Published Paper Details:

Face recognition (FR) is one of the most extensively adopted biometric approach for identity verification. The applications of face recognition can be seen in various sectors such as, army, business, public security etc. However, there systems were not able to detect the faces accurately when covered with scarf and masks, which degrade their overall performance. In this paper, an effective approach is proposed to detect the occluded faces with different head pose angles. To do so, the images are selected from the available dataset or can be taken in real time through camera. These selected input images are then analyzed carefully to detect the various face points such as eye, nose, mouth etc. Once the faces are detected from the input images they are converted to gray scale and HSV color images, so that their statistical features like mean, standard deviation, variance, skewness and kurtosis can be evaluated. These statistical features obtained from gray scale and HSV colored images are then concatenated together to form a single feature matrix. The feature matrix is then given to the CNN model for training and testing purpose. The performance of the proposed CNN model is simulated in the MATLAB environment in terms of various performance factors such as accuracy, sensitivity, specificity, precision, recall and Fscore. The simulation outcomes obtained proved that the proposed CNN model is effective and accurate in identifying the masked faces.