An Expert System for Classification and Detection of Improper Posture
Shripad S. Bhatlawande, Anusha Agrawal, Adhiraj Jagdale, Swati Shilaskar · 2022 3rd International Conference for Emerging Technology (INCET) · 2022
Posture is something that is often neglected when a person is focused on said activity and hence there are several Artificial Intelligence based systems that have been developed to intelligently detect a poor posture and immediately alert the user. This paper presents an exhaustive comparative analysis of posture detection and recognition models such as accelerometer sensor based, Kinect sensor based and computer vision-based models that are currently available. The powers of AI to detect the posture are also reviewed. Further, the paper proposes three different posture recognition systems that performed better than the models that currently exist. They also overcome several limitations faced by the current systems. The proposed systems used four classification algorithms which were SVM, KNN, Random Forest and Decision Tree. Of these, SVM was found to be the best performing algorithm for the first system which yielded an accuracy of 99.87%. The second gave an accuracy of 100% for all the four classification algorithms. Third system gave an accuracy of 100% for SVM, KNN and Random Forest. It performed with an accuracy of 99.3% for decision tree classifier. This paper also highlights the use of the Grid Search tuning method to obtain the optimum hyperparameters for each classification algorithm.