Addressing the cold-start problem in facial expression recognition. Jorro-Aragoneses, J., L., Díaz-Agudo, B., & Recio-García, J., A. In 23rd International Conference ¡, ICCBR 2015, volume 9343, pages 197-211, 2015. Springer.
abstract   bibtex   
In our previous research [5] we proposed a CBR approach to infer the\nemotional state of the user through the analysis of a picture taken from\nthe front facing camera of her mobile device. We demonstrated that\ndifferent people express emotions with different gestures and got the\nbest accuracy using a personal case base with self pictures of the same\nuser. However, in the cold start situation, where pictures of the\nquerying user are not available, the CBR system uses a generic case base\n(GCB) made of pictures of anonymous people. Although the performance\nusing the GCB was acceptable on average there were several users with a\nvery low accuracy. In this paper we compare our GCB to other reference\npicture catalogues and evaluate our CBR approach with state-of-the-art\nFacial Expression Recognition (FER) algorithms. Results point out that\nour approach is only suitable for GCB including semantically similar\nusers. We use an ontology to group together users with similar\ndemographic and physiological information: sex, age and ethnic group. We\nevaluate our CBR approach with small and specialized case bases where\npictures are semantically similar to the target population and\ndemonstrate that it efficiently increases the accuracy in the cold start\nsituation and minimizes the noise in the case base.
@inproceedings{
 title = {Addressing the cold-start problem in facial expression recognition},
 type = {inproceedings},
 year = {2015},
 pages = {197-211},
 volume = {9343},
 publisher = {Springer},
 id = {8fc352e1-c2e8-3baf-9fa3-15ad6b299edc},
 created = {2019-06-24T13:44:22.494Z},
 file_attached = {false},
 profile_id = {fa40a510-5b8d-3050-adba-9fab4285e880},
 group_id = {ae46f76f-8a49-34b5-8b7c-d25295ac4e0c},
 last_modified = {2019-06-24T13:44:22.494Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {true},
 hidden = {false},
 source_type = {inproceedings},
 private_publication = {false},
 abstract = {In our previous research [5] we proposed a CBR approach to infer the\nemotional state of the user through the analysis of a picture taken from\nthe front facing camera of her mobile device. We demonstrated that\ndifferent people express emotions with different gestures and got the\nbest accuracy using a personal case base with self pictures of the same\nuser. However, in the cold start situation, where pictures of the\nquerying user are not available, the CBR system uses a generic case base\n(GCB) made of pictures of anonymous people. Although the performance\nusing the GCB was acceptable on average there were several users with a\nvery low accuracy. In this paper we compare our GCB to other reference\npicture catalogues and evaluate our CBR approach with state-of-the-art\nFacial Expression Recognition (FER) algorithms. Results point out that\nour approach is only suitable for GCB including semantically similar\nusers. We use an ontology to group together users with similar\ndemographic and physiological information: sex, age and ethnic group. We\nevaluate our CBR approach with small and specialized case bases where\npictures are semantically similar to the target population and\ndemonstrate that it efficiently increases the accuracy in the cold start\nsituation and minimizes the noise in the case base.},
 bibtype = {inproceedings},
 author = {Jorro-Aragoneses, Jose L. and Díaz-Agudo, Belén and Recio-García, Juan A.},
 booktitle = {23rd International Conference ¡, ICCBR 2015}
}

Downloads: 0