DOA (Direction Of Arrival) estimation algorithm of virtual array has been developed quickly for its low computation and high resource utilization. However, speech signal DOA estimation algorithm of virtual array is seldom reported. To improve the accuracy of speech signal DOA estimation, an improved DOA estimation algorithm of virtual array speech signal is presented in this paper. The first step of this method is combining the energy and entropy to detect speech. Then, select two speech frames from the detected signals following the minimum power spectrum variance criterion and the maximal spectrum entropy criterion, estimate DOA of the speech source with the two selected frames. In the end, conclude the final result by calculating the weighting average of the first two DOA estimation results. In this paper, DOA estimation results of variance selected frame, DOA estimation results of spectrum entropy selected frame and the weighting average of the first two results are presented and compared. The experimental results and analysis indicate that the method proposed in this paper has better performance.
