Human Activity Recognition and People Count for a SMART Public Transportation System

In public transportation networks, it is desirable to detect and count people around bus stations. To achieve this goal, a unified dynamic human activity recognition and people counting (HARC) is applied to a public dataset for Wi-Fi-based activity recognition (WiAR) and a performance evaluation is carried out using a machine learning-based data processing framework. The processing results reported in this paper show that the accuracy achieved by HARC is 94% when the Adam optimizer method is used.