An approach to recognition of human patterns in movement
Alejandro Pazos, A. Rivas-Feal · 2002
The system the authors propose here aims to establish an automatization of the data acquisition process from video sequences in sporting environments. All of this is devoted to perform an analysis of human movement in an affordable way. One has to take into account that if one is working with very narrow environments (e.g. sport movements and specifically cyclic movements), the main problem, that is shape recognition and interpretation from images, reduces the complexity and thus the analysis can be treated in an easier way. Under this condition, a database for the storage of graphic patterns aids the system in situations with a nonclear solution, and at the same time allows one to save the environmental conditions of the image sequence. These conditions can change a lot of times and can make the recognition process difficult. The authors have also used artificial neural networks for pattern association between patterns from the image, in some cases incomplete, and those stored in a database. This process will be preceded, of course, by filtering operations and typical techniques of image processing for improving the features of each frame from the sequence to be recognized.