JURNAL INFORMATIKA ISSN: 1978-0524 Vol. 12, No 2, July 2018, pp. 71-75 71 DOI: 10.26555/jifo.v12i2.a12742 W : http://journal.uad.ac.id/index.php/JIFO E : jifo@uad.ac.id Facial recognition using deep learning Abdulrazak Yahya Saleh a,1,* , Kirthanaa A/P Jiva Rattinami a,2 a Faculty of Cognitive Science and Human Development, Universiti Malaysia Sarawak, 94300, Sarawak, Malaysia 1 ysahabdulrazak@unimas.my* I. Introduction Facial expression is one of the non-verbal communication methods to understand the mood/mental state of a person [1][2]. Facial recognition technology work utilizes several measurements and technologies to scan faces, including thermal imaging, 3D face mapping, cataloguing unique features, analyzing geometric proportions of facial features, mapping distance between key facial features, and skin surface texture analysis [3]. The aim of this research, presented in the article, is to recognize four basic emotional states: happy, sad, angry, and disgust based on facial expressions. Convolutional Neural Network can improve the accuracy performance in related tasks and few recent works on facial expression recognition successfully utilize Convolutional Neural Network for feature extraction and inference [4][5][6][7][8]. II. Method This section reviews the important foundation of developing Convolutional Neural Network and discusses how the algorithm has been utilized for enhancement. In the first part, an introduction of the Convolutional Neural Network has been presented. The second part focuses on the data set used in this research. Finally, the third part focuses on the proposed method. A. Convolutional Neural Network (CNN) Convolutional Neural Network (CNN) one of the most popular techniques used in image recognition and computer vision systems today. The historical roots traced back to the 1980s, when Kunihiko Fukushima proposed a neural network architecture inspired by the feline visual processing system [9][10][11][12][13][14].The primary aim of Convolutional is to extract features from image where preserves spatial relationship between pixels by learning image features using Small Square of input data [15][16][17]. Added on, Convolutional Neural Network composed of learnable weights and biases. Each neuron receives several inputs, takes a weighted sum over them, pass it through an activation function and responds with an output [18][19]. There are five structures of the Convolutional Neural Network which are convolutional layer used to detect features, non-linear layer, pooling or down-sampling layer decreases the number of weights and controls over-fitting, flattening layers function as set up data for classical neural network, and fully-connected layer is the standard neural network used for classifications [20][21]. Moreover, in the article indicted by [22] stated there are four fundamental functions of the Convolution Neural Network. First, input layers customarily hold the pixel values of the image. Then, Convolutional Neural Network decides the output of neurons, which are connected to local regions of the input through their weight and connected region. Withal verbally expressed that rectified linear unit (ReLu) apply activation function such as sigmoid to the output of the anterior layers. Next, pooling layer function as down-sample gives input dimensionally and reduces the number of ABSTRACT In this article, the researcher presented the results of recognition of four emotional states (happy, sad, angry, and disgust) based on facial expressions. A deep learning method with a Convolutional Neural Network algorithm for recognizing problems has been proven very effective way to overcome the recognition problem. A comparative study is carried out using MUAD3D dataset from Faculty of Computer Science and Information Technology, Universiti Malaysia Sarawak for evaluating accuracy performance of this dataset. More discussion is provided to prove the effectiveness of the Convolutional Neural Network in recognition problems. Keywords: Classification Facial Recognition Convolutional Neural Network