In this thesis, we aim to contribute to studies regarding the relationship between Anterior Cruciate Ligament (ACL) and the valgus position of the knee, by tracking 3D coordinates of key knee joints in multiple human subjects when performing physical activities. These subjects are instructed to perform dynamic physical protocols under the monitoring of 2-3 Azure Kinect sensors. It is hypothesized that the valgus position may have an intrinsic relationship in distinguishing an injured ACL from a healthy ACL, and the establishment of this relationship would allow for more accurate assessments of knee rehabilitation treatments, especially in athletes. To analyze the valgus position, we extract raw data from our Azure Kinect setup and transform it into Point Clouds. Afterwards, background filtering and registration of all sensors is performed. In the end, an end-to-end neural network based on Point Cloud data is trained in a supervised fashion to estimate spatial positions of key knee joints, allowing the valgus position to be tracked over time.