A Modified SSA Function for Real-Time Sound Source Localization

PurposeThe accuracy of real-time sound source localization is an important issue in acoustics. The steered sample algorithm (SSA), an algorithm developed based on the reciprocity of wave propagation, has a higher spatial resolution than the steered response power - phase transform (SRP-PHAT) algorithm. The algorithm can also render an accurate sound source localization under limited array elements and low signal-to-noise ratio. However, the actual implementation of the algorithm is usually based on an expensive grid search process, which makes the computational cost a serious problem.MethodsAn improved implementing algorithm based on beetle swarm optimization (BSO) is proposed for SSA, which can effectively reduce the computational cost. The modified algorithm has a good convergence speed as the SSA algorithm has fewer local extremums.ResultsExperimental results demonstrate that compared with steered sample algorithm (SSA), the proposed algorithm has almost the same localization performance and robustness with lower computational cost.ConclusionThis paper proposes an improved algorithm about SSA algorithm. Compared with SRC algorithm, the amount of computation is reduced more than twice. Meanwhile, the proposed algorithm inherits the anti-noise performance of SSA. Under the condition of low SNR, the positioning success rate and RMSE performance are excellent. Under the condition of high reverberation, the improved algorithm needs more particles to ensure the positioning performance. When the number of particles is not less than 70, the localization success rate of the proposed algorithm is highly consistent with the conventional SSA, and the RMSE is slightly less than the conventional SSA. Compared with SRC algorithm, the proposed algorithm better inherits the robustness of original SSA and improves the positioning accuracy.

Automated summary

This paper proposes an improved algorithm for the SSA algorithm. Compared with the SRC algorithm, the computational cost is reduced by more than twice. Meanwhile, the proposed algorithm inherits the anti-noise performance of SSA.