Using Spiking Forward Propagation Through Time to Improve Sign-Language Recognition

This paper investigates the use of Spiking Forward Propagation Through Time (SFPTT) for sign-language recognition on the SL-Animals dataset. Sign-language recognition requires accurate modeling of spatial and temporal patterns, making it well suited for Spiking Neural Networks (SNNs), which naturally process sequential information through spike-based dynamics. Compared with conventional neural networks, SNNs offer sparse event-driven computation, lower memory usage, and potential energy efficiency. Experimental results show that SFPTT significantly outperforms previous SNN training methods, reaching 92.76% average validation accuracy and demonstrating strong potential for efficient real-time gesture recognition systems.