Name
Abstract | ||
|---|---|---|
This paper presents a partially connected Multilayer Perceptron (PCM) neural network as an optimal new MLP with a supervised algorithm and three hidden layers to detect face emotions. Compared with the traditional MLP, the proposed network shows improvements in speed, accuracy, and computational time. Six emotions have been considered in this study, namely anger, surprise, fear, sadness, normal, and happiness. The image dataset with the fixed background is used to test and train the network. Canny edge detection algorithm is employed to separate regions of the face including eyes, mouth, and eyebrows from the background. A binary image is extracted to represent the areas of eye, mouth, and eyebrow. The feature extraction is carried out by Legendre coefficients of Legendre polynomials which are selected for the best accuracy. Results show that the partial network is faster and has a lower computational complexity compared with the full connection network. It also has a faster convergence with higher accuracy. |
Year | DOI | Venue |
|---|---|---|
2021 | 10.1016/j.neucom.2020.12.070 | Neurocomputing |
Keywords | DocType | Volume |
Partial connection MLP,Facial expression,Polynomial Legendre,Curve fitting | Journal | 434 |
ISSN | Citations | PageRank |
0925-2312 | 0 | 0.34 |
References | Authors | |
0 | 2 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Gholamreza Karimi | 1 | 7 | 4.86 |
| Mehdi Heidarian | 2 | 0 | 0.34 |