As electromagnetic waves fade rapidly when passing through the human body, interferences may significantly degrade communication over wireless body area networks (WBANs). However, due to the limited power and resource restrictions in WBAN devices, existing interference mitigation methods may not fully address this issue. In this study, considering the architecture and equipment capacity of WBANs, the clustering mechanism is improved using unsupervised machine learning during the coloring process. Accordingly, a novel unsupervised coloring algorithm (UCA) for spectrum allocation in multiple WBANs is presented in this paper. The nodes in the region are separated by clustering, and the channel group is assigned to the cells by coloring to achieve UCA. We simulated the performance of UCA in different scenarios with different node densities for both indoor and outdoor environments. The experimental results showed that the proposed UCA achieved convergence results with accurate and inaccurate positioning. It outperformed the existing methods by exhibiting less time complexity, lower interference strength, and faster algorithm convergence. Moreover, the proposed UCA was robust against topology changes.

由于电磁波在穿过人体时会迅速消失，因此干扰可能会严重降低无线人体局域网（WBAN）上的通信。但是，由于WBAN设备中有限的功率和资源限制，现有的干扰缓解方法可能无法完全解决此问题。在这项研究中，考虑到WBAN的体系结构和设备容量，在着色过程中使用无监督机器学习来改进聚类机制。因此，本文提出了一种新颖的无监督着色算法（UCA），用于在多个WBAN中进行频谱分配。该区域中的节点通过聚类分离，并且通过着色将通道组分配给单元以实现UCA。我们针对室内和室外环境在不同节点密度的情况下，模拟了UCA的性能。实验结果表明，所提出的UCA实现了收敛，定位准确，不准确。它具有更少的时间复杂度，更低的干扰强度和更快的算法收敛性，从而优于现有方法。此外，提出的UCA对于拓扑更改具有鲁棒性。