Reconfigurable phase-only acoustic holography with surrogate model based on soft-GAN

Abstract

Acoustic holography (AH) has shown great potential for application in various fields, including biomedical, lab-on-a-chip, and industrial scenarios, due to its advantages of high penetration ability, versatility in operating mediums, low energy consumption, and excellent biocompatibility. In this paper, we propose a phase-only acoustic holography (POAH) algorithm with a novel surrogate model based on soft-GAN to dynamically construct reconfigurable acoustic field patterns. Specifically, the acoustic field is encoded onto the phase-only profile of a phased array of transducers (PAT), enabling high-fidelity acoustic field reconstruction. Benefiting from the concept of physics-based deep learning (PBDL), we incorporate a well-defined physics rule with a generative learning method, thereby eliminating the need for an annotated POAH dataset. Moreover, a soft evaluation scheme is proposed to enhance the algorithm’s flexible exploration capability. Experimental results on various acoustic field patterns demonstrate the effectiveness of the proposed algorithm, showing superior accuracy and real-time performance in constructing acoustic holograms. These results suggest significant potential for precise and contactless physical applications.