Розпізнавання сигналів за допомогою нейронних мереж

Abstract

Human sensory organs function as multi-component receptors that perceive images and generate multi-channel pulse signals, and recognizing them is crucial for the development of next-generation neural networks known as neuromorphic intelligent systems. Spiking Neural Networks (SNNs) are more effective because they can detect dynamic signals without converting them into feature vectors. This work presents a method for detecting multi-dimensional signals using SNNs, detailing their structure, models and training procedures. SNNs can be trained digitally or by special circuits that adjust synaptic weights without digital processors