CONCLUSIONS
Establish a model to allow separating semantic knowledge from spatial and temporal parameters
Separate the motion analysis from recognizing object
Combine other useful information
Current technique is a time-consuming process
