Hierarchical classification using ML/DL for sussex-huawei locomotion-transportation (SHL) recognition challenge

Abstract

In this paper, our team, SensingGO, presents a hierarchical classifier for Sussex-Huawei Locomotion-Transportation (SHL) recognition challenge. We first separate the original data into motorized activities and non-motorized activities in the first layer of the classifier by using accelerometer data. For the non-motorized activities, we calculate auto-correlation values with accelerometer data as input features. For the motorized activities, we take magnetometer and barometer with mean, maximum, standard deviation values as input features. Finally, we integrate the recognition results of each layer of the classifier, and the average F1-score is 50% to the validation data.