An Improved 3-D Sound Source Localization of Varied Sources Using Oblique Square Pyramid Array

Sound source localization (SSL) is widely applied in noise control, fault localization, and explosives and weaponry localization in defense. Existing SSL methods have a tradeoff between affordability and accuracy. Moreover, they are time-consuming, require high signal-to-noise ratio (SNR), and do not converge in certain regions of 3-D space. This article presents an improved 3-D SSL method that identifies spatial coordinates of sound sources using an oblique square pyramid array of five microphones. Its novelty is: sensor array design, optimized wavelet denoising, a theoretical model that does not use projection angles and solution of nonconvergence. Simulations and experiments demonstrate that our method is more accurate than the existing time delay (TD) 3-D SSL method (error reduction 11.1%), existing numerical 3-D SSL method (error reduction 0.9%), our previous 3-D SSL method (error reduction 4.5%), and phase difference method (error reduction 8.8%). It accurately localizes multiple sound sources (8.9% error), works accurately (2.8% mean error) with SNR 15 dB onward, and converges in complete 3-D space. Wavelet function, wavelet decomposition level, and microphone spacing affect our method’s performance and must be optimized. This article notably advances the research status on SSL.