The use of accelerometry to assess physical activity has become more common in recent years. However, accelerometer devices are unable to determine dynamics of energy expenditure (EE) due to the lack of precision and temporal resolution. In the present study, we propose a novel method to predict EE based on transfer function (TF) analysis between oxygen uptake (VO₂) and triaxial accelerometer signals. Ten healthy subjects performed incremental treadmill walking test consisted of 4-min incremental speeds ranging from 2.5 km/h to 6.5 km/h with the increment of 1km/h. The amplitude of the three dimensional vector of body acceleration (A3D) was linearly related with VO₂ during the speeds ranging from 2.5 to 5.5 km/h. Then, we employed pseudorandom binary sequence exercise test switching the speeds between 2.5km/h and 5.5km/h to determine the individual TF relating A3D to VO₂ by applying autoregressive with extra input model. To evaluate the accuracy of the estimated TF, VO₂ response was predicted by convolving the determined impulse response with corresponding A3D obtained from square wave transition treadmill walking test at the speed from 2.5 km/h to 5.5 km/h. The average root mean square error between predicted and measured VO₂ during on-step transition was 1.18 ml/kg/min (range 0.79 to 1.75 ml/ kg/min). Gait frequency can be precisely estimated by the phase derivative of A3D signal with the error less than 1.5%. We conclude that VO₂ on-kinetics can be estimated from A3D by using an individual TF estimate together with the gait frequency. The proposed method should be useful for estimating VO₂ dynamics during walking and may be beneficial to the estimation of EE during rehabilitation usually held in situations without a gas analyzer. Further research should be needed whether the method could be applied to the ambulatory activities of walking and/or running.

DESCENTE SPORTS SCIENCE Vol.32/THE DESCENTE AND ISHIMOTO MEMORIAL FOUNDATION FOR THE PROMOTION SPORTS SCIENCE