Abstract: The three-dimensional multiple signal classification (MUSIC) algorithm has the disadvantages of slow calculation speed and large calculation amount in estimating the position of sound source. In this paper, a three-dimensional algorithm for the localization of near-field sound source is proposed based on the chicken swarm optimization (CSO) algorithm. Firstly, the receiving model of near-field signal is established, and the spatial spectrum function in the MUSIC algorithm is selected as the fitness function. Through continuous iteration and local search, the chicken individual is sorted continuously with the fitness value as the index, and finally the position of the optimal chicken individual, which is the coordinates of the near-field sound source to be measured, is obtained. The results of simulation and experiment show that the proposed algorithm has high localization accuracy, high calculation efficiency and good real- time performance compared with the three-dimensional MUSIC algorithm; the average time for three-dimensional localization of sound source is only 1.9 % of that of the three-dimensional MUSIC algorithm in simulation and 3.2% in experiment.