User and Period Independent Transportation Mode Detection for Wheelchair Users

Abstract

Transportation mode detection (TMD) is an important research area in human activity recognition. It can improve the mobility and accessibility of people by providing a better understanding of their mobility patterns, thereby enhancing their quality of life and social inclusion. Although previous studies of TMD for people without mobility disabilities exhibited, the performance of TMD models on new users and periods was limited. This issue would be more important for people with mobility disabilities. This study investigated the negative impact of user and period differences on the performance of TMD for wheelchair users (wTMD) and suggested a method to address these challenges. Our main findings are (1) user and period differences degraded the wTMD performance from 94.28% to 59.32%; (2) the multi-DenseNet with a soft voting ensemble provided a 76.49% accuracy to data from different users and periods. We expect that our understanding of wTMD will aid in the design of more generalized wTMD models.

Keywords

• Transportation mode detection
• deep learning
• smartphone
• mobility disability