Neural Network Assisted Signal Classification for Integrated Sensing and Communication Systems

Sammanfattning

The Integrated Sensing and Communication (ISAC) system has emerged due to its advantages of efficient spectrum utilization and low hardware cost. In ISAC, wireless communication and radar sensing coexist, making the classification between communication and sensing signals crucial for improving the system’s efficiency, as it directly impacts the performance of ISAC systems and has not been adequately addressed in existing literature. In this thesis, we propose a Neural Network (NN) based framework to classify communication and sensing signals. The proposed framework is built on the mathematical observation that the Hankelization matrix of Orthogonal Frequency Division Multiplexing (OFDM) signals exhibits a low-rank property when the Signal-to-Noise Ratio (SNR) is infinite. By extension, we infer that the Hankelization matrix of a communication or sensing channel retains significant information even in real-world environments. Consequently, we designed our network to take the singular values of the Hankelization matrix as input and the one-hot coded vector as output. This design choice allows the network to function effectively with small input vectors and limited training dataset sizes. To validate our approach, we conducted extensive simulations. The results demonstrate that our network outperforms existing classification methods across various scenarios, highlighting its robustness and reliability. The key findings of this thesis include the successful classification of signals with higher accuracy and efficiency compared to traditional methods. The impact of this research is substantial, as it provides a novel solution for signal classification in ISAC systems. The enhanced classification capability can lead to more efficient spectrum utilization and improved performance of ISAC systems. Future research can build on this framework to explore other types of signal classification or improve the robustness of the network further.