Dynamic vision sensor-oriented brain-like gesture sequence identification method

Abstract

The invention relates to a dynamic vision sensor-oriented brain-like gesture sequence identification method. The method comprises the following steps of address event expression data capture, space-time convolution, pulse pooling, space-time feature learning and gesture sequence identification by using a learned DoubleSTS model. The method has the characteristics of high noise robustness, high precision, high efficiency, rapid convergence, time sensitivity, brain imitation and the like, and each gesture in the gesture sequence can be successfully recognized.

Description

technical field [0001] The invention relates to a brain-like gesture sequence recognition method, in particular to a brain-like gesture sequence recognition method oriented to a dynamic visual sensor. Background technique [0002] In recent years, brain-like computing has gradually become a major research hotspot. In the field of neuromorphic vision, by simulating the biological retina, researchers have developed a series of neuromorphic vision sensors, also known as silicon eyes, event cameras, and more. Unlike traditional cameras that output frame images, neuromorphic vision sensors only respond to dynamic changes in the scene and encode light intensity changes into asynchronous spatiotemporal event stream data, which has the advantages of low power consumption, low latency, and high dynamic range. Therefore, most existing methods cannot be directly used to process this kind of event stream data. SNNs are expected to enable low-power asynchronous event information integr...

AI Technical Summary

Problems solved by technology

[0004] In recent years, biologically-inspired SNN-based supervised learning algorithms have emerged, which can be classified as threshold-based or membrane-potential-based plasticity rules. Although membrane-potential-based methods have shown better learning effects than threshold-based methods, such method still lacks biological interpretability, and the rule sometimes fails to find extreme points like that

Images