In order to utilize spatial feature information for network sorting of the synchronous frequency hopping signal, this paper proposes a FH signal DOA estimation algorithm based on STFD and unitary ESPRI. Firstly, the paper exchanges the received data for time frequency by the WVD and SPWVD combined timefrequency analysis method. Then the effective hop of the frequency hopping signals is extracted and the spacetime frequency matrix (STFD) is established. Finally, the frequency hopping signal DOA is estimated by using the unitary ESPRIT algorithm. This method transforms the covariance matrix from the complex field to the real domain by the unitary transformation to reduce the computational complexity and uses the conjugate information of the data to double the data length to improve the accuracy of the estimate. The simulation results show that the DOA estimation performance of this algorithm is prior to that of ESPRIT when SNR is higher than 2 dB.