Action Recognition Using Supervised Spiking Neural Networks

TL;DR

This paper demonstrates a supervised spiking neural network that effectively recognizes human hand gestures from DVS camera data, achieving high accuracy by employing surrogate gradient descent.

Contribution

It introduces a novel application of surrogate gradient descent for training SNNs on dynamic gesture recognition tasks, bridging the gap between biological plausibility and practical performance.

Findings

Achieved 97.2% accuracy on gesture recognition

Successfully applied surrogate gradients to SNNs for dynamic data

Demonstrated energy-efficient processing of temporally dynamic inputs

Abstract

Biological neurons use spikes to process and learn temporally dynamic inputs in an energy and computationally efficient way. However, applying the state-of-the-art gradient-based supervised algorithms to spiking neural networks (SNN) is a challenge due to the non-differentiability of the activation function of spiking neurons. Employing surrogate gradients is one of the main solutions to overcome this challenge. Although SNNs naturally work in the temporal domain, recent studies have focused on developing SNNs to solve static image categorization tasks. In this paper, we employ a surrogate gradient descent learning algorithm to recognize twelve human hand gestures recorded by dynamic vision sensor (DVS) cameras. The proposed SNN could reach 97.2% recognition accuracy on test data.