Object tracking based on hierarchical temporal memory classification

Abstract

With recently advances in technology (hardware and software) there is an interest of humanity to have machines that behave like humans do. One aspect that researchers have to overcome is how to imitate the cognitive processes of the brain; cognitive processes like visual pattern recognition, speech recognition, space comprehension and so on. This task needs an algorithm that receives raw information from the environment, thus a signal processing method is needed to convert the raw input into useful information. Computer Vision is an interesting eld of research, because the process of capturing images is simple and the hardware to process these images is available with current technology. A natural cognitive process is tracking objects, for example humans (and animals) can focus on an object of their interest and follow it with their eyes; humans do this very easily but it is a very challenging problem for computers. This research focuses on the eld of video tracking process using an emerging technology like Hierarchical Temporal Memory (HTM). HTM is a machine learning technique that tries to imitated the Neocortex of the human brain, and then emulate cognitive processes. This research is based on creating a video tracking algorithm that tries to imitate the cognitive process of the brain. Di erent approaches have been developed to face the video tracking problem, this research was done using HTM network to achieve this purpose.