Gestures, the visible body movements that are ubiquitous in human behavior, are key elements of natural communication. Understanding them is fundamental to designing computing applications with more natural forms of interaction. Both sign languages and everyday gestures reveal the rich signal capacity of this modality. However, although research is developing at fast pace, we still lack in-depth understanding of the elements that create the underlying symbolic signals. This is partly due to lack of tools for studying communicative movements in context. We introduce a novel approach to address this problem based on unobtrusive depth cameras and developed an infrastructure supporting naturalistic data collection. While we focus on sign language and gestures, the tools we developed are applicable for other types of body based research applications. We report on the quality of data collection, and we show how our approach can lead to novel insights and understanding of communicative movements.