Analysis of the neural network recognition characteristics of 6 basic facial expressions

In order to develop an “Active Human Interface” that realizes heart-to-heart communication between intelligent machine and human being, we have been investigating the machine recognition of human emotions from facial expressions. By using a neural network(NN), we obtained a high recognition ratio of about 90% for 6 basic facial expressions, but we find that, depending on the NN learning information, there exist certain characteristics between correct/incorrect recognition ratios. This paper analyses these characteristics by using the action units (AUs) used in the psychological field for describing the basic muscle movement of human face and analyzing the relationship between the NN input and output information. We found, in this paper, how AUs influenced the recognition results by assigning O value to each AU using NN input information