The purpose of this study was to develop a new algorithm for classifying physical activity into either locomotive or household activities using a triaxial accelerometer. Sixty-six volunteers (31 men and 35 women) participated in this study and were separated randomly into validation and cross-validation groups. All subjects performed 12 physical activities (personal computer work, laundry, dishwashing, moving a small load, vacuuming, slow walking, normal walking, brisk walking, normal walking while carrying a bag, jogging, ascending stairs and descending stairs) while wearing a triaxial accelerometer in a controlled laboratory setting. Each of the three signals from the triaxial accelerometer was passed through a second-order Butterworth high-pass filter to remove the gravitational acceleration component from the signal. The cut-off frequency was set at 0.7 Hz based on frequency analysis of the movements conducted. The ratios of unfiltered to filtered total acceleration (TAU/TAF) and filtered vertical to horizontal acceleration (VAF/HAF) were calculated to determine the cut-off value for classification of household and locomotive activities. When the TAU/TAF discrimination cut-off value derived from the validation group was applied to the cross-validation group, the average percentage of correct discrimination was 98.7%. When the VAF/HAF value similarly derived was applied to the cross-validation group, there was relatively high accuracy but the lowest percentage of correct discrimination was 63.6% (moving a small load). These findings suggest that our new algorithm using the TAU/TAF cut-off value can accurately classify household and locomotive activities.
Copyright 2010 Elsevier B.V. All rights reserved.