Derya Ozkan, Louis-Philippe Morency: “Self-Based Feature Selection for Nonverbal Behavior Analysis”

August 24, 2010 | Istanbul, Turkey

Speaker: Derya Ozkan, Louis-Philippe Morency
Host: 20th International Conference on Pattern Recognition (ICPR 2010)

One of the key challenge in social behavior analysis is to automatically discover the subset of features relevant to a specific social signal (e.g., backchannel feedback). The traditional approach for feature selection focuses on finding the relevant behaviors from a dataset made of multiple human interactions. The problem with this group-based approach is that it oversees the inherent behavioral differences among people (e.g., culture, age, gender) by focusing on the average model. In this paper, we present a feature selection approach which first looks at important behaviors for each individual, called self-based features, before building a consensus. To enable this approach, we propose a new feature ranking scheme which exploits the sparsity of probabilistic models when trained on human behavior problems. We validated our self-based approach on the task of listener backchannel prediction and showed improvement over the traditional group-based approach. Our technique gives researchers a new tool to analyze individual differences in social nonverbal communication.