An Ada-Random Forests based grammatical facial expressions recognition approach

The automatic recognition of facial expressions has a tremendous impact in many research fields, especially in the field of sign language since facial expressions contribute towards the formation of grammatical structure of the language that reduce the ambiguity of the sign language understanding. However, the automatic recognition of grammatical facial expressions is still a very challenging task due to the signer-based variation of the grammatical facial expressions, and the co-occurrence of manual and non-manual signs. This paper presents a novel Ada-Random Forests framework for recognizing the grammatical facial expressions used in Brazilian sign language. In this approach, an Ada-Boost feature selection algorithm is applied to select compact feature subsets from the numerous raw extracted features to reduce the computational time as well as to improve the recognition rate of the system; then, selected features are fed to a robust random forests classifier, given their capability to handle high-dimensional and unbalanced data, to recognize the grammatical facial expressions. The evaluation results of the experiments conducted on the first publicly available benchmark data set on Brazilian sign language indicate that the proposed technique improve the recognition metric by as much as 7.5% over the previously applied technique.