Akbar Mirzaei

A study to identify the holistic nature of human facial expression processing

Is face processing “holistic” or the parts of the face are processed independently, and changes in configurations of the parts have no impact on processing? It has been claimed that face is recognized more holistically than other types of objects. In fact, face identification relies not on the recognition of individual features, but on the recognition of those features into a holistic face representation. While few face researches would disagree that the recognition of facial identity is holistic, there is less consensus about whether the perception of facial expression is based on the holistic perception of the entire face or on the analytic perception of particular parts.

To shed some insight into these questions, our approach is to make use of psychology and psychophysics experiments such as composite effect and perform the corresponding experiments in some face processing models. Since neural networks models are also considered cognitive models and they provide us with some degree of biological plausibility, we mostly tend to rely on these models. After using preprocessing methods such as Gabor filters and Principle Component Analysis and by comparing the performance of both results (our results and excremental results in psychophysics) we may propose the answer to our question about the holistic nature of facial expression processing. In order to simulate psychophysics experiments we need to model the attention to the target half of the face. We proposed three methods to model the attention to the target half and compared the results of those to the psychophysics experiments results.

Finally, we would like to know whether psychophysics experiments and our simulation using EMPATH model have the same results or not. Since they have the same results we conclude that the EMPATH model is a biologically plausible model and the facial expression processing is a holistic processing.


A. Mirzaei, R. Hosseini

Simulation-based investigation of psychophysical experiments based on EMPATH model for validating the holistic nature of facial expression recognition

International Conference on Computer Engineering and IT, In persian, 2016