Metsis, VangelisNgu, Anne Hee HiongSaeednejad, Mahya2022-06-012022-06-012021-05Saeednejad, M. (2021). <i>Real-time movement classification and analysis from RGB – D video data</i> (Unpublished thesis). Texas State University, San Marcos, Texas.https://hdl.handle.net/10877/15833Physical therapy is a form of rehabilitative treatment that includes specially designed and prescribed exercises to help patients recover from diseases that disturb their movements of daily life or enhance their physical abilities. In physical therapy, tracking the movements of different body parts and classifying the movements during the exercises has great importance. The goal of this thesis is to create an interactive system able to recognize physical therapy exercises from patients’ movements and count the repetitions of them in real-time. To achieve this objective, we used a combination of a Long Short-Term Memory (LSTM) model and Dynamic Time Warping (DTW) algorithm for the classification and counting the repetitions of exercises using 3D skeleton tracking data captured by Microsoft Kinect. First, we developed methods for preprocessing and normalization of a multi-dimensional skeleton sequence. Second, we explored ways of increasing the precision of offline human movement classification methods; finally, we created a real-time system for exercise recording and classification. In addition, we proposed a 3D DTW algorithm for 3D skeleton sequence pattern matching and counting the repetition of movements, and we designed and implemented an interactive application that tracks the actions of an individual, transfers information for processing, and displays the results.Text54 pages1 file (.pdf)enComputer scienceDeep learningReal-time classificationThesis