Efficient face and facial expression recognition model

This paper contains a brief explanation of the model used to recognize faces and predict the condition such as happy, sadness, fear, anger, disgust, natural, surprise. Local Directional Number Pattern (LDNP) technique is proposed to extract the features of the face and categorized with one of the algorithms for classification. LDNP encrypts directional information of the facial texture in a compact way to eight different directions through calculating the edge responses in the neighborhood, by choice most positive and negative directions of those edge responses to produce a descriptor. Positive and negative directions provide sufficient information on the structure of the neighborhood. It allows distinguishing the intensity changes for example from light to dark and dark to light. The LDNP is a method more compact than previous methods in which they can extract many facial features that integrate with each other into a feature vector, which is then, used it as a face descriptor. This approach is extended in this work to improve recognition accuracy. In the existing method, Support Vector Machine (SVM) with LDNP approach which used for facial expression. The performance of SVM is destitute. In the proposed model Feed Forward Neural Network (FFNN), is used to improve the accuracy performance.