With the rapid development of the retail industry, enhancing customer experience and operational efficiency has become increasingly critical, where technological integration is key. This study introduces an innovative framework for human behavior recognition that combines GNN and RFID technology. By embedding RFID signals into the graph structure, we effectively capture the spatial dependencies hidden in the data. Furthermore, the Spline Convolution technique is utilized to address the spatial dependencies of the signals, achieving accurate and robust human behavior recognition. Facing challenges such as dynamic changes in data dimensions in the retail environment, over-smoothing issues in graph neural networks, and the effective fusion of multi-dimensional features, we adopted a graphbased modeling approach. We constructed an adjacency matrix with small-world characteristics using the TopK mechanism and Pearson correlation coefficients, and introduced Inception structures and residual connections to increase network width, thereby mitigating over-smoothing phenomena. The introduction of BiLSTM readout methods further enhanced the model's ability to process time series information. Experimental results demonstrate the framework's excellent performance in human behavior recognition tasks, with high accuracy and strong robustness, proving not only theoretically feasible but also highly effective in practical applications. Through qualitative analysis, we have improved the interpretability of the framework, providing retailers with a powerful tool for gaining in-depth insights into customer behavior, which helps to optimize customer experience and enhance operational efficiency.

This paper delves into the inherent graphtopological structure utilized by algorithms such as LandMark in the context of RFID indoor positioning. We uncover the essence of these algorithms, which leverage the implicit topological relationships within signal features for message passing and positioning accuracy. Despite the theoretical advantages, practical applications of the LandMark algorithm have been hindered by issues related to signal propagation, the limitations of topological structures, neighbor tag selection, and simplistic weight distribution methods. To address these limitations, we propose a series of innovative improvements. Our approach includes data preprocessing techniques like B-spline interpolation and normalization to mitigate environmental noise and enhance signal integrity. We introduce the concept of spatiotemporal graphs that map signals into a high-dimensional space, allowing for the construction of dynamic graph structures that more accurately capture the temporal dynamics of signals. Furthermore, we employ the PNAConv algorithm, a graph neural network technique, to refine the message passing and feature aggregation process, optimizing the selection of neighboring tags. Our experiments, conducted across various datasets, demonstrate that our model maintains low error rates, showcasing its high precision and robustness in diverse environments. The results not only validate the effectiveness of our improved algorithm but also highlight the importance of understanding and exploiting the graph-topological structure inherent in signal-based positioning systems.

With the popularization of Bluetooth technology, its security issues become especially critical. As a typical security threat, Bluetooth man-in-the-middle attack can cause serious damage to the integrity and privacy of data transmission. To address this challenge, this paper proposes a real-time Bluetooth signal anomaly detection model based on graph neural network (GNN). The model focuses on detecting and identifying potential man-in-the-middle attacks from complex Bluetooth signals. We first construct a Bluetooth signal dataset containing rich scenarios that model diverse communication environments and potential attack patterns. Using a graph representation learning approach, we convert the signals into graph structures, enabling the use of graph topology to analyze signal properties. Then, we design a novel graph convolution operation that captures spatio-temporal features in the signals and efficiently extracts patterns associated with attack behaviors. In the experimental section, we demonstrate the significant advantages of the proposed model in terms of detection accuracy, robustness and real-time performance by comparing it with existing methods. The model not only exhibits high accuracy in anomaly detection on the dataset, but also meets the demand of real-time monitoring in terms of processing speed. In addition, we also evaluate the generalization ability of the model and demonstrate its stability and reliability on different environments and devices.