Abstract: The cutoff distance of the traditional density peak clustering algorithm which is selected according to experience used to realize blind estimating of Walsh code soft spread spectrum signals results in unstable performance. For this problem, weighted Euclidean distance as data similarity and adaptive determination of cutoff distance according to the data distribution instead of artificial selecting is proposed. The improved algorithm has the characteristics of high efficiency and stable performance. The Matlab simulation results show that the proposed algorithm can automatically determine the cutoff distance and accurately estimate the pseudo-code sequence size and pseudo-code sequence when the signal-to-noise rate is 0-10 dB.

Key words: Walsh code, soft spread spectrum, blind estimation, improved density peak clustering