Comparative analysis and selection of the geometry of the microphone array based on MEMS microphones for sound localisation

The subject of this article is the design and optimization of the geometric configuration of omnidirectional MEMS microphone arrays for sound localization tasks. The goal is to determine the most effective array architecture and beamforming algorithms to achieve compactness, accuracy, and balanced omnidirectional coverage. The tasks to be addressed include analyzing spatial-frequency characteristics of various microphone array geometries (Uniform Linear Array, Uniform Planar Array, Uniform Circular Array, and Uniform Concentric Array), comparing beamforming algorithms (delay-and-sum, differential, and superdirective), and evaluating their performance under isotropic noise fields and coherent noise sources. The methods used involve the application of both established and author-derived analytical models for transfer functions and directivity coefficients, as well as experimental validation using a prototype device built on a Raspberry Pi 5 platform with an Adafruit PCA9548 8-Channel STEMMA QT expansion board and SPH0645LM4H-B omnidirectional MEMS microphones. The results show that similar geometric configurations of microphone arrays from omnidirectional microphones can be used for sound localization tasks at low frequencies because they are characterized by good values of Array Directivity and HPBW. This means creating a sufficiently narrow main beam, where the level of the sidelobe SLL does not differ from that of the main lobe at high frequencies. The best configurations were URA Microphone Arrays with n = 8 and d = 23 cm. Conclusions. Differential beamforming algorithms have demonstrated superior performance in isolation of target signals in challenging acoustic environments. The Uniform Circular Array (UCA) combined with DAS or EF DAS algorithms provides reliable omnidirectional coverage and balanced frequency response, making it ideal for applications requiring uniform sensitivity. Optimizing the spacing and radius of the microphone arrays further enhances directivity and minimizes sidelobe levels. In future work, we will focus on improving array designs using SSL reduction methods to expand localization accuracy across a wider frequency range.