Threshold memristor-based inverse radial basis neuron for motion detection

Visual motion detection plays a crucial role in computer vision applications including human behavior understanding and traffic monitoring. Traditional motion detection methods lack the selective response mechanism of biological visual neurons and cannot effectively distinguish between moving objects and background information. This paper proposes an inverse radial basis neuron (IRBN) circuit based on a threshold-switched memristor that realizes ``near-suppression far-enhancement'' response characteristics opposite to conventional radial basis functions. The voltage-frequency responses of IRBN can be modulated through adjustment of the memristor dielectric constant and circuit parameters. Application of IRBN characteristics to gait energy image (GEI) processing achieves suppression of low-grayscale background regions and enhancement of high-grayscale motion silhouettes with an edge preservation index of 9.466, which is 2 to 30 times higher than traditional image processing methods. This work provides a new implementation approach for biomimetic motion vision detection.