Our research about Energy Efficient Hand Gesture Recognition by Event camera + SNN presented at ITS European Congress 2020
As the title explained, we realized hand gesture recognition on event camera data by using Spiking Neural Networks (SNN). The point of this research is to use a method to convert from trained Deep Neural Networks (DNN) including Convolutional Neural Networks (CNN) to SNN. In general, it is difficult to apply Back Propagation (BP), which is one of the most common Neural Networks training method, for SNN because of its asynchronous signal propagation. However, thanks to the SNN conversion method, we can train CNN by using BP, then convert the trained CNN to SNN for inference. In this research, we showed a converted SNN for gesture recognition with low energy consumption, at the same time with keeping same level of recognition performance as CNN.