An Effective Wireless Communication Architecture Using Ultra Lightweight Automatic Modulation Classification Model Over Cortex M4 Based Edge Device

The future advancement in computation, IoT and ultra-edge computing devices will require support of efficient communication methods. The presented research focuses on two parts. The first one focuses on a novel automatic modulation classification (AMC) technique developed using 1D-Convolutional Neural Network useful for STM32L4 series. The approach effectively balances performance with the stringent energy and computational constraints of the ultra-edge devices. The approached has shown 95.42% reduction in Memory requirement while achieving high classification accuracy of 79.01% along with having 99.63% at high SNR (10dB to 20dB), 92.90% at medium SNR (0dB to 10dB), and 44.78% at low SNR (-10dB to 0dB). The second part highlights a combination of multiple modulation schemes in a sequential manner, which is improvising the security of wireless communication. The model also demonstrates robust performance against noise and other channel impairments. This research provides a scalable, efficient AMC solution for ultra-edge devices, enhancing the practicality and security of intelligent communication systems. By optimising for the STM32L4 platform, this work represents a significant advancement in deploying advanced signal processing techniques on resource constrained devices.