Lakoza S. Inertial system for estimation of human motion parameters

The thesis describes that monitoring of motor activity in the diagnosis and treatment of neurological diseases movements in sports medicine during rehabilitation, movements detecting in virtual reality requires use of objective measuring instruments such as inertial system for estimation of human motion parameters. It is described the physical principles and sensor systems which are used for development of systems for estimation of human motion parameters. Their main advantages and disadvantages are characterized. The basic approaches, which provide ISEHMP development, are considered. It is adduced a number of tasks that need to consider during development of ISEHMP. The paper describes the concept of the human skeleton biomechanical model and its using in ISEHMP and in methods for estimating of human position in the open space. Calibration issues of system for estimation of human motion parameters are analyzed. It is highlighted several stages: calibration of segment linear dimensions of skeleton biomechanical model; calibration of IMU's frame orientation relative to segment frame in static poses; functional joint axes definition by performing certain types of motor activity; system calibration using closed kinematic chain in skeleton biomechanical model. It is shown that problems of functional joint axes calibration for ISEHMP are currently solved for most significant cases and mainly related to peculiar qualities of human body functioning for specific biomechanical research. It is described the algorithm for estimation of the individual body segments orientation using IMU data. It is shown that most of the algorithms ignore the presence of dynamic errors and external disturbances. It is shown that in the known literature hasn't considered the problems of cross-linking between channels correction of AHRS, the application of SINS algorithms for estimation of human motion parameters. The thesis describes the used human skeleton biomechanical model, key joints are defined, it is described joints' degrees of freedom, and it is associated appropriate frames with segments. In work AHRS algorithms are developed using Kalman filtration for orientation parameters such as direction cosine matrix. It is formulated, theoretically grounded and in way of simulation confirmed the principle of correction channels separation for AHRS. It is got simplified error model of AHRS. AHRS' errors analysis shows the necessity of using SINS algorithms to estimate human motion parameters. In the study it is done the development of complex ISEHMP algorithm invariant to segments' accelerated movement. Correction signals are generated using human skeleton biomechanical model. It is completed an optimal mathematical model of accelerometers using method of group accounting of arguments. By result of experiments was found that the optimal model by regularity criterion should consider offset, scale factor and two coefficients that consider installation errors and sensors' cross-sensitivity. Allan variations were used to estimate the parameters of sensors' noise mathematical model that are used for synthesis of an extended Kalman filter algorithm for sensors calibration. Sensor' noises include a flicker noise, random walk on acceleration, velocity random walk, and ramp noise. The paper discusses the issues of using scalar calibration for the autonomous final finding of parameters of accelerometers and magnetometers mathematical model. It is obtained graphs and performed their analysis which showed the conditions and capabilities of scalar calibration for low accuracy sensors. It is shown sensitivity of scalar calibration to the presence of high level cross-linking in the sensor signal. Experimental study is done for parameters of sensor included in the IMUs of ISEHMP. It is got the values of sensors calibration matrix. These values are used as etalons for comparison. Resulting sensors errors are given. It has been shown that for value of sensor' model errors lower 10% it can be improved the calibration accuracy of accelerometers and magnetometers using method of scalar calibration iteratively. For accuracy improvement of sensor' mathematical model parameters estimation it is proposed to form a resultant signal sensor model combining scalar calibration results and standard calibration results performed on high-precision equipment. It is proposed for scalar calibration process convergence to perform write-off of previously estimated sensors' offset level. It has been investigated the accuracy of the AHRS algorithm developed for ISEHMP. The simulation of developed ISEHMP's complex algorithm invariant to limbs acceleration is done. Modeling scheme is described. Stand-imitator has been done to verify the accuracy and correctness of embedded ISEHMP algorithms. Stand mimics part of the upper limb. The stand allows to measure angles of rotation and flexion-extension. It has been done series of field experiments for ISEHMP accuracy checking. In thesis it has been experimentally shown that using of complex algorithm in ISEHMP compared to using AHRS algorithm makes it possible to reduce system error. Keywords: estimation of motion parameters, inertial system, AHRS, calibration, skeleton biomechanical model.