Radar Angular Resolution via Neural Network Signal Processing

The patent replaces the mechanical/physical approach of using massive antenna arrays with a computational/neural network-based approach. The neural network processes reflected signals to generate magnitude and angle images, achieving high angular resolution through algorithmic processing rather than physical antenna expansion. This substitution resolves the contradiction by achieving improved measurement precision without increasing device complexity.

The patent uses neural network models to create virtual representations of the radar imaging process. By training the neural network on reflected signals and reference signals, the system creates a computational model that can generate accurate magnitude and angle images without requiring physical duplication of antenna elements. This copying approach achieves high resolution through software-based signal processing.

The patent substitutes physical antenna expansion with computational processing. Instead of increasing the physical volume of the radar system to improve angular resolution, the invention uses neural network-based signal processing to achieve the same resolution improvement, thereby maintaining a compact radar size while enhancing measurement precision.

The patent transitions from a spatial dimension solution (physically larger antenna array) to a computational dimension solution (neural network processing). By moving the resolution enhancement from the physical domain to the computational domain, the system achieves high angular resolution without increasing the radar's physical volume.