Wireless sensing-based human-vehicle recognition (HVR) methods have become a research hotspot due to their non-invasive, low-cost and ubiquitous advantages. However, existing methods usually have limited applicability and low recognition performance. Furthermore, the antenna deployment height and the carrier frequency of wireless signals are also key factors that affect HVR performance, but few works deeply explore their impact on HVR tasks. To address these issues, inspired by the powerful feature learning capability of recently-merged deep learning techniques, this paper proposes an attention-based convolutional neural network (CNN) model named wireless-based lightweight attention deep learning method (Wi-LADL) for HVR. Wi-LADL takes received signal strength (RSS) as input, since RSS contains the useful temporal feature of the gait for HVR. Wi-LADL consists of several lightweight CNN modules in series and a convolutional block attention mechanism (CBAM) to learn high-level features from RSS signals. Experimental results on the developed five-category RSS dataset show that the proposed Wi-LADL not only achieves higher accuracy, but also exhibits lower computational complexity compared with state-of-the-art HVR methods. In particular, Wi-LADL obtains the best performance with an average accuracy of 98.8% at 2.4 GHz band and an antenna height of 0.8 m on HVR tasks. Specifically, when classifying one-pedestrian, two-pedestrians, one-bicycle, two-bicycles and one-car, the proposed method achieves recognition accuracy of 97.06%, 97.14%, 100%, 100% and 100%, respectively. To facilitate the development of wireless sensing technology, the developed new RSS dataset has been publicly available at https://github.com/TZ-mx/WiParam.
25 January 2024 in IEEE Transactions on Intelligent Transportation Systems
In the realm of traffic parameter measurement, wireless sensing-based human-vehicle recognition methods have been pivotal due to their low cost and non-invasive nature. Traditionally, these methods have relied on the 2.4 GHz frequency band, often neglecting the rich potential of the sub-GHz bands. Furthermore, the energy attributes of wireless signals are influenced by both antenna height and carrier frequency, yet few studies have explored their impact on human-vehicle recognition performance. Addressing this critical research gap, this study introduces an innovative convolutional neural network-based method that leverages both sub-GHz bands and variable antenna heights. Specifically, this paper focuses on two key aspects: received signal strength signal-to-noise ratio analysis and wireless sensing-based human-vehicle recognition performance analysis. Experimental results demonstrate that the optimal human-vehicle recognition performance is achieved with 2.4 GHz wireless signals and an antenna height of 0.8 m, resulting in an average vehicle recognition accuracy of 95.96%. Besides, the dataset with various carrier frequencies and antenna heights has been publicly available at https://github.com/TZ-mx/mixed-RSS-dataset .
10 January 2024 in IEEE Internet of Things Journal
Wireless-sensing-based human–vehicle classification (WHVC) offer cost-effective advantages and enhance the detection efficiency of traffic parameters in intelligent transportation systems (ITSs). Existing WHVC methods primarily utilize channel state information (CSI) or received signal strength (RSS) features extracted from the surrounding wireless signals. Although CSI data provides more detailed and accurate channel information compared to RSS data, extracting and processing CSI is more challenging than RSS. Moreover, for applications that do not require fine-grained human–vehicle classification, such as intelligent street lighting systems, RSS-based WHVC has the advantages of easy implementation and low cost. Therefore, investigating the performance of CSI- and RSS-based WHVC methods in different application scenarios could provide valuable insights for the WHVC domain. To address this issue, this article proposes a deep-learning-based WHVC method, which employs deep learning as a tool to evaluate the performance of RSS and CSI methods in various classification tasks. Specifically, this article collects CSI and RSS data for seven different classification tasks in real traffic road scenarios and evaluates these tasks using a convolutional neural network (CNN)-based deep learning model designed in this article. Experimental results demonstrate that for road user categories less than four, RSS-based WHVC achieves higher accuracy than CSI-based WHVC. However, as the number of categories increases, CSI-based WHVC exhibits superior accuracy compared to RSS-based WHVC. Additionally, the developed data set is publicly available at https://github.com/TZ-mx/mixed-dataset .
With the development of Intelligent Transportation Systems (ITS), wireless sensing-based vehicle recognition (WsVRE) methods have emerged as a research hotspot in recent years. Compared to traditional vehicle recognition methods, WsVRE methods offer advantages such as strong interference resistance, low equipment costs, and ease of deployment, thereby elevating the intelligence level of ITS. Among the representative wireless sensing technologies, channel state information (CSI) based WsVRE methods have been widely applied in ITSs. However, existing CSI-based WsVRE methods have mainly focused on studying the impact of vehicle recognition algorithms on recognition performance while neglecting the influence of wireless signal attributes on the same. To address this gap and inspired by the recent advances in deep learning techniques, this paper proposes a deep learning based WsVRE method. The study employs deep learning as a tool to investigate the vehicle recognition performance under different antenna deployment positions. Furthermore, to enhance vehicle recognition accuracy, a novel CSI subcarrier filtering method is introduced. Experimental results demonstrate that the WsVRE method performs optimally with wireless signals at an antenna height of 0.6 m, achieving an accuracy of up to 97.14%.
1 January 2026 in IEEE Transactions on Instrumentation and Measurement
Accurate and real-time detection of human motion and respiration is critical for applications such as in-home health monitoring and elderly care. However, existing solutions often face challenges in multi-task sensing, continuous behavior recognition, and computational efficiency. This paper proposes CTSMR, a continuous two-stage framework based on wifi channel state information (CSI) for joint motion and respiration monitoring. The system first distinguishes between static and dynamic states, and subsequently performs respiration classification or activity recognition using a unified neural architecture. A differential threshold segmentation algorithm is introduced to enable efficient processing of continuous CSI streams, while a lightweight two-stage network, DualSenseNet, ensures low-latency inference. Experimental results demonstrate that CTSMR achieves 98.79% accuracy in respiration classification and 96.98% in activity recognition. The system maintains stable performance across varying distances and environments, with a total processing latency of 1.156ms per sample, including 0.201ms for model inference. Unlike prior isolated methods, CTSMR introduces a unified framework for continuous multi-task sensing, addressing the trade-off between high-accuracy detection and low-latency deployment in practical settings.