Zobraziť v EDS

Moth-flame optimization based deep feature selection for facial expression recognition using thermal images.

Uložené v:
Podrobná bibliografia
Názov: Moth-flame optimization based deep feature selection for facial expression recognition using thermal images.
Autori: Chatterjee, Somnath, Saha, Debyarati, Sen, Shibaprasad, Oliva, Diego, Sarkar, Ram
Zdroj: Multimedia Tools & Applications; Jan2024, Vol. 83 Issue 4, p11299-11322, 24p
Predmety: FACIAL expression, ARTIFICIAL neural networks, FEATURE selection, THERMOGRAPHY, CONVOLUTIONAL neural networks, SOCIAL media
Abstrakt: Automatic human facial expression recognition, a challenging research problem, has many real-life applications in the fields of social media, digital marketing, and healthcare, etc. Facial expression recognition using machine learning is a challenging task due to the variability in facial expressions across different people and situations, as well as the impact of lighting conditions and camera angles on the appearance of a face, thereby making it difficult to accurately recognize expressions. In this paper, we developed a two-stage facial expression recognition system using thermal images. In the first stage, we used a lightweight convolution neural network model called MobileNet, to generate features from the input images. In the second stage, we employed a nature-inspired meta-heuristic, called the Moth-flame Optimization algorithm, to select the optimal subset of features obtained from the MobileNet model. To increase the overall performance, a memory-based variant built on the Grunwald-Letnikov approach is designed. The proposed two-stage model is evaluated on the IR database, a publicly available standard dataset on thermal image-based facial expressions. The proposed model efficiently eliminates the redundant features, and the overall model achieves an accuracy of 97.47% on the said dataset while using only 29% features generated from the MobileNet model. [ABSTRACT FROM AUTHOR]
Copyright of Multimedia Tools & Applications is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Databáza: Complementary Index
Popis
Abstrakt:Automatic human facial expression recognition, a challenging research problem, has many real-life applications in the fields of social media, digital marketing, and healthcare, etc. Facial expression recognition using machine learning is a challenging task due to the variability in facial expressions across different people and situations, as well as the impact of lighting conditions and camera angles on the appearance of a face, thereby making it difficult to accurately recognize expressions. In this paper, we developed a two-stage facial expression recognition system using thermal images. In the first stage, we used a lightweight convolution neural network model called MobileNet, to generate features from the input images. In the second stage, we employed a nature-inspired meta-heuristic, called the Moth-flame Optimization algorithm, to select the optimal subset of features obtained from the MobileNet model. To increase the overall performance, a memory-based variant built on the Grunwald-Letnikov approach is designed. The proposed two-stage model is evaluated on the IR database, a publicly available standard dataset on thermal image-based facial expressions. The proposed model efficiently eliminates the redundant features, and the overall model achieves an accuracy of 97.47% on the said dataset while using only 29% features generated from the MobileNet model. [ABSTRACT FROM AUTHOR]
ISSN:13807501
DOI:10.1007/s11042-023-15861-5