The proliferation of unauthorized Unmanned Aerial Vehicles (UAVs) poses significant security risks, necessitating robust detection systems. However, in practical long-range surveillance scenarios, UAV signals often deteriorate due to severe attenuation and complex environmental interference, rendering traditional detection methods ineffective. To address this, this letter proposes a novel signal enhancement framework comprising an improved adaptive Wiener filter. We introduce a dynamic noise spectrum estimation strategy coupled with a decision-directed phase compensation mechanism. This approach effectively suppresses non-stationary background noise while reconstructing high-fidelity time-frequency features by rectifying phase distortions. Experimental results on a proprietary real-world dataset demonstrate that the proposed method significantly improves signal quality, enabling high-accuracy detection using YOLO models even in ultra-low Signal-to-Noise Ratio (SNR) regimes compared to raw data-based baselines.

Analog/mixed-signal (AMS) computation can be more energy efficient than digital approaches for deep learning inference, but incurs an accuracy penalty from precision loss. Prior AMS approaches focus on small networks/datasets, which can maintain accuracy even with 2b precision. We analyze applicability of AMS approaches to larger networks by proposing a generic AMS error model, implementing it in an existing training framework, and investigating its effect on ImageNet classification with ResNet-50. We demonstrate significant accuracy recovery by exposing the network to AMS error during retraining, and we show that batch normalization layers are responsible for this accuracy recovery. We also introduce an energy model to predict the requirements of high-accuracy AMS hardware running large networks and use it to show that for ADC-dominated designs, there is a direct tradeoff between energy efficiency and network accuracy. Our model predicts that achieving < 0.4% accuracy loss on ResNet-50 with AMS hardware requires a computation energy of at least ~300 fJ/MAC. Finally, we propose methods for improving the energy-accuracy tradeoff.