Effects of Facial Mood Expressions on Face Biometric Recognition System's Reliability

Facial changes due to the user's emotional condition or facial mood expressions can affect the reliability of facial recognition system. Emotionally expressive faces can be typically classified by seven classes which are angry, disgusted, fearful faces, happy, normal, sad and surprised ones. In this paper we have presented one by one comparison between all samples from different moods in order to answer the following questions: 1- if the difference between samples matching score with different emotional facial expressions is statically significant? (Which mood can put the reliability of system in risk more than other ones?) 2- For each mood, what are the best and worst comparison scenarios from the biometric system's perspective? In order to answer the first and second questions, images with different expressions from Jaffe Database have been used. For calculation of matching scores, two state of the art methods, the face recognition dlib -deep cnn based - code and commercial code -Verilook- have been implemented. The obtained results showed in comparison with neutral faces, the mood influenced pictures can lead a statistically significant change in matching score. We have also reported that, for each mood which mood can be the least and most probable error source.