A Comparative Analysis of the Micro-Facial Movement Detection Accuracy Depending on the Fuzzy Spatio-Temporal Feature Descriptor Selection

The problem of the true human emotion recognition from his or her facial micro-expressions is considered. The experiments with the help of the micro-facial movement detection software pipeline proposed in the previous research works are carried out. In the present research work the FHOOF and FHOFO spatio-temporal descriptors are selected to implement the feature extraction stage. The dependencies of the micro-facial movement detection accuracy on adjustable parameters of these algorithms are investigated. The comparative analysis of the obtained accuracy in dependence of the selected versions of the FHOOF or FHOFO descriptor, the types of its input blocks, and the classifiers is performed.