Face Detection and its Features Extraction using Convolution Neural Network Model

Face emotion is an utmost efficient way of nonverbal communication and indication of sentimental status, mentality and intent. In this thesis, we focused on computer vision-based schemes to identify and recognize the faces and expressions, respectively. In this work a CNN based prototype is presented to estimate the age and to classify the gender of detected faces along with its emotions. The proposed model is implemented in three stages namely a hybrid feature extraction technique to identify the facial expressions by using SVM classifier. A combination of SIFT and Deep learning-based model is presented for identifying facial emotions. Whale optimization algorithm is applied to optimize the performance of face emotion recognition. This model gives a convolution neural network and introduces a novel technique for classification of facial emotions, age prediction and male-female identification. The data set for the analysis is considered from WideResNet architecture and Caffe models. The proposed model is tested on a static emotion identification benchmark dataset, and empirical investigation demonstrates that it improves emotion recognition significantly. An accuracy of 98.79% is reported and compared with the existing methods. The reported model is found to be suitable for many applications which involves human computer interaction with the adoption of emerging technologies AI and ML.