Abstract: In order to solve the problem of carrier frequency blind estimation of PSK signals in electronic reconnaissance, a new estimation method was proposed. The phase shift keying(PSK) signal was divided into several overlapping intervals which had equal length, and the spectrum concentration measures of every interval were extracted by the FFT. And then, using the grid-density clustering, the spectrum concentration measures were classified into two categories: the narrowband spectrum interval and the wideband spectrum interval. The narrowband spectrum interval was regarded as the characteristic class. The spectrums of the characteristic class were accumulated to estimate the carrier frequency of PSK signal. The proposed method had avoided the non-linear operation in the traditional PSK signal carrier frequency estimation algorithm. Thus, the signal to noise ratio (SNR) threshold was remarkably decreased. Moreover, the proposed method did not need the prior knowledge of the signal, which was suitable to the electronic reconnaissance occasion. Experimental results had verified the validity of the proposed estimation method in low SNR.