Hybrid deep learning based spectrum sharing strategy for cognitive radio based IoT in 5G environment
Jayesh Kumar Dabi, Priyadarshi Ashok Dahat · 2025
Spectrum sensing (SS) is essential to the cognitive radio network (CRN), which will be used in the next generation of wireless communication systems. Various spectrum sensing techniques have been proposed over the last decade, but there are many challenges associated to these strategies. Some techniques fail miserably when Signal-to-noise ratio (SNR) is changing as they necessitate previous knowledge of the PU signals. Due to their reliance on thresholds under particular signal-noise model assumptions, the detection success of these approaches is totally dependent on the sensor&s;s accuracy. Thus, among the most coveted objectives for wireless researchers is the development of an intelligent and reliable spectrum detecting system. When dealing with time-series data, however, multilayer learning models are not optimal because of the high computational cost and high rate of misclassification they produce. The authors propose a hybrid model that integrates Extreme Learning Models (ELM) and Convolutional Neural Network (CNN) for learning spectral features of time and other environmental activity statistics such as distance, energy, and duty cycle duration in order to enhance sensing performance.