| dc.description.abstract | Features related to gait are fundamental metrics of human motion [1]. Human gait has been shown to be a valuable and feasible clinical marker to determine the risk of physical and mental functional decline [2], [3]. Technologies that detect changes in people’s gait patterns, especially older adults, could support the detection, evaluation, and monitoring of parameters related to changes in mobility, cognition, and frailty. Gait assessment has the potential to be leveraged as a clinical measurement as it is not limited to a specific health care discipline and is a consistent and sensitive test [4].
A wireless technology that uses electromagnetic waves (i.e., radar) to continually measure gait parameters at home or in a hospital without a clinician’s participation has been proposed as a suitable solution [3], [5]. This approach is based on the interaction between electromagnetic waves with humans and how their bodies impact the surrounding and scattered wireless signals. Since this approach uses wireless waves, people do not need to wear or carry a device on their bodies. Additionally, an electromagnetic wave wireless sensor has no privacy issues because there is no video-based camera.
This thesis presents the design and testing of a radar-based contactless system that can monitor people’s gait patterns and recognize their activities in a range of indoor environments frequently and accurately. In this thesis, the use of commercially available radars for gait monitoring is investigated, which offers opportunities to implement unobtrusive and contactless gait monitoring and activity recognition. A novel fast and easy-to-implement gait extraction algorithm that enables an individual’s spatiotemporal gait parameter extraction at each gait cycle using a single FMCW (Frequency Modulated Continuous Wave) radar is proposed. The proposed system detects changes in gait that may be the signs of changes in mobility, cognition, and frailty, particularly for older adults in individual’s homes, retirement homes and long-term care facilities retirement homes. One of the straightforward applications for gait monitoring using radars is in corridors and hallways, which are commonly available in most residential homes, retirement, and long-term care homes. However, walls in the hallway have a strong “clutter” impact, creating multipath due to the wide beam of commercially available radar antennas. The multipath reflections could result in an inaccurate gait measurement because gait extraction algorithms employ the assumption that the maximum reflected signals come from the torso of the walking person (rather than indirect reflections or multipath) [6].
To address the challenges of hallway gait monitoring, two approaches were used: (1) a novel signal processing method and (2) modifying the radar antenna using a hyperbolic lens. For the first approach, a novel algorithm based on radar signal processing, unsupervised learning, and a subject detection, association and tracking method is proposed. This proposed algorithm could be paired with any type of multiple-input multiple-output (MIMO) or single-input multiple-output (SIMO) FMCW radar to capture human gait in a highly cluttered environment without needing radar antenna alteration. The algorithm functionality was validated by capturing spatiotemporal gait values (e.g., speed, step points, step time, step length, and step count) of people walking in a hallway. The preliminary results demonstrate the promising potential of the algorithm to accurately monitor gait in hallways, which increases opportunities for its applications in institutional and home environments. For the second approach, an in-package hyperbola-based lens antenna was designed that can be integrated with a radar module package empowered by the fast and easy-to-implement gait extraction method. The system functionality was successfully validated by capturing the spatiotemporal gait values of people walking in a hallway filled with metallic cabinets. The results achieved in this work pave the way to explore the use of stand-alone radar-based sensors in long hallways for day-to-day long-term monitoring of gait parameters of older adults or other populations.
The possibility of the coexistence of multiple walking subjects is high, especially in long-term care facilities where other people, including older adults, might need assistance during walking. GaitRite and wearables are not able to assess multiple people’s gait at the same time using only one device [7], [8]. In this thesis, a novel radar-based algorithm is proposed that is capable of tracking multiple people or extracting walking speed of a participant with the coexistence of other people. To address the problem of tracking and monitoring multiple walking people in a cluttered environment, a novel iterative framework based on unsupervised learning and advanced signal processing was developed and tested to analyze the reflected radio signals and extract walking movements and trajectories in a hallway environment. Advanced algorithms were developed to remove multipath effects or ghosts created due to the interaction between walking subjects and stationary objects, to identify and separate reflected signals of two participants walking at a close distance, and to track multiple subjects over time. This method allows the extraction of walking speed in multiple closely-spaced subjects simultaneously, which is distinct from previous approaches where the speed of only one subject was obtained. The proposed multiple-people gait monitoring was assessed with 22 participants who participated in a bedrest (BR) study conducted at McGill University Health Centre (MUHC).
The system functionality also was assessed for in-home applications. In this regard, a cloud-based system is proposed for non-contact, real-time recognition and monitoring of physical activities and walking periods within a domestic environment. The proposed system employs standalone Internet of Things (IoT)-based millimeter wave radar devices and deep learning models to enable autonomous, free-living activity recognition and gait analysis. Range-Doppler maps generated from a dataset of real-life in-home activities are used to train deep learning models. The performance of several deep learning models was evaluated based on accuracy and prediction time, with the gated recurrent network (GRU) model selected for real-time deployment due to its balance of speed and accuracy compared to 2D Convolutional Neural Network Long Short-Term Memory (2D-CNNLSTM) and Long Short-Term Memory (LSTM) models. In addition to recognizing and differentiating various activities and walking periods, the system also records the subject’s activity level over time, washroom use frequency, sleep/sedentary/active/out-of-home durations, current state, and gait parameters. Importantly, the system maintains privacy by not requiring the subject to wear or carry any additional devices. | en |
