Using the methods of the wavelet transform and the nonlinear dynamics, the b ehavior of chaotic signals in phase space is studied. It is indicated that, in p hase space reconstruction, the wavelet transform of chaotic time series is essen tia lly a projection of strange attractor on the axis of the space that filter vect ors opened, which in correspondance with the method of phase space reconstructio n proposed by Packard et al. The experimental results show that, after d oing wavelet transform, the architecture of attractor trajectory is similar to t he original one, and the nonlinear invariants such as correlation dimension and Kolmogorov entropy are reserved. These results show that wavelet transform is ef fective for studying chaotic signal.