Frozen gait (FG) is an increasingly prevalent concern in individuals with Parkinson’s disease (PD) that limits mobility and increases the risk of falls. Traditional FG detection and monitoring methods using clinical observations and wearable sensors face limitations, such as inflexibility, lack of portability, inaccessibility to individuals, and the inability to provide continuous monitoring in real-life environments. To address these challenges, this experimental study presents the development of a software-defined radio (SDR)-based radio frequency (RF) sensing platform for continuous FG monitoring. Data were collected through multiple experiments involving various physical activities, including FG episodes. The acquired data were processed using advanced signal-processing (ASP) techniques to extract relevant wireless channel state information (WCSI) patterns. The physical activities were classified using machine learning and deep learning models developed on the dataset prepared from the SDR-based RF sensing system. The results demonstrated that the deep learning models outperformed the machine learning models. The bidirectional gated recurrent unit (BiGRU) achieved the highest accuracy of 99.7%. This indicates that the developed system has the potential for accurate, real-time monitoring of FG and other PD symptoms. The proposed RF sensing platform using SDR technology and artificial intelligence (AI) offers an intelligent and continuous monitoring solution, addressing the limitations of traditional methods. This system provides portable, continuous detection of FG events, potentially improving patient care, safety, and early intervention.

Pressure ulcers or bedsores are a common health challenge among immobile patients, often leading to severe complications if not addressed promptly. The existing solutions mostly rely on direct contact and inconvenient methods that lack effective and privacy-preserving systems suitable for continuous monitoring. Furthermore, these methods frequently fail to provide accurate posture detection necessary for early intervention. This study addresses these limitations by introducing a non-contact and privacy-respecting solution that harnesses the capabilities of Wireless Channel State Information (WCSI) sensing by exploiting the Software Defined Radio (SDR) technology and Artificial Intelligence (AI). The proposed system aims to detect patient postures intelligently, contributing to bedsores while ensuring privacy and comfort with improved accuracy. The WCSI represents various human postures by conducting multiple experiments in a controlled lab environment. Advanced signal processing techniques are applied to clean the collected dataset and extract the prominent posture patterns. An intelligent sensing system is developed using Machine Learning (ML) and Deep Learning (DL) algorithms for classifying different postures to prevent bedsores. The developed ML and DL models were evaluated on a dataset prepared from the sensing system. The results indicate a trade-off between various performance metrics and computational efficiency. Among ML algorithms, the Fine Gaussian Support Vector Machine (FGSVM) outperforms others with the highest accuracy of 99.84%, indicating its reliability. While using DL algorithms, Bidirectional Long Short-Term Memory (Bi-LSTM) achieves the highest accuracy of 99.98%. The finding suggests ML models are ideal for computationally constrained scenarios, while DL models have high accuracy, and thus highlights the intelligent sensing system’s potential to mitigate pressure ulcers effectively.