The convergence of radar and communication systems in 6G networks creates a critical challenge: spectrum congestion. To address this, we propose a deep learning-based method for efficient waveform classification, crucial for resource-constrained cognitive radio-enabled Internet-of-Things (CR-IoT) devices. We introduce WaveNet, a cost-efficient deep convolutional neural network that effectively learns underlying radio features from time-frequency representations generated by a smooth pseudo Wigner-Ville distribution (SPWVD). WaveNet leverages innovative modules, notably Cost-Efficient Feature Awareness (CEFA). CEFA integrates two key structural blocks: grouped-of-kernel-wise residual connections and dual asymmetric channel attention. These modules significantly reduce network complexity while maintaining classification accuracy. Extensive simulations on an impaired signal dataset containing eight radar and communication waveform types demonstrate WaveNet's effectiveness and robustness.
The Matlab code and dataset provided here are included in the under-review paper at IEEE Internet of Things Journal
Thien Huynh-The, Van-Phuc Hoang, Jae-Woo Kim, Minh-Thanh Le, and Ming Zeng, "WaveNet: Towards Waveform Classification in Integrated Radar-Communication Systems with Improved Accuracy and Reduced Complexity," IEEE IoTJ, 2024.
The dataset can be download on Google Drive (please report if not available)
If there is any error or need to be discussed, please email to Thien Huynh-The via thienht@hcmute.edu.vn