Exploring the Potential of Spiking Neural Networks in UWB Channel Estimation
Although existing deep learning-based Ultra-Wide Band (UWB) channel estimation methods achieve high accuracy, their computational intensity clashes sharply with the resource constraints of low-cost edge devices. Motivated by this, this letter explores the potential of Spiking Neural Networks (SNNs) for this task and develops a fully unsupervised SNN solution. To enable a comprehensive performance analysis, we devise an extensive set of comparative strategies and evaluate them on a compelling public benchmark. Experimental results show that our unsupervised approach still attains 80% test accuracy, on par with several supervised deep learning-based strategies. Moreover, compared with complex deep learning methods, our SNN implementation is inherently suited to neuromorphic deployment and offers a drastic reduction in model complexity, bringing significant advantages for future neuromorphic practice.