Human activity monitoring is a fascinating area of research to support autonomous living in the aged and disabled community. Cameras, sensors, wearables, and non-contact microwave sensing have all been suggested in the past as methods for identifying distinct human activities. Microwave sensing is an approach that has lately attracted much interest since it has the potential to address privacy problems caused by cameras and discomfort caused by wearables, especially in the healthcare domain. A fundamental drawback of the current microwave sensing methods such as radar is non-line-of-sight and multi-floor environments. They need precise and regulated conditions to detect activity with high precision. In this paper, we have utilised the publicly available online database based on the intelligent reflecting surface (IRS) system developed at the Communications, Sensing and Imaging group at the University of Glasgow, UK (references 39 and 40). The IRS system works better in the multi-floor and non-line-of-sight environments. This work for the first time uses algorithms such as support vector machine Bagging and Decision Tree on the publicly available IRS data and achieves better accuracy when a subset of the available data is considered along specific human activities. Additionally, the work also considers the processing time taken by the classier in training stage when exposed to the IRS data which was not previously explored.

This paper presents a study on contactless localization for activity recognition based on radio-frequency sensing. The focus of this study is on the quantitative analysis of the collected data, which is in the form of channel state information (CSI). The proposed method utilizes a software-defined radio (SDR) system in combination with an ensemble learning technique to localize and identify daily living activities such as leaning, sitting, standing and walking. Specifically, SDR device, Universal Software Radio Peripheral (USRP) models X300/X310 are utilized to collect data on the aforementioned activities. The data is collected from an empty space and a participant performing daily living activities in different territories. The acquired data is labelled based on the region, zone and performed activity. The CSI data is subsequently preprocessed and fed into an ensemble-based machine learning algorithm for classification. Furthermore, the stability analysis of the proposed method is performed to evaluate its reliability and robustness. The performance of the algorithm is evaluated using various metrics, including a confusion matrix, accuracy, cross-validation score and training time. The results demonstrate that the proposed ensemble-based approach achieves a high accuracy of up to 90% in activity recognition, highlighting the effectiveness of the proposed method in contactless localization for activity recognition.

Non-invasive indoor human activity detection using radio waves has attracted the interest of researchers, contributing to a range of new applications including smart healthcare. Localisation of activities can assist in developing advanced healthcare systems able to identify the location of patients. Radio frequencies have been shown in numerous studies as a non-invasive method to identify human activity. This is achieved by observing the signal propagation described in the Channel State Information (CSI). This paper presents experimental results using Universal Software-Defined Radio Peripheral (USRP) devices to identify and localise a single human subject performing activities by utilizing the CSI of radio frequencies. The experiments are carried out to retrieve CSI samples observing a single subject perform no-activity, sitting, standing, and leaning forward actions in various positions in a room. Additional CSI is captured for the subject walking in two directions across the observed area. Giving a total of 6 activities spanning the monitored area. CSI is also collected while the monitored area is empty for further comparison. Artificial intelligence is used to make classifications on collected CSI. The proposed approach uses a Super Learner (SL) algorithm that can identify the location of different activities with 96% accuracy, outperforming existing benchmark approaches.

Current approaches to activity-assisted living (AAL) are complex, expensive, and intrusive, which reduces their practicality and end user acceptance. However, emerging technologies such as artificial intelligence and wireless communications offer new opportunities to enhance AAL systems. These improvements could potentially lower healthcare costs and reduce hospitalisations by enabling more effective identification, monitoring, and localisation of hazardous activities, ensuring rapid response to emergencies. In response to these challenges, this paper introduces theTransparentRFIDTag Wall (TRT-Wall), a novel system taht utilises a passive ultra-high frequency (UHF) radio-frequency identification (RFID) tag array combined with deep learning for contactless human activity monitoring. TheTRT-Wallis tested on five distinct activities: sitting, standing, walking (in both directions), and no-activity. Experimental results demonstrate that theTRT-Walldistinguishes these activities with an impressive average accuracy of $$95.6\%$$95.6% under four distinct distances (2, 2.5, 3.5 and 4.5 m) by capturing the RSSI and phase information. This suggests that our proposed contactless AAL system possesses significant potential to enhance elderly patient-assisted living.

We present RFSensingGPT, an integrated framework for radio frequency (RF) sensing that combines technical question-answering, code retrieval, and spectrogram analysis through retrieval-augmented generation (RAG). Our framework addresses the fundamental challenge of applying large language models to RF sensing applications, where specialized domain knowledge is underrepresented in general training corpora. The system leverages a filtered RedPajama dataset containing RF-relevant technical documents, processed through a hybrid retrieval mechanism that combines vector-based similarity search with best match (BM25)-based query fusion. Performance evaluation using document collections ranging from 5K to 80K demonstrates that RAG consistently maintains superior faithfulness across all dataset sizes (0.9033-0.9779 vs 0.8162-0.8506, average improvement of 13.0%) compared to baseline LLM implementations. Our hierarchical chunking approach using MarkdownHeaderTextSplitter achieves optimal precision (0.31-0.32) at lower k-values while maintaining correctness scores of 4.0-5.0. The framework integrates CLIP-based vision models for RF pattern recognition, achieving 93.23% accuracy in radar data analysis tasks. Implementation benchmarks show efficient processing with minimal GPU memory requirements (0.66GB) even at scale. Through a comprehensive evaluation of the embedding models, RFSensingGPT establishes a new benchmark for technical query understanding and RF spectrogram analysis in the emerging field of integrated sensing and communications systems for 6G networks.

Complex software systems that support organizations are updated regularly, which can erode system architectures. Moreover, documentation is rarely synchronized with the changes to the software system. This creates a slew of issues for future software maintenance. To this goal, information extraction tools use exact approaches to extract entities and their corresponding relationships from source code. Such exact approaches extract all features, including those that are less prominent and may not be significant for modularization. In order to resolve the issue, this work proposes an enhanced approximate information extraction approach, namely, fact extractor system for Java applications (FESJA) that aims to automate software modularization using a fact extraction system. The proposed FESJA technique extracts all the entities along with their corresponding more dominant formal and informal relationships from a Java source code. Results demonstrate the improved performance of FESJA, by extracting 74 (classes), 43 (interfaces), and 31 (enumeration), in comparison with eminent information extraction techniques.