Correction for the
errors by riding vehicles to estimate VO2 during walking by tri-axial
accelerometry
Toshiaki
Yamazaki1,2, Hirokazu Gen-no2,3, and Hiroshi Nose1,3
1Department of Sports Medical Sciences, Shinshu
University Graduate School of Medicine, Matsumoto 390-8621, Japan; 2Kissei
Comtec Co., Ltd., Matsumoto 390-1293; 3Jukunen Taiiku Daigaku
Research Center, Matsumoto 390-8621, Japan
Recently,
we have developed a device to estimate VO2 during uphill and
downhill walking by using tri-axial accelerometry and barometry (MSSE, in
press). Here, we propose an algorithm to correct the errors by passive body
movement by riding vehicles in the measurement. Forty two subjects with
tri-axial accelerometer (JD Mate, Kissei Comtec, Japan) on their waist walked on
a treadmill while tri-axial accelerations (Ax, Ay, and Az) were measured every min. Other 5 subjects with JD Mate rode on a car,
a train, and a bike while the accelerations were recorded as in the walking
test. Then, we performed discriminant analysis on acceleration data. The
rates of correct classification were 98.5%, 93.8%, 89.4%, and 95.3%, for
walking and riding a car, a train, and a bike, respectively. The wrong
classification into walking category from riding a car, a train and a bike were
3.9%, 0.0%, and 1.6%, respectively.
Thus, this algorithm made it possible to estimate mere VO2 by
walking in our daily activity.
Keywords: triaxial accelerometer, discriminant analysis