For the situation where the current signal extraction algorithms for the photon Doppler velocimetry (PDV) data are weak in the performance of anti-noise, a method based on trajectory search with the short time Fourier transformation (STFT) spectrum of the PDV data is proposed to extract the valid target signal. The Algorithm uses the iterative MeanShift progress weighting the search direction to get the maximum value points of the density function and these points can be regarded as the trajectory points of valid signal. Experimental results show that, when the proper initial point is given, the algorithm can accurately extract the valid signal from the short-time Fourier transform spectrum which contains a large amount of noise and interference signal, this signal extraction algorithm for PDV has strong robustness in strong noise condition.