Recurrence Quantification Analysis of Crowd Sound Dynamics

When multiple individuals interact in a conversation or as part of a large crowd, emergent structures and dynamics arise that are behavioral properties of the interacting group rather than of any individual member of that group. Recent work using traditional signal processing techniques and machine learning has demonstrated that global acoustic data recorded from a crowd at a basketball game can be used to classify emergent crowd behavior in terms of the crowd's purported emotional state. We propose that the description of crowd behavior from such global acoustic data could benefit from nonlinear analysis methods derived from dynamical systems theory. Such methods have been used in recent research applying nonlinear methods to audio data extracted from music and group musical interactions. In this work, we used nonlinear analyses to extract features that are relevant to the behavioral interactions that underlie acoustic signals produced by a crowd attending a sporting event. We propose that recurrence dynamics measured from these audio signals via recurrence quantification analysis (RQA) reflect information about the behavioral dynamics of the crowd itself. We analyze these dynamics from acoustic signals recorded from crowds attending basketball games, and that were manually labeled according to the crowds' emotional state across six categories: angry noise, applause, cheer, distraction noise, positive chant, and negative chant. We show that RQA measures are useful to differentiate the emergent acoustic behavioral dynamics between these categories, and can provide insight into the recurrence patterns that underlie crowd interactions.

Evacuation behaviors and emergency communications: An analysis of real-world incident videos

Emergencies such as fires and terrorist attacks pose risks of injuries and fatalities, which can be exacerbated by delayed, ill-informed, or unmanaged responses. Effective emergency communication strategies could be used to better inform people and reduce these risks. This research analyzes videos of real-world emergencies to: (a) identify people's observed behaviors that increase risk during evacuations, and (b) examine which emergency communication strategies might reduce risk behaviors. We analyzed 126 publicly available videos of emergency evacuations in different emergencies (e.g., fire, terror attack, evacuation alarm, perceived threat). We found evidence of three types of risk behaviors (delayed response, filming, running) and four emergency communication strategies (evacuation alarm, staff guiding people to exits, general prerecorded message, live announcement). Our analyses suggest that having staff guide people to exits is the most effective strategy for promoting faster and more effective responses. However, neither live announcements nor pre-recorded messages were associated with delayed responses, while evacuation alarms were associated with more delayed responses than other communication strategies. Although people filming the incident was unrelated to staff interactions, it occurred more with alarms sounding and prerecorded messages, suggesting that these emergency communications might not prevent filming. Compared to no communications, all emergency communication strategies reduced running during evacuations. We discuss the implications of this research for identifying effective emergency communication strategies and reducing risk-increasing evacuation behaviors.

Collective Motion in Human Crowds

The balletic motion of bird flocks, fish schools, and human crowds is believed to emerge from local interactions between individuals, in a process of self-organization. The key to explaining such collective behavior thus lies in understanding these local interactions. After decades of theoretical modeling, experiments using virtual crowds and analysis of real crowd data are enabling us to decipher the 'rules' governing these interactions. Based on such results, we build a dynamical model of how a pedestrian aligns their motion with that of a neighbor, and how these binary interactions are combined within a neighborhood in a crowd. Computer simulations of the model generate coherent motion at the global level and reproduce individual trajectories at the local level. This approach yields the first experiment-driven, bottom-up model of collective motion, providing a basis for understanding more complex patterns of crowd behavior in both everyday and emergency situations.