关键词: 1/f噪声, 最小递归二乘法（RLS）, 混沌, 信号跟踪

Abstract: A simple improved RLS algorithm is given to track a discrete-time noised 1/f signal. Formal RLS algorithm or rapid computing RLS algorithm in the limited accuracy under the conditions of disorder and convergence is no essential difference, they are limited in the number of iterations and the RLS filter will lead to weight divergence, particularly in tracking non-stationary signals even more so. In view of this, this paper introduces a non-linear function of RLS filter input data from the inverse correlation matrix to be amended. Experiments show that the algorithm has a good track non-stationary signals as well as the characteristics of chaos with the 1/f noise of the signal capabilities. It can effectively reduce the tracking error, as well as the average variance, and can input data in accordance with the rapid changes in the adjustment of filter coefficients. Performance is better than the regular RLS algorithm. FBm noise for tracking the dynamic process of adjustment factor memory, this paper derives an expression of relation between memory factor and input signal autocorrelation matrix, providing theoretical basis for the RLS algorithm which is used to dynamically adjust the memory factor to track the fBm process.

Key words: 1/f noise, Recursive Least Squares（RLS） algorithm, chaos, signal tracking