A Smart Chair Design for Recognizing Human-Object Interactions using Pressure Sensors

An important part of identifying the rich variety of human behavior in the visual world is the identification of human object interactions (HOIs). The complete behavior of a human can be understood by recognizing the human, the object, and the activity being performed by the human with the object. Although recent progress has been made in improving HOIs recognition in a completely supervised environment, there is a significant room for possible human-object interactions. In this paper, a smart chair is designed using pressure sensors to identify the object, user, and the activity based on sitting posture of the individual. For this purpose, 8 participants are asked to interact with two different objects i.e., laptop and a book while sitting on a chair and pressure sensor data are recorded. The individuals are then asked to perform three different activities with the object (laptop) and data are recorded. Data are recorded in 10 trials for object and activity. Six different features are extracted from the acquired data to classify the objects, users, and their activities using Naïve Bayes (NB), multilayer perceptron (MLP), J48, and support vector machine (SVM) classifiers. Highest accuracy of 98.75% is achieved to recognize two objects using MLP classifier. Whereas, MLP also recognizes the three activities with a highest accuracy of 80%. Moreover, an accuracy of 98.75% is achieved for identifying users of the object using NB classifier.