Delay and Sum Beamforming（DSB） is widely used in angle of arrival estimation of microphone array signals. However, this method is not ideal for azimuth estimation of multiple speech signal sources due to the problem of grid lobe in speech signal sources. In addition, in the actual complex environment, it is affected by noise and reverberation, which makes azimuth recognition more difficult. In order to solve these problems, an improved DSB method is proposed, which combines signal frequency and microphone array spacing to select the frequency points in the sub-segment, and then weighs the data covariance matrix. At the same time, experiments are carried out in the simulation and actual environment, and the results show that, compared with the unimproved DSB method, the computational complexity of this method is reduced to 18.37%, and the amount of computation is effectively reduced. In the simulation experiment, the average angle positioning deviation is reduced by 27.3%, 21.4% and 36%, respectively, under different reflection coefficients of 0.2, 0.4 and 0.6, respectively. In the actual environmental experiment, the maximum azimuth angle estimation deviation is 9° and the minimum azimuth angle estimation deviation is 1.35°, which is lower than 12.1° and 3° of the unimproved algorithm.

%U http://cea.ceaj.org/EN/10.3778/j.issn.1002-8331.1910-0123