Kinematics Modeling of Human Motion using System Identification Technique Ahmad, R., Yaacob, M.S., Che Omar, M.B. Department of Applied Mechanics, Universiti Teknologi Malaysia Abstract Image processing techniques from motion captured images are accurate and cost effective method to give a set of data that defines the location of specified limb at every sequence of human motion. From this set of data, system identification was done to model the human motion. This project is a study on how performance of a model is influenced by the type of model whether it is a linear model or non-linear model and a single variable model or multi variable model. Two types of parameter estimator was used which were the least square estimate and recursive least square estimate. The study also was conducted to see how the number of lags can give effects to the model. The objective is to formulate a predictive model to analyze human motion. Simulation was done through the model to see the result and performance of model whether it can be a model for human motion representation. 1. Introduction Walking and running are the most important and common human movement. Developing a computer model of the motion has been the goal of many researches. It is a challenge to model the human motion. Motion analysis entails measurement, analysis and assessment of the movement features that are associated with the walking or running task. Significant technical and intellectual progress has been made in the area of motion analysis over the past new decades, especially because of the advance in computing speed which in turn has aided the development of more advanced movement recording system that require less data processing time. Improved computing speed has also made feasible and inspired increasingly complex and innovative motion data analysis technique [1]. This field has been studied by many professional from varies discipline such as science, engineering, sports or medicine. Each discipline has own research interests and motivations of study. An example where work in this area can be used is biomechanics field include a gait pattern classification and recognition task including categorization of normal and pathological gaits. The field of human motion study can be divided into two sections; kinematics and kinetics. In this project, only the kinematics motion of lower limbs was studied. Kinematics describes the pattern and temporal aspects of motion such as positions, angles, velocities, and accelerations of body segments and joints during motion [2]. It is the study of body motion without reference to the forces that causes the motion. The objectives of this project are: 1. To apply the video digitizing and image processing technique with capability to capture images of real human motion (two dimensional) and process the images to give a set of data that defines limbs location at each sequence of the motion. 2. To determine a good model structure and parameter estimation method that gives adequate and high predictive model of human motion. 3. To simulate the human motion model. Other methods that available for human motion data acquisition are footswitches, gait mats, goniometer, accelerometer or magnetic tracking system [3]. Rezaul et al. [4] control of limb movements using a kinematics movement plan by artificial neural network. It used a model of feed-forward neural network where the input consisted of kinematics profiles of the lower limb and the individual muscle activations formed the output vectors. However, data acquisition method was not mentioned. Kinematics and muscle activity data is from a total of eleven different walking conditions. This paper is organized as follows. In the second section, it is an overview of human motion and the data acquisition technique that has been conducted. The third section is a discussion about the system identification while the fourth section is the discussion about the developed model of human motion. The final Second Asia International Conference on Modelling & Simulation 978-0-7695-3136-6/08 $25.00 © 2008 IEEE DOI 10.1109/AMS.2008.178 338