The purpose of this thesis is to implement and study selected methods to increase the quality of the process of human gait analysis on the basis of data collected with the Kinect v2 depth sensor. In addition, the scope of the thesis includes the creation of a number of tools that provide the possibility of saving data received directly from the device and its subsequent processing.

As part of the thesis, two methods for improving the quality of data obtained from the Kinect v2 were examined. The first method is based on applying an appropriately selected Butterworth filter. Filter parameter selection was based on the data received from the device. The second data quality improvement method is based on smoothing approximation of the time course of the hip joint angle. Taking advantage of the fact that the course of the angle between the joints over time has an approximately sinusoidal shape, the thesis tested whether by approximating the waveform of this angle with a sinusoidal function and then appropriately approximating the position of the knee joints we are able to reduce the measurement error. To test the selected methods, the thesis used data recorded by Kinect v2 and the Vicon system simultaneously.

As part of the thesis, an application with a graphical user interface implemented using WIN32 API was created to record and read data from the Kinect v2 depth sensor. In addition, a set of scripts in MATLAB language was created to preprocess the data collected from the Kinect v2 and bring it to the desired format by appropriately rotating the coordinates of the recorded points.