A person's social intelligence impacts their physical and mental health, and the productivity levels of the individuals involved, for example, in workplace interactions. To promote successful social interactions, this dissertation explores the use of sensor technology and machine learning algorithms to monitor and quantify nonverbal behavior indicators in real time. This dissertation conducts extensive convergence research between psychology, communication science, and engineering and establishes a new real-time human/group interaction monitoring platform. From sensor selection to data collection and algorithm design, existing human behavior monitoring systems vary widely in the type of methods employed for their design. Many of these systems were trained with data collected in controlled environments, making them not practical for real-life scenarios. Moreover, existing systems lack the capabilities needed to recognize behaviors in a manner that could support machine-augmented social intelligence. To address these issues, the developed human/group interaction monitoring platform combines a real-time enabled multi-sensor system with a machine learning framework that establishes training and algorithm design methods for behavior recognition. Methods for the execution of human studies, collection of natural human behavior data, and data annotation procedures were also established to train machines to recognize human behaviors impacting the quality of social interactions. The contributions of this dissertation, which can be universally applied to other behavior studies, will advance the design of human behavior monitoring systems for group interactions and facilitate future real-time feedback to increase self-awareness and promote successful social interactions.
Read
