Communication through chaotic modeling of languages Murilo S. Baptista, 1,2,* Epaminondas Rosa, Jr., 3 and Celso Grebogi 1,2,4 1 Institute for Plasma Research, University of Maryland, College Park, Maryland 20742 2 Institute for Physical Science and Technology, University of Maryland, College Park, Maryland 20742 3 Nonlinear Dynamics Laboratory, Department of Physics, University of Miami, Coral Gables, Florida 33146 4 Department of Mathematics, University of Maryland, College Park, Maryland 20742 Received 2 March 1999 We propose a communication technique that uses modeling of language in the encoding-decoding process of message transmission. A temporal partition time-delay coarse graining of the phase space based on the symbol sequence statisticsis introduced with little if any intervention required for the targeting of the trajectory. Message transmission is performed by means of codeword, i.e., specific targeting instructions are sent to the receiver rather than the explicit message. This approach yields ierror correction availability for transmission in the presence of noise or dropouts, iitransmission in a compressed format, iiia high level of security against undesirable detection, and ivlanguage recognition. PACS numbers: 05.45.Vx, 05.45.Gg I. INTRODUCTION Recent developments in communicating with chaos 1–4 have produced a wealth of potential practical applications including synchronization 5–7, encoding-decoding tech- niques 1–4,8–11, noise filtering 12, and signal masking and recovery 13,14. This is so because chaotic systems have peculiar properties that make them natural candidates to play a significant role in nonlinear communication systems. One of these properties, the sensitivity of the dynamics to small perturbations, is useful for targeting the trajectory in phase space to specific regions to which particular symbols have been assigned. This targeting feasibility provides cha- otic systems with a natural type of dynamics to be used in communication. The symbol sequence to be followed by the chaotic trajectory corresponds then to the information to be transmitted 1–4,8,9,11,15. Indeed, the ergodicity or the eventual visit of the trajectory to all partitions without any targeting or controlof chaotic systems has been used re- cently 14in a chaotic communication scheme. Symbolization of a chaotic trajectory can be useful for extracting relevant information about the system under con- sideration. Correlation function computing 16,17, param- eter estimation 18, and data compression 19are examples of symbolic dynamics 20application toward a better under- standing of the system dynamics. Also, different signals gen- erated by the same dynamics can be identified with the help of the conditional entropy 21obtained from the symbolic dynamics of the chaotic process. Of course, the symbolic sequence generated by a chaotic trajectory depends on how the phase space is partitioned. It also depends on the time delay interval sampling rate for symbol sequence construc- tion, which has been used to measure correlation lengths from given symbolic sequences 19. Much emphasis has been placed on the characterization of the complexity of symbol sequences based on patterns and transmission rules estimated from symbolic time series 22. In this work we present a language approach for a chaotic communication system. The text message to be transmitted is generated by a chaotic process that respects the grammar of a language. Symbols are assigned to judiciously chosen re- gions of phase space, and the chaotic trajectory is controlled to visit these regions generating a symbol sequence that cor- responds to the desired message. The message itself is not transmitted. Rather, what is transmitted is a set of instruc- tions, the codeword, that enables the receiver to decode the message. A temporal partition is introduced as a time-delay coarse graining 19of the phase space. The phase space is divided into a number of cells to which different symbols are assigned 16,17. As the chaotic trajectory visits these re- gions, symbols are generated, producing a symbol sequence that corresponds to a message to be transmitted. The parti- tions are chosen in such a way that the message is consistent with the grammar of a language. For the purpose of illustra- tion we use an artificial language created as an approxima- tion to a real language in terms of statistical structure. We assume a communication system consisting basically of a transmitter where the message is encoded, a communication channel that carries the message from one place to another, and the receiver where the message is decoded. Transmitter and receiver have complete knowledge about the dynamical system being used. The procedure involves a minimum of information transmission, is secure against unwanted detec- tion, and is robust against noise and dropouts. This paper is organized as follows. In Sec. II we introduce concepts and definitions related to languages, paying special attention to their statistical structure. In Sec. III, we show how this statistical structure is used in the construction of the dynamical model process. Section IV details how the com- munication system is built based on language modeling, and a technique for optimal transmission of information is pre- sented in Sec. V. In Sec. VI, we introduce a language recog- nition scheme and explain how this proposed communication system is secure against undesired decoding. Section VII proposes an error correcting code that is able to recover in- formation when the transmission is corrupted by noise or lost *Permanent address: Instituto de Fı ´sica, Universidade de Sa ˜o Paulo, C.P. 66318, 05315-970 Sa ˜o Paulo, SP, Brazil. PHYSICAL REVIEW E APRIL 2000 VOLUME 61, NUMBER 4 PRE 61 1063-651X/2000/614/359011/$15.00 3590 © 2000 The American Physical Society