The Multi-Event-Class Synchronization (MECS) Algorithm
This work addresses the need for modeling non-verbal multimodal signals in Human-Computer Interaction, representing an incremental advancement in synchronization analysis.
The paper tackles the problem of measuring synchronization between events from different classes in multiple time series, proposing the Multi-Event-Class Synchronization (MECS) algorithm, which extends event synchronization techniques to handle intra-class and inter-class synchronization, macro-events, and macro-classes, and demonstrates its application on artificial and real-world multimodal behavior data.
Synchronization is a fundamental component of computational models of human behavior, at both intra-personal and inter-personal level. Event synchronization analysis was originally conceived with the aim of providing a simple and robust method to measure synchronization between two time series. In this paper we propose a novel method extending the state-of-the-art of the event synchronization techniques: the Multi-Event-Class Synchronization (MECS) algorithm. MECS measures the synchronization between relevant events belonging to different event classes that are detected in multiple time series. Its motivation emerged from the need to model non-verbal multimodal signals in Human-Computer Interaction. Using MECS, synchronization can be computed between events belonging to the same class (intra-class synchronization) or between events belonging to different classes (inter-class synchronization). In the paper we also show how our technique can deal with macro-events (i.e., sets of events satisfying constraints) and macro-classes (i.e., sets of classes). In the last part of the paper, we apply the proposed method to two types of data i) artificial and 2) real-world case study concerning analysis of human multimodal behavior.