Multi-modal fusion in ergonomic health: bridging visual and pressure for sitting posture detection

As the contradiction between the pursuit of health and the increasing duration of sedentary office work intensifies, there has been a growing focus on maintaining correct sitting posture while working. This paper introduces a sitting posture recognition system that integrates visual and pressure modalities. The system uses differentiated pre-training for bimodal models and a feed-forward feature fusion module, combining laptop camera data with thin-film pressure sensor mat data in office scenarios. It achieved an F1-Macro score of 95.43% on a dataset with complex composite actions, improving over systems that rely solely on pressure or visual modalities and over a uniform pre-training strategy.

CCF Transactions on Pervasive Computing and Interaction (TPCI), 6, 380-393