Deep neural networks have been recently putting a breakthrough in pattern recognition, machine learning and artificial intelligence. This paper emphasizes a study based on deep learning framework contributing to the field of expression recognition. The proposed model involves a technique using deep for human facial expression recognition. Images are first preprocessed with normalization manipulation to remove illumination and facilitate enhancement using hat-filtering. At that point, a weighted, focus symmetric nearby paired example (CS-LBP) is connected to each face hinder by piece. The CS-LBP pieces are connected to form an element vector of the face picture. The deep network is trained using the layer-wise strategy. We use the CIFAR-10 dataset is used for training and testing. A database of real images is used for testing the algorithms. GUI has been created which compares trained and tested dataset and specifies the type of expression in the command window.