<![CDATA[The escalating complexity and pervasive noise in contemporary communication systems have increasingly rendered traditional signal processing methods insufficient for reliable real-time analysis. This research addresses a fundamental void in existing literature by proposing a novel and lightweight deep learning framework, primarily centered on Convolutional Neural Networks (CNNs) for the joint classification and denoising of communication signals. Distinct from prior methodologies that often segregate these crucial tasks our model integrates both objectives within a highly optimized, unified architecture engineered for ultra-low-latency inference, notably achieving a 30–50% reduction in inference time compared to deeper CNN-RNN hybrids or Transformer-based architectures. The framework's effectiveness was comprehensively evaluated using both synthetic and real-world datasets, including RadioML2018.01A which encompasses a diverse range of modulation schemes and signal-to-noise ratio (SNR) levels. Experimental results conclusively demonstrate that the proposed CNN achieved an impressive 96.8% classification accuracy significantly enhanced signal quality to an average of 22.3 dB SNR, and maintained an average processing latency of merely 11.3 ms. These figures consistently demonstrate superior performance compared to traditional baselines including FFT, SVM, and LSTM. Despite these promising results, the current model was primarily trained and evaluated under Additive White Gaussian Noise (AWGN) conditions, and future work will explore its generalization to real-world scenarios involving multipath fading, Doppler shifts, and dynamic channel interference. This study represents a significant leap forward in developing robust, efficient, and intelligent solutions essential for next-generation communication signal processing, particularly for real-time applications in resource-constrained environments.]]>
