Selection of Embedding Parameters in Phase Space Reconstruction

A joint algorithm is proposed for computing embedding dimension and delay time in phase space reconstruction. This algorithm based on delay time window method and mutual information method takes delay time window length into consideration, the delay time and embedding dimension can be obtained simultaneity. In simulation, by using the extremely simple noise reduction method, this algorithm has good results in phase space reconstruction from noisy signals. The simulated experimental results indicate that the relative error of the correlation dimension of Lorenz chaotic time series is decreased from 0.83% by the conventional algorithm to 0.44% by using this algorithm.