MoveYou: feasibility study of movements checking for physiotherapy through a new algorithm for point cloud manipulations captured in real time

• Grant number: 19/16672-4
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: September 01, 2020
• End date: June 30, 2021
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Alexandre Greluk Szykman
• Grantee: Alexandre Greluk Szykman
• Company: Alexandre Greluk Szykman
• CNAE: Outras atividades de telecomunicações ; Desenvolvimento de programas de computador sob encomenda
• City: São Caetano do Sul
• Associated researchers: Anna Heloisa Greluk Pawluczkowicz
• Associated scholarship(s): 20/12996-7 - MoveYou: feasibility study of movements checking for physiotherapy through a new algorithm for point cloud manipulations captured in real time, BP.PIPE

Abstract

The increasing demand for physical therapy and motor rehabilitation treatments in the last decade has sparked academic and commercial interest in the development of technological solutions that provide support to professionals in the field. During that period, various apparatuses, algorithms and electronic games were created focusing on the verification of previously recorded professionals' movements against those of patients, performed in real time, in order to automate / quantify the treatment feedback. Overall, improvements in quality and acceleration in treatments were noted. However, all the developed solutions have two aspects in common: 1. They are based on algorithms for inference of body joints from point cloud analysis, and 2. They present problems related to the precision of inferred movements. On the other hand, recent studies focusing on recording, confrontation and temporal parameterization of point clouds have shown results with clearly higher accuracy than those focused on joints inference. Our preliminary research suggests that it is possible to converge the know-how gained with studies based directly on point cloud operations to an algorithm that checks movements by comparing these point clouds, eliminating the joints inference step. Although the accuracy found in this proposed new paradigm is proven to be greater, at least one new research problem is predicted: Inferred joint checking can easily be done by people of different height and body mass index; with point cloud checking, those factors directly influence the result. To work around this problem, we suggest an algorithm that reads a series of point clouds recorded by the expert and transforms it, frame by frame, into the patient's physical form. The algorithm should change the shape of the point clouds, but keep the motion recorded by the expert. In this way, the point clouds of the two actors can be confronted, also frame by frame: patient versus the simulation resulting from the physical form of the patient exercising the movements proposed by the specialist. The expected result is a system that allows Physiotherapy professionals to quantitatively and assertively analyze whether the movements performed by the patient are consistent with those expected for treatment. For the test of the solution, we will use previous studies on the quality of movement checking to define quantitative (through software checking) and qualitative (checking by a Physiotherapy professional) indicators. In addition to confirming the effectiveness of the algorithm, we expect to obtain new indicators, that will guide its refinement in order to improve its efficiency. By achieving the desired accuracy, our solution can revolutionize the environment in which Physical Therapy exercises are explained and checked. The proposed algorithm is part of the project named MoveYou. With enhancements enabled by technologies such as Virtual and Augmented Reality, the solution proposed by MoveYou can be further expanded and add value to learning systems and movement analysis of other Physical Education professionals, among other applications. (AU)