Emotions play a significant role in an individual's social interactions, as they influence perceptions, thoughts, and behavior. These emotions reveal themselves to others through changes in attitude, facial expressions, vocal inflections, gestures, or even postures. The advent of computers has drastically changed the ways of social interaction , now expanding into the digital world. Affective computing encompasses all research attempts to understand and simulate human emotions to improve interactions between humans and machines. This field, blending psychology and artificial intelligence, offers methods for automated recognition, interpretation, and synthesis of affective states. So far, most of the research has focused on facial and vocal expressions. The object of this thesis is to explore the relatively uncharted territory of detecting and producing expressive animations through body movements and postures. It consists of three main contributions: (1) a systematic review of research to identify a list of factors involved in body expression, (2) a method for automatic recognition of body expressions from human motion capture, and (3) a synthesis method that can modulate the emotional expression of any virtual character animation. The approaches proposed in this thesis are characterized by low computation and resources costs, making them real-time functional. They rely on concepts intelligible to a human user and are fully explainable end-to-end, which distinguishes them from recent AI approaches based on deep neural networks, which offer only partial control to the user.