Frameworks of Transportation Mode Detection for People with Handicaps: HCI Data Collection Methods, AI Analyses, and Social Science Findings

Abstract

With the aging population, the number of people with mobility is increased for a decade. The people with mobility disability can easily lose life satisfaction and quality of life because they are limited to mobility. Therefore, knowing their pattern of mobility is important to enhance their mobility and increase the life satisfaction. Although Transportation Mode Detection (TMD) technology have potential to solve this problem, previous studies were limited to people with mobility disability such as wheelchair users. Therefore, this thesis developed and suggested the framework of TMD for expanding to people with mobility disability. First, we developed the smartphone application, called ‘TMD-APP’ for data collection. By applying this application, experimenter can configurate data collection considering people with mobility disability and participants also can collect data. In this thesis, we collected 33,788 minutes of TMD data including wheelchair modes from people with and without mobility disability. We also classified seven transportation modes using a recurrent neural network, known as long short-term memory, with accuracies of 95.82% in people without mobility disability and 96.70% in people with people with mobility disability. However, the generalization problem, called intraclass variability, was suggested by previous studies. Especially, the performance of TMD models on new users and periods was limited. This issue would be more important for people with mobility disabilities. This thesis 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. In this thesis, we found that user and period differences degraded the wTMD performance from 94.28% to 59.32% and DenseNet with a soft voting ensemble provided a 76.49% accuracy to data from different users and periods. Previous TMD studies mainly focused on improving detection performance. Therefore, based on our framworks, TMD is applied to investigate the social implication of mobility information. In this thesis, we developed wTMD using global positioning system (GPS) data and apply it to determine how transportation behaviors affect the life satisfaction of wheelchair users. First, we proposed a wTMD technology from GPS data collected by this thesis. We conducted regression analyses on existing in-the-wild GPS dataset of wheelchair users. The result shows that the portion of subways in an individual’s travel time is directly connected to wheelchair users’ life satisfaction in Seoul, South Korea. This thesis uses computational techniques to predict the life satisfaction of wheelchair users. We suggested that the framework of TMD developed in this thesis can help people with mobility disability. We believe that the framework can detect the pattern of mobility pattern, and this finding can aid to design the urban planning and accessibility for people with mobility disability. Furthermore, this framework can be beneficial for enhancing the quality of life and enabling the social inclusion of people with mobility disabilities.