ASIE: An Asynchronous SNN Inference Engine for AER Events Processing

Neuromorphic computing based on spiking neural network (SNN) shows good energy-efficiency. However, the conventional convolution operation required in processing the output of the frame-based vision sensor makes SNN inefficient because the convolution requires full-image processing. Frameless vision sensors (also called event-based vision sensors) can address this problem because they produce spikes only when events occur in the image. The image encoding of frameless vision sensors is Address Event Representation (AER), an emerging asynchronous protocol. AER events are particularly well-suited for SNN processing. Indeed, computation and communication in SNN are asynchronous because of the asynchronous nature of AER events. This paper presents an SNN hardware inference engine based on an asynchronous Processing Element (PE) array with AER events as input. The engine uses a convolution algorithm based on AER events. This design also uses distributed storage in the PE array to reduce the cost of memory access. The experimental results show that the design can achieve a recognition accuracy of 97.5% for the MNIST AER dataset. This design reduces by 9 the computing activity compared to a baseline systolic array based accelerator.