A Real Time Method Based on Deep Learning for Reconstructing Holographic Acoustic Fields from Phased Transducer...

Area of Science:

• Acoustics and Ultrasonics
• Signal Processing
• Machine Learning

Background:

• Phased transducer arrays (PTA) generate holographic acoustic fields by controlling ultrasonic waves.
• Reconstructing PTA phase from a holographic acoustic field is a complex inverse problem.
• Existing iterative methods are time-consuming and mathematically challenging.

Purpose of the Study:

• To develop a novel, efficient deep learning method for reconstructing holographic acoustic fields from PTA.
• To address the challenges of focal point distribution imbalance and randomness in holographic sound fields.

Main Methods:

• A deep learning approach was employed to reconstruct the holographic sound field from PTA.
• A novel neural network structure with attention mechanisms was designed to handle focal point distribution.
• The network learns to determine the transducer phase distribution required for generating the target acoustic field.

Main Results:

• The deep learning method successfully reconstructs the holographic sound field with high efficiency and quality.
• The determined transducer phase distributions accurately enable PTA to generate the desired holographic acoustic fields.
• The proposed method demonstrates real-time performance and higher accuracy compared to existing AcousNet methods.

Conclusions:

• Deep learning offers a powerful and efficient solution for the inverse problem of holographic acoustic field reconstruction.
• The attention-based neural network effectively manages focal point distribution complexities.
• This approach provides a significant advancement over traditional iterative methods, enabling real-time applications.