A CNN-LSTM Model for IMU-based Energy Expenditure Estimation under Various Walking Conditions

Energy expenditure (EE) is important quantities in evaluating the intensity of physical activity and health status. EE can be estimated based on signals from wearable sensors such as inertial measurement unit (IMU) and electromyography (EMG) sensor. Although many algorithms have been developed to estimate EE during daily life [1] – [4] , only a few algorithms have evaluated the performance of EE according to walking conditions (speed and incline) [2] – [4] . This paper presents a combined model of convolutional neural network (CNN) and long short-term memory (LSTM) for an IMU-based EE estimation under various walking conditions. The results of this study suggest the optimal IMU location for estimating EE during level walkings/runnings and inclined walkings.