Abstract

Activity recognition systems automatically recognize the daily activities of residents in smart home environments. Their development has been the subject of numerous studies, indicating a well-established area of research. Developing HAR systems for practical use encounters numerous challenges. Deep learning approaches require labeled datasets from the target environment that are large enough. Acquiring sufficient activity labels is an expensive and time-consuming task. The problem could be addressed by transferring knowledge from other environments. This process is known as transfer learning. In the chapter, some scenarios from the activity recognition environments, which illustrate these conditions, are described. Before knowledge transfer can be utilized, some issues need to be addressed. Source and target environments can differ to a great extent and need to be unified. This chapter discusses methods for handling differences. The chapter reviews some influential works in the field of transfer learning for activity recognition with a greater focus on recently published literature.