Gait Databases

Naturally, the success and evolution of a new application relies largely on the dataset used for evaluation. The early gait databases were collected about 10 years before the time of writing. Then, computers had little power and memory costs were comparatively high. Clearly, this was before digital video and acquisition was based on analogue camcorder technology which resulted in frames being digitized individually. Since techniques were in their infancy, as in face recognition, early databases only had few subjects. The idea then was to determine whether recognition could be achieved at all — or not. At that stage we were not interested in the ramifications of recognition. There were two early databases which were developed independently: the UCSD data was recorded outdoors and the Southampton data was indoors, with subjects wearing special trousers. The current databases are considerably more advanced, but certainly benefited in their development for the early approaches.