Accurate and Efficient Sitting Posture Recognition and Human-Machine Interaction Device Based on Fabric Pressure Sensor Array and Neural Network

This article presents a novel approach to interact with users through posture recognition, leveraging its advantages of convenience and real-time feedback to enhance user engagement and personalized experiences. In contrast to traditional methods that rely on camera-based posture detection, this study proposes a deep learning-based framework for posture recognition by classifying the distribution of body pressure under different sitting positions. The system integrates a large-area, highly flexible fabric pressure sensor array into the chair, which collects data on posture-specific pressure patterns for training and identification purposes. A deep learning algorithm, specifically the LeNet architecture, is employed to classify 49 different sitting positions based on angular variations, including body tilt to the left or right, standard posture, and forward or backward leaning. The proposed approach achieves an impressive accuracy rate of 99.86%. Furthermore, the application of this posture recognition system in VR devices enables intelligent chair control for VR games. This research provides strong support for future advancements in chair design and human-computer interaction technologies, enhancing ergonomic designs in various domains such as automotive seats, office chairs, and medical seating, while simultaneously improving user comfort and well-being. A new human-machine interaction method is presented using fabric pressure sensors and deep learning, achieving 99.86% accurate sitting posture classification. It enables race car control based on posture recognition.image