Authors:
Carlos Polvorinos-Fernández
1
;
Luis Sigcha
2
;
María Centeno-Cerrato
1
;
Elena Muñoz-Bellido
1
;
César Asensio
3
;
Juan M. López
4
;
Guillermo de Arcas
1
and
Ignacio Pavón
1
Affiliations:
1
Instrumentation and Applied Acoustics Research Group, Mechanical Engineering Department, ETS Ingenieros Industriales, Universidad Politécnica de Madrid, Madrid, Spain
;
2
Department of Physical Education and Sports Science, Health Research Institute, & Data-Driven Computer Engineering (D2iCE) Group, University of Limerick, Limerick, Ireland
;
3
Instrumentation and Applied Acoustics Research Group, Department of Audiovisual Engineering and Communications, ETS. de Ingeniería y Sistemas de Telecomunicación, Universidad Politécnica de Madrid, Madrid, Spain
;
4
Instrumentation and Applied Acoustics Research Group, Department of Physical Electronics, Electrical Engineering and Applied Physics, ETS. de Ingeniería y Sistemas de Telecomunicación, Universidad Politécnica de Madrid, Madrid, Spain
Keyword(s):
Wearables, Gait, Machine Learning, Accelerometer, Gyroscope.
Abstract:
Human gait is a biomechanical process vital to health, with abnormalities often linked to neurological disorders like Parkinson's disease (PD). In PD patients, arm swing during walking becomes asymmetric and reduced in amplitude, providing a potential biomarker for early diagnosis and monitoring disease progression. This pilot study focuses on detecting variations in arm swing amplitude and asymmetry using data collected from smartwatches worn by 24 participants under different gait conditions. Participants walked while carrying progressively heavier loads (0 kg, 2 kg, and 4 kg) to simulate restricted arm swing. Machine learning models were developed to classify these conditions using accelerometer and gyroscope data. Results showed that the K-Nearest Neighbours algorithm performed best, achieving up to 94.3% accuracy. Although the models effectively distinguished between load and no-load conditions, it was difficult to differentiate between different load levels. These findings high
light the potential of wearable devices for PD gait analysis, though further refinement and testing with PD patients are needed for clinical application.
(More)