CANNY EDGE DETECTION AND CONTRAST STRETCHING FOR FACIAL EXPRESSION DETECTION AND RECOGNITION USING MACHINE LEARNING

Facial expression recognition in the world is challenging due to various unconstrained conditions. Although existing facial expression classifiers have been almost perfect for analyzing constrained frontal faces, they fail to perform well on partially occluded faces common in the wild. This paper deals with Facial expression detection and recognition through the Viola-jones algorithm and HCNN using the LSTM method to avoid those challenges in recent work. For face detection, basically, utilize the face detection Viola-Jones algorithm and it recognizes the occluded face and it helps to perform the feature selection through the whale optimization algorithm, once after compression and further, it minimizes the feature vector given into the HCNN and LSTM model for efficiently identifying the facial expression. In existing work, the feature extraction stage exact finding of the corner becomes a very difficult task, to solve this issue need to use a corner detection algorithm to enhance the feature extraction (corner points) from the face image. One of the main drawbacks of WOA is that it is not good at exploring the search space. To overcome those issues, the work introduced an improved framework for facial image recognition. In which first edges are detected using the canny edge detection operator. Then improved linear contrast stretching is used for image enhancement. Then in feature extraction, the author proposes Hybrid SIFT with Double d-LBP (Dd-LBP) to obtain the features that are illumination and pose independent. For face detection, utilize the face detection Viola-Jones algorithm and it recognizes the occluded face and significant features are selected by using self-learning chicken swarm optimization, it minimizes the feature vector is given into the Hybrid HCNN and LSTM model for efficiently identifying the facial expression. Experimental results demonstrate the efficiency of the proposed work in terms of accuracy, precision, recall and f-measure.