Modulation Classification Through Convolutional Spiking Neural Networks with Data Fusion

As the volume of information shared on the Internet continues to surge, the communication industry is actively exploring methods to optimize spectrum utilization. Furthermore, the wireless communication sector is evolving to align with 5G/6G and advanced standards for space communication, addressing the escalating data demands from applications like autonomous vehicles, IoT, and intelligent systems. The integration of Software Defined Radio and Cognitive Radio opens new applications of Artificial Intelligence in the communication sector, enabling tasks such as spectrum monitoring, dynamic spectrum access, adaptive resource allocation, and more. A critical task within this domain is automatic modulation classification, crucial for effective spectrum monitoring and management. Presently, Artificial Neural Network (ANN)-based deep learning models leverage high-performance clusters and graphics processing units to classify modulation types swiftly and with enhanced accuracy, despite increased power consumption. However, it is worth noting that many classification applications run on power-constrained devices. To address this challenge, we investigate the application of low-power spiking neurons in the deep learning architecture.