IMU based gait tracking

Analysis of human gait is needed in different fields, especially biomedical. Diagnosis and treatment of diseases such as muscle disorders, spinal disorders, joint diseases, parkinsonism or walking on bipedal robots and human walk can be example of these fields. The task of the balance system are, to prevent undesired displacement of the center of mass from the equilibrium position on the support floor, to ensure correct positioning of the body position on the move, to control the eye movements to provide a clear view as the person or the environment moves. The balancing system has a complex structure which includes three main parts, collecting datas from various body systems, processing datas and decision making, applying the decision. Visual datas are provided by visual system, datas such as pressure, touch are provided by somatosensory system and datas which is about gravity or rotation are provided by vestibular system. Collected datas are processed in the brain and decisions are decided. Decided decisions are turned movement by muscle-skeleton system. If there is a problem at any stage of the balance process, the patient may experience difficulties in performing activities of daily living, gait disturbances, falls, and in some cases complaints of drowsiness.It is critical for people with the mentioned complaints to correctly analyze their gaits in the diagnosis and treatment of their illnesses. There are three main approaches used in anaysis and recognition of human gait which can be listed as: image processing, floor sensors and sensor placed on the body. In the process of image processing, the movements of the persons are recorded with one or more cameras and various filters are applied to analyze the gait. In some applications, optical markers are placed on the moving person. The problems of the image processing method are the separation of the moving image from environment image and the inability to repeat the natural movement because of the image is taken in a fixed area usually in the treadmills. At studies using floor sensors, the floor is equipped with sensors which capable of measure the applying forces. The problem with this method is ignoring the parameters such as the posture, the movement of the legs, etc., since only the applied force is measured. In the methods in which the body sensors are used, the movement sensors are placed to the joint points on the human body. Sensors are small in size and lightweight, so they do not interfere with the natural walk. There is no need for special environment to record data. The main problem of this sensors is effect of offset error, which is varying with time, is high while calculating the position. Within the scope of the project, researchers in the field of engineering and medicine worked together. Medical doctors have contributed to the determination of gait physiology and parameters that may be effective in disease detection. Engineers have been working on filtering data collection data and obtaining meaningful results. In addition, the collected data were used to diagnose the disease using machine learning algorithms. This thesis aims developing analyze dynamic vestibular system algorithm and designing balance detecting device. Pressure sensors and inertial measurement units will used in this project and collected datas from various sensors correlates with each other. This system is used for patients who has vestibular diseases, when detecting and treating illnesses. This thesis is part of the project aiming to develop a dynamic vestibular system analysis algorithm that will help the diagnosis and treatment of diseases of people with vestibular system diseases by combining the data collected from pressure sensors and inertial measurement units and designing the balance detection device. In this study, it is aimed to determine the parameters that can be used in gait analysis within the scope of the project, to collect data from the sensors of healthy persons and people who have balance problems, to establish wireless communication between sensors and computers, to evaluate these datas in MATLAB environment and to obtain predetermined parameters, and to calculate the possible correlations by evaluating the parameters obtained from the sensors. In the first part, the human balancing system is mentioned and the methods used in gait tracking are explained together with their advantages and disadvantages. Studies which is done in similar areas in the literature are examined, hardware and software tools which is used, data collection and processing algorithms are explained. In the second part, gait physiology is examined and parameters to be obtained from the collected data are explained. Among these parameters, criterias and calculations which is used are given in determining the parameters that may be distinguishable in disease diagnosis. In the third part, firstlt general information is given about the inertial measurement units used. The hardware and software features of the sensors have been examined. Tools used to transfer data to and from the computer are described. How the sensors are positioned on the body is given. Afterwards, the criteria to be considered when selecting the data collection environment, the measurements made are explained. After all the system components are explained, the general structure of the system is explained in order of the operations performed when data is collected. At the fourth part the post-data analysis steps recorded are explained. Firstly, what is the free acceleration data obtained from the sensors and why it is used. It is explained together with the results of the proposed method for eliminating the offset and misalignment errors and mistakes encountered in obtaining the position data from the foot sensors afterwards. In the next step, the problems encountered in calculating the position data from the knee sensors are discussed in terms of differences from the method used in the foot sensors. The algorithms used to solve these problems are described along with the results of the applied filtering methods. Angular data measured from knee and waist sensors are shown. The parameters to be used in the gait analysis after receiving measurements are shown together with the results. Incorporation of data obtained from IMU and pressure sensors has been included. Finally, the road map used for the calculation of the calculated parameters in the machine learning is briefly explained. In the fifth section, the results of the collected data are evaluated, interpretation of the results and future work are included.