysis of Mixed Chaotic Signals for Communications Systems K. Umeno ∗ National Institute of Information and Communications Technology, 4-2-1 Nukui-kitamachi, Koganei, Tokyo, 184-8795, JAPAN (Received 10 February 2006) We consider the blind source separation of the mixed chaotic sequences and propose a new type of chaos-based communications systems where the signal separation at receiver is achieved purely by means of Independent Component Analysis (ICA). It is shown that chaotic sequences that can be identified with bit sequences due to symbolic dynamics cor- respondence can be cleanly separated without any knowledge of the decoding sequences as well as the mixing matrix. PACS numbers: 84.40.Ua Keywords: independent component analysis, chaotic communications 1. Introduction Code Division Multiple Access (CDMA) is a currently used data transmission technique for the third generation mobile communication sys- tem. In CDMA, each user is assigned a particular spreading code-sequence and the multiple access capability is due to this coding. Since all the users simultaneously access the same frequency band, the multiple access interference, namely, the in- terference noise from other users, must be limited. To this end, orthogonal spreading codes are pri- marily necessary and binary pseudo noise (PN) sequences such as Gold codes are practically used. The chaos-based CDMA is a promising al- ternative of the conventional CDMA where the spreading code sequences are the chaotic se- quences. Besides the communication security, the potential advantage of chaos-generated pseudo se- quences (PN) over the conventional PN sequences such as Gold codes are: (1) due to the ergodic the- ory, analytical estimation of system performance is possible. (2) infinitely many different chaotic sequences that are orthogonal are available by changing the initial condition of dynamical sys- tems. The second advantage may allow more ac- cessible users. In fact, it is shown that the num- * E-mail: umeno@nict.go.jp ber of simultaneously accessible users of asyn- chronous CDMA communication system can be improved by 15% if the digital filter constructed by Lesbegue spectral analysis of ergodic theory is incorporated into the system [1]. Although the chaos-based CDMA has al- ready been proven to be a practical communi- cation system with advantages, it would be inter- esting to apply other signal separation techniques instead of the widely used despreading procedure as the latter is time-demanding and the improve- ment of communication speed is one of the re- quired conditions for the fourth generation mo- bile communication system. In this paper, we apply as such an attempt Independent Compo- nent Analysis (ICA) [2] to chaos-based CDMA for signal separation at receiver for communica- tions systems, and analyze the blind source sep- aration of the mixed chaotic signals. The ICA for the CDMA system with Gold codes has been investigated in the literature [5]. However, be- sides our main purpose, the ICA of the mixed chaotic signals seems interesting in its own right because the analytical relations of chaotic signals are known, and thus they may be useful to test the ICA algorithm itself. Nonlinear Phenomena in Complex Systems, 10:2 (2007) 170 - 175 170