Gait regularity assessed by wearable sensors: Comparison between accelerometer and gyroscope data for different sensor locations and walking speeds in healthy subjects

• Published in: Journal of biomechanics Vol. 113; p. 110115
• Main Authors: Scalera, Giovanni Marco, Ferrarin, Maurizio, Rabuffetti, Marco
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 02.12.2020; Elsevier Limited
• Subjects: Accelerometer; Accelerometers; Accelerometry; Adult; Ankle; Autocorrelation; Correlation coefficients; Data acquisition; Fitness equipment; Gait; Gyroscope; Gyroscopes; Healthy Volunteers; Human motion; Humans; Hypotheses; Inertial platforms; Inertial sensing devices; Microelectromechanical systems; Motor functions; Pelvis; Periodicity; Regularity; Rehabilitation; Sensors; Standard deviation; Treadmills; Velocity; Vertebrae; Walking; Walking Speed; Wearable computers; Wearable Electronic Devices; Windows (intervals); Wrist; Accelerometer; Gyroscope; Motor functions; Rehabilitation; Gait; Regularity
• ISSN: 0021-9290; 1873-2380; 1873-2380
• DOI: 10.1016/j.jbiomech.2020.110115

Inertial measurement units (IMU), including accelerometers and gyroscopes, can support the assessment of gait regularity, relevant for an effective walking. Gait regularity is typically quantified by an autocorrelation analysis of trunk/pelvis accelerations. A methodological upgrade fosters a multi-sensor approach including upper and lower limbs. Very few studies dealt with gait regularity using gyroscope data and no comparison between the two inertial sensors has been published. Therefore, this study compares gait regularity assessment by autocorrelation analyses performed on accelerometer and gyroscope data simultaneously acquired. Twenty-five adult healthy subjects walked steady-state on treadmill at three speeds (3.6, 5.0, 6.4 km/h), with rest between. Four IMUs were firmly attached on the trunk, pelvis, wrist and ankle. The autocorrelation method was applied to time-windows of the signal vector magnitude and resulted, on average for each trial, in its regularity index (RI) and periodicity index (PI), i.e. the stride time. Results showed that both sensors identified the same PI (Spearman correlation coefficient = 0.999), and evidenced that, for matched sensor locations and gait speeds, the accelerometer-based RI was larger/equal to the gyroscope-based RI on 86.3% of all conditions (overall median of gyroscope-to-accelerometer RI ratio was 91.1%). The two sensors gave always statistically different RI, with the exception of the ankle at the lowest speed and wrist at higher speeds. Such results help remove potential confounders from analyses performed with different sensors and support the use of accelerometers for gait regularity assessment, not excluding that gyroscopes may be more suitable for other human movements.