Chaotic modulation has recently been proposed as an alternative to the conventional spread spectrum and code division multiple access system. It uses a chaotic dynamical system to modulate the signal of transmission by embedding it in the bifurcating parameter. The transmitted signal then occupies a wide bandwidth as desired. One advantage of this chaotic modulation communication is that it does not require synchronization which is a complicated procedure for a conventional SS system. The main difficulty of the chaotic modulation technique is to design a reliable receiver so that the signal of transmission can be demodulated after passing through a noisy channel. In this paper, adaptive filters are used for chaotic demodulation, that is, to estimate the signal of transmission in an on-line fashion. The two widely used adaptive filtering algorithms: least mean square (LMS) and recursive least square (RLS), are considered. Based on simulation, both LMS and RLS receivers are demonstrated to be more accurate than the inversion approach in the noisy environment.