An Approach to Microphone Array Geometry Transform

Microphone array signal processing is significantly influenced by array parameters such as the number of sensors, array aperture, and array topology. These factors affect both the development of algorithms and their performance limitations. Consequently, applying algorithms designed for one array geometry to signals captured by different geometries can be challenging. To address this issue, we introduce a novel method referred to as the microphone array geometry transform, or simply the microphone array transform. This approach involves two key steps: 1) encoding the array observations in the frequency and angle domains, and 2) generating observations for the target array using array manifold vectors and the frequency-angle domain source signals. We formulate this problem as one of convex optimization and solve it using the well-established alternating direction method of multipliers (ADMM). Simulation results justify the effectiveness of the proposed method.