Current and future trends of wearable sensors in gait analysis
By Dr Han Yi Chiew
Gait analysis, which is the study of human walking patterns, is useful in various fields, including rehabilitation, sports performance optimisation, and medical diagnosis. Wearable sensors, typically integrated into shoes, clothing, or body-worn devices, have gained significant popularity in gait analysis for their portability and cost-effectiveness.
These sensors are usually small and include accelerometers, gyroscopes and magnetometers, while pressure sensors and even electromyography sensors may be included depending on the applications.
While cameras have been employed for gait analysis, their use typically requires a controlled laboratory environment where subjects move within a predefined space, maintaining a direct line-of-sight to the cameras. In contrast, wearable sensors are used in real-world settings surrounded by items such as walls and furniture. This is particularly useful for assessing gait in everyday life, such as monitoring elderly individuals for fall risk or evaluating patients’ gait recovery level.
Nevertheless, the need for cameras, particularly optical motion capture systems, persists in laboratory settings to validate the performance of wearable sensors. Once these wearable sensors have successfully undergone validation in the lab, they can then be employed in daily life independently, without ongoing need for cameras.
The availability of motion sensors in smartphones has further popularised wearable sensors in gait analysis, which is evident in existing applications that estimate the number of walking steps and recognise human activities such as walking, jogging, and stair climbing. Future applications include but are not limited to identifying user identity based on the gait pattern and subsequently auto-locking or unlocking the phone for enhanced security and convenience.
The signal processing techniques for wearable sensors revolve around mathematical modelling over the past decade. This involves a comprehensive understanding of the pros and cons of different types of sensors, then derives mathematical equations to fuse these sensors to mitigate respective drawbacks. For instance, integration of gyroscope readings introduces drifting issue in angle estimations, while accelerometers result in inaccurate angle estimations during motions. Fusing the sensors mathematically through algorithms such as Madgwick’s filter can counterbalance each sensor’s disadvantages and yield accurate results.
Current trend focuses mostly on machine learning for automated gait analysis and predictive modeling, for example, activity recognition and gait abnormality detection. This includes applying, fine-tuning and modifying deep learning algorithms to achieve accurate and real-time estimations.
Since machine learning is involved, handling of seen and unseen data in training is crucial. In gait analysis, this does not only involve splitting the dataset by a percentage, but also completely excluding training data associated with the test individual to generate a generalised model. These, however, are still classification tasks where the results are usually in binary format such as normal or abnormal gait and walking or standing.
Future trend appears to be shifting towards estimations expressed in decimals, offering numerical assessments of the gait, akin to devices like glucometers that generate precise numerical results.
The current and future trends in wearable sensors for gait analysis demonstrate a significant shift towards more practical and versatile applications. Wearable sensors have revolutionised gait analysis by allowing assessments in real-world settings, enabling monitoring of everyday activities. The current focus on machine learning for automated gait analysis is paving the way for more accurate and real-time estimations.
As the field advances, it is poised to promise greater insights into human mobility and health. Emerging applications, such as identity recognition based on gait patterns to enhance security, foreshadow a future filled with increasingly diverse and sophisticated possibilities.
Dr Han Yi Chiew serves as both a lecturer and programme coordinator for the Bachelor of Technology in Computer Systems and Networking at Curtin University Malaysia. His expertise lies in gait analysis, with numerous research publications featured in IEEE journals. In addition, he actively engages in various projects involving embedded systems engineering, Internet of Things, and the development of trading bots using artificial intelligence. Dr. Han can be contacted by email at hanyichiew@curtin.edu.my.