Remote Gait Analysis Using Ultra-Wideband Radar Technology

Research Background

Gait analysis, the study of the coordinated biomechanical patterns of human walking, is not only a critical component of biomechanics research but also provides valuable information on health status. In recent years, there has been a growing interest in developing new at-home gait analysis solutions that allow individuals to undergo comprehensive gait evaluations in the comfort of their homes. This advancement benefits healthy individuals looking to optimize their gait performance and assists patients with chronic musculoskeletal diseases (such as arthritis and back pain), acquired brain injuries (such as stroke and traumatic brain injury), and neurodegenerative diseases (such as Parkinson’s and Alzheimer’s disease).

Existing gait analysis technologies are mainly divided into wearable (direct) and non-wearable (remote) categories. Wearable systems, due to user non-cooperation issues, especially for the elderly, are often unsuitable for long-term monitoring. Although marker-based motion capture technology is currently the gold standard in gait evaluation, its spatial and high cost requirements make it difficult to use in home environments. Hence, non-wearable, markerless technology is especially favored. Electromagnetic radar systems are a particularly promising markerless monitoring solution, and ultra-wideband (UWB) radar, due to its time-of-flight measurement capabilities, has already seen success in numerous health-related applications, such as vital signs analysis, sleep stage classification, cardiovascular monitoring, daily activity recognition, and fall detection. These developments highlight UWB radar’s ability to extract physiological and behavioral biomarkers.

Research Source and Author Information

This study was authored by Charalambos Hadjipanayi, Maowen Yin, Alan Bannon, Adrien Rapeaux, Matthew Banger, Shlomi Haar, Tor Sverre Lande, Alison H. McGregor, and Timothy G. Constandinou from the Department of Electrical and Electronic Engineering, Dementia Research Centre, Biomimetics Lab, and Department of Surgery and Cancer at Imperial College London, and the Department of Informatics at the University of Oslo. The research was published in the 2023 issue of the IEEE Transactions on Biomedical Engineering journal.

Research Purpose and Methods

The primary goal of this study is to explore the feasibility of ultra-wideband radar in markerless remote gait analysis. Specifically, by analyzing Doppler information from three monostatic ultra-wideband radar sensors, the study aims to extract ten clinically significant spatiotemporal gait characteristics. These gait parameters were compared with data from an optical motion capture system to verify their accuracy.

Data Collection and Experimental Setup

In the experimental setup, three independent monostatic impulse radio ultra-wideband (IR-UWB) radars (XeThruTM X4M03 from Novelda AS) were used, and data were collected along a specified path (Figure 1a). All radars were configured to simultaneously transmit pulses at a carrier frequency of 7.29 GHz with a sampling rate of 500 Hz. Ground truth motion data were collected using a marker-based 3D motion capture system composed of 28 high-resolution infrared cameras tracking the positions of 27 reflective markers. This allowed the study to record participants’ gait information while walking steadily.

Algorithm Steps

1. Preprocessing and Doppler Analysis: Radar signal preprocessing included noise suppression and adaptive clutter suppression using an exponential moving average filter. This was followed by digital quadrature demodulation of the radar signals to extract the raw baseband signal.
2. Time-Frequency Map Extraction: A three-dimensional range-Doppler-time (RDT) mapping was obtained by computing the short-time Fourier transform of the radar signals.
3. Gait Event Detection: Gait cycle events, including heel strike (HS) and toe-off (TO) times, were detected based on the simulated maximum envelope trajectory.
4. Extraction of Spatiotemporal Gait Parameters: Ten clinical gait parameters, including step length, stride length, stride time, and walking speed, were extracted. These parameters were used to calculate gait variability and asymmetry.

Results and Discussion

The results show that nine key gait parameters can be consistently estimated with an accuracy of 90-98%, capturing 94.5% of the participants’ gait variability and 90.8% of bilateral symmetry. Further correlation analysis and Bland-Altman analysis revealed strong correlations and high consistency between radar-based gait parameters and ground truth values.

Analysis of Radar Configuration Effects

The performance of the algorithm was compared across different numbers of radars (single, dual, and triple radar configurations). Results indicated that the more radars used, the higher the overall accuracy. In the optimal triple-radar one-dimensional localization configuration, spatial feature extraction accuracy was the highest.

Research Conclusion

The study demonstrates that ultra-wideband radar can deliver gait analysis quality comparable to the gold standard. It has the potential to shift from costly and cumbersome lab-based gait analysis tools to non-intrusive and affordable home deployment solutions, enabling researchers and healthcare applications to conduct long-term continuous monitoring. Although initial results are promising, further research is needed to assess the system’s performance on patients with gait abnormalities in actual application scenarios to achieve broader clinical and home applications.