ABSTRACT

302Recent advances in pattern recognition have allowed engineers and scientists to jointly address automatic analysis of human behavior via computers. This chapter explores a number of different aspects and open challenges in the field of human behavior analysis (HBA) in the context of Big Data. Human behavior is recognized in different levels with the high-level knowledge in addition to the context. Human pose recognition, action recognition, and interaction recognition are considered as different levels, depending on the number of actors involved. The goal of this chapter is to recognize human behavior analysis in different situations by employing various learning models. Here, the data have been received from the different cameras/locations in streaming; thus, we need an effective learning algorithm that can deal with such situation. Over time, new data become available and the decision structure needs to be revised accordingly. Therefore, different learning models have been identified and studied in the machine learning community under several hierarchies called generative learning, discriminative learning, imitative learning, graphical models, etc. As a result, this chapter addresses the different learning models with respect to HBA for large video data sets, as well as the pros and cons of each learning model. This work also discusses the state-of-the-art learning methodologies that are required to analyze the behavior with respect to Big Data. This chapter also discusses the applications and presents a case study with respect to behavior analysis in large volumes of data.