Towards optimal spatiotemporal wavefront shaping for the cocktail party problem with inverse design of an acoustic reconfigurable metasurface in disordered media

Multiple-user multiple-input multiple-output applications have recently gained a lot of attention. Here, we show an efficient optimization formulation for the design of all the temporal and spatial degrees of freedom of an acoustic reconfigurable metasurface for the cocktail party problem. In the frequency domain, the closed-form least square solution matches the optimal time reversal solution for multiple emitter-receiver pairs, optimizing for each frequency independently. This is more efficient than solving in the time domain where the time convolution mixes all the degrees of freedom into a resource-intensive optimization. We illustrate this methodology by optimizing the frequency response of a design for two pairs of emitters-receivers using the Green’s functions of disordered media that are measured experimentally. We report strong performance that will be put in perspective in future work, where we will analyze the robustness of the design to noise in the data and design the convolutional filters that match the optimal frequency response for experiment validation of the design.