An Improved Intelligent Cognitive Radio Spectrum Sensing System using Concept Bottleneck and Deep Learning Models
Mohamed Diab, Zaki Nossair, Hesham Badr · 2023
This article investigates the use of concept bottleneck models in improving the performance of intelligent cognitive radio spectrum sensing systems. We propose a new explainable radio frequency intelligent system using concept bottleneck models that provide inherent decision explanations. In the context of cognitive radio spectrum sensing problems that only use raw IQ data as input, the proposed system has achieved a better level of performance than convolutional neural network algorithms in terms of both explainability and accuracy. Our proposed system has the potential to classify primary user existence that is not encountered during training, even in scenarios of low SNR levels for signals of interest.