CAS(ME)2: A database of spontaneous macro-expressions and micro-expressions

Abstract

Micro-expressions are facial expressions that are characterized by short durations, involuntary generation and low intensity, and they are regarded as unique cues revealing one’s hidden emotions. Although methods for the recognition of general facial expressions have been intensively investigated, little progress has been made in the automatic recognition of micro-expressions. To further facilitate development in this field, we present a new facial expression database, CAS(ME)2, which includes 250 spontaneous macro-expression samples and 53 micro-expression samples. The CAS(ME)2 database offers both macro-expression and micro-expression samples collected from the same participants. The emotion labels in the current database are based on a combination of Action Units (AUs), self-reports of every facial movement and the emotion types of the emotion-evoking videos to improve the validity of the labeling. Baseline evaluation was also provided. This database may provide more valid and ecological expression samples for the development of automatic facial recognition systems.