Optimal irregular microphone distributions with enhanced beamforming performance in immersive environments

Abstract

Complex relationships between array gain patterns and microphone distributions limit the application of optimization algorithms on irregular arrays. This paper proposes a Genetic Algorithm (GA) for microphone array optimization in immersive (near-field) environments. Geometric descriptors for irregular arrays are proposed for use as objective functions to reduce optimization time by circumventing the need for direct array gain computations. In addition, probabilistic descriptions of acoustic scenes are introduced for incorporating prior knowledge of the source distribution. To verify the effectiveness of the proposed optimization, signal-to-noise ratios are compared for GA-optimized arrays, regular arrays, and arrays optimized through direct exhaustive simulations. Results show enhancements for GA-optimized arrays over arbitrary randomly generated arrays and regular arrays, especially at low microphone densities where placement becomes critical. Design parameters for the GA are identified for improving optimization robustness for different applications. The rapid convergence and acceptable processing times observed during the experiments establish the feasibility of this approach for optimizing array geometries in immersive environments where rapid deployment is required with limited knowledge of the acoustic scene, such as in mobile platforms and audio surveillance applications.

Computational strategy for accelerating robust sound source detection in dynamic scenes

“…Simulated array recordings allowed for testing performance over a broad range of source placements and conditions. Details of the simulator are described in [7,[15][16][17]. …”

Stochastic allocation strategy for irregular arrays based on geometric feature control

“…Figure 2 gives the optimal arrays resulted from computer-aided heuristic searching 19 and hyperbola cluster design method. The hyperbola areas are marked by dashed lines with different colours.…”

Optimization for randomly described arrays based on geometry descriptors

“… G , r i ,r s  is the relationship functions between geometric descriptors and performance matrices through nonlinear regression analysis and Monte Carlo simulations [14]. For a given focal point ri and noise source at rs ,…”

“…In our experiments, the size of initial population, ratio of elites selection, ratio of mutation and crossover, and standard deviation of perturbation in mutation are the relevant factors which need to be adjusted to maintain robust performance of optimization procedure in specified scenes. The details of GA setting and objective functions have been provided in paper [5][6][7]14].…”