Combining Multi-branch Pyramid Pooling Module and BiGRU for Human Energy Expenditure Estimation Using Wearable Sensors

With the increasing emphasis on Physical Activity Energy Expenditure (PAEE), many studies have proposed various methods to evaluate the energy expenditure. However, the low prediction accuracy restricts the wide application of these approaches. In this paper, we propose an innovative architecture that Combining Multi-branch Pyramid Pooling Module(PPM) and BiGRU for Human Energy Expenditure, called Multi PPM-BiGRU, which simultaneously utilize the acceleration data from multiple body locations. To evaluate our proposed model, extensive experiments are carried out using publicly available datasets from two distinct populations: the GOTOV dataset, consisting of data from older people, and the JSI dataset, consisting of data from young people. We compare our approach with existing approaches on different units. The experimental results demonstrate that our model can achieve better performance in terms of accuracy for the energy expenditure estimation.