Step length and gait speed estimation using a hearing aid integrated accelerometer: A comparison of different algorithms

• DOI: 10.36227/techrxiv.24182496
• Publication Date: 2023-09-29T20:23:15Z
• Authors: Ann-Kristin Seifer; Arne Küderle; Eva Dorschky; Hamid Moradi; Ronny Hannemann; Bjoern Eskofier
• Keywords: Feature (linguistics); Motion Capture

<p>Abstract: Gait is an indicator of a person’s health status and abnormal gait patterns are associated with higher risk falls, dementia, mental disorders. Wearable sensors facilitate long-term assessment walking in the user’s home environment. Earables, wearable that worn at ear, gaining popularity for digital assessments because they unobtrusive can easily be integrated into daily routine, example, hearing aids. A comprehensive analysis pipeline ear-worn accelerometer includes spatial-temporal parameters currently not existing. Therefore, we propose compare three algorithmic approaches to estimate step length speed based on data: biomechanical model, feature-based machine learn- ing (ML) models, convolutional neural network. We evaluated their performance bout level compared it optical motion capture system. The ML model achieved best precision 4.8 cm level. For speed, learning approach absolute percentage error 5.3% (± 3.9%). find able clinically relevant precision. Furthermore, insensitive different age groups sampling rates but sensitive speed. To our knowledge, this work first contribution estimating using accelerometers. Moreover, lays foundation framework enabling continuous monitoring home.</p>