A machine learning based approach to human observer behaviour analysis in CCTV video analytics & forensics

Human observer behaviour analysis in image and video inspection in many areas of practical application is conducted based on using data captured by eye tracking devices. Such data is analysed using statistical approaches leading to the creation of useful information and the ability to make decisions about the content. CCTV observer behaviour analysis is one example of a most widely used application. Unfortunately, the information and knowledge that such statistical approaches to data analysis can create is rather limited, especially the trends and patterns of data cannot be easily analysed. Thus, important information and knowledge that the data can provide may not be identifiable. In this paper, we proposed a novel approach to human observer eye tracking data analysis based on machine learning algorithms. Further, in order to conduct a more detailed and practically useful data analysis, we specifically analyse the attention human observers given instructions to search for specified content. We provide experimental results to demonstrate the significance and novelty of the information and knowledge that this novel approach to data analysis can provide. To the authors' knowledge, there is no work in literature that has proposed the use of machine learning in eye tracking data analysis.