**6. Conclusions**

A low-computational cost system for children emotion recognition in an unobtrusive way was proposed in this study, which is composed of low-cost cameras, making possible its extension for research into developing countries. As a first stage, our proposal was tested on visual and thermal

images of children interacting with the mobile social robot N-MARIA, achieving promising results (85.75% of accuracy) to locate specific face landmarks, as well as to recognize (or infer) five emotions. All children had a hopeful interaction with the robot, which demonstrated our system is useful to stimulate positive emotions in children, and able to trigger a profitable interaction with them. In future works, this proposal will be integrated in N-MARIA, aiming to know online the children's emotion and making control decisions based on emotions. Additionally, other methods will be explored for facial tracking, in order to reduce the influence of head-pose during the emotion recognition. Furthermore, unsupervised methods for automatic label assignments and classifier learning will be evaluated in our dataset to obtain a robust recognition system for processing patterns of high uncertainty, such as facial expressions and emotions.

**Author Contributions:** Conceptualization, C.G., C.V., D.D.-R. and D.F.; Methodology, C.G. and D.F.; Software, D.F., V.B., A.F., G.B., C.V. and D.D.-R.; Formal Analysis, C.G. and D.D.-R.; Writing—Original Draft Preparation, C.G., D.D.-R., C.V. and D.F.; Writing—Review and Editing, E.C. and T.B.-F.

**Funding:** This research was funded by FAPES/Brazil, gran<sup>t</sup> numbers 72982608 and 645/2016.

**Acknowledgments:** Authors would like to thank the financial support from CAPES, CNPq and FAPES/Brazil (project numbers: 72982608 and 645/2016), and UFES for the technical support.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
