Development of intelligent techniques for collaborative motion and fall recognition using acceleration and depth sensor data

In order to better understand human movements, these movements must be identified accurately and efficiently, and identified movements must be accurately identified and classified. Falling action, which can cause serious health problems especially in the elderly, must be detected correctly and early and necessary first aid must be provided. For this purpose, in this thesis, intelligent techniques have been developed to detect human movements and to detect the fall event accurately and efficiently in order to provide emergency first aid to falling individuals. Kinect depth data and acceleration data were used in the study. Applications were made via MATLAB platform. Human movements and falls were detected and classified using acceleration data. Kinect depth data was used to confirm the fall to avoid confusion with sudden movements such as jumping and falling action, and to more accurately detect the fall event. First of all, machine learning techniques were used to determine human daily activities from acceleration data. In addition, Long Short Term Memory (LSTM) based deep learning method was used to detect motion from acceleration data and the results of these two methods were compared. It is thought that better results can be obtained if more than one classifier is selected properly from a classifier for motion detection, so particle flock optimization technique has also been applied. From the acceleration data, One Dimensional Local Binary Pattern (1D-LBP) and Extreme Learning Machine (ELM) based fall detection method have been proposed. At the same time, the fall detection technique using control graph and Kinect depth images has been developed. As a result, machine learning algorithms, LSTM based deep learning technique and particle flock optimization technique were used for motion detection. 1D-LBP and ELM methods and control graph method were used for fall detection and effective results were obtained.