Performance analysis for facial expression recognition under salt and pepper noise with median filter approach
Facial expression provides an important behavioural measure for studies of emotion, cognitive processes, and social interaction. Facial expression recognition has recently become a promising research area. In face recognition, the simple process of face recognition system should go through image dat...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 語言: | English English English |
| 出版: |
2013
|
| 主題: | |
| 在線閱讀: | http://eprints.uthm.edu.my/6631/1/24p%20AZRINI%20IDRIS.pdf http://eprints.uthm.edu.my/6631/2/AZRINI%20IDRIS%20COPYRIGHT%20DECLARATION.pdf http://eprints.uthm.edu.my/6631/3/AZRINI%20IDRIS%20WATERMARK.pdf |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | Facial expression provides an important behavioural measure for studies of emotion, cognitive processes, and social interaction. Facial expression recognition has recently become a promising research area. In face recognition, the simple process of face recognition system should go through image data retrieval, face detection, facial feature extraction and face recognition. However, some researches focus on the part of face recognition system, such as face detection, face recognition, or algorithms dealing with certain drawbacks issues such as illumination, occlusion, noise, and angle. Thus, in this research we have considered the facial changes as represented by face emotions from JAFFE Database results for different noise levels. The proposed system consists of three modules. The first module read the images face of three different emotions such as happy, fear and surprise. Those images are flawed with different level of salt and pepper noise. Then filter is applied on the corrupted images with average and median filter. The second module constructs PCA that are responsible for feature extraction, while the third module extracts the features by processing the image and measuring dimensions of PCA using k-NN and NN. Using the proposed classifiers, some experimental results have been obtained. It is found that the highest percentage of accuracy is 71.43 % and 83.96 % for K-NN and NN classifier respectively. |
|---|