Emotion Recognition Model Based on Facial Expressions, Ethnicity and Gender Using Backpropagation Neural Network

Introduction

No aspect of our mental life is more important to the quality and meaning of our existence than emotions. They are what make life worth living, or sometimes ending. So it is not surprising that most of the great classical philosophers had recognizable theories of emotion (William & Klaus, 2009).

While facial expression is the key to the knowledge of feelings and emotions, we need to detect these emotions from facial expressions to reach the main objective and goal of developing and improving the interaction between computer and human, and make it naturally specially it is the latest challenges in human-robot interaction (Terrence, Illah, & Kerstin, 2002).

In fact, there are wide ranges of human emotions, which can express the internal emotion; these emotions can be divided into two groups basic and secondary. The basic emotions consist of six emotions happy, sad, fear, disgust, angry and surprise, and the other emotions are secondary emotions (e.g., skeptical). In addition, each group is divided into two sets, positive and negative. The basic emotions classifies happy as positive emotion and all other emotions as negative emotions (Miyamoto & Uchida, 2010).

Many studies try to improve the interaction between computer and human by recognizing the human emotions using face expressions (Fischer, 2004; Karthigayan et al., 2008; Ratliff & Patterson, 2008; Waller, et.al, 2008), speech (Douglas-Cowie et al., 2003; Rudra et al., 2008) or etc. (Horlings, 2008).

Gender is one of the most important factors in emotion detection systems. Gender differences were important in a variety of cultures and ethnic groups. The gender was even more important than ethnicity in shaping people’s emotional ideology, emotional experience, and habits of emotional expression. Men and women may share roughly the same sorts of emotional experiences in their relationships, but they differ in how freely they express their feelings. Men tended to express their positive and negative emotions less frequently and less intensely than they were experienced. Women tended to be somewhat more direct in their emotional expression (Elain et al., 2009).

An ethnic group (or ethnicity) is a group of people whose members identify with each other, through a common heritage, often consisting of a common language, a common culture (often including a shared religion) and an ideology that stresses common ancestry or endogamy (Giger, Davidhizar, Purnell, Harden, Phillips, & Strickland, 2007). In general, emotional communication is generally more accurate among people who share similar cultural backgrounds (Beaupré, 2005; Elfenbein & Ambady, 2002). Appearance of emotions are universal across individuals as well as human ethnics and cultures, but the way of emotion expression is that vary from one ethnic group to another and the differences cross-cultural in the familiarity effect on recognition accuracy for different types of expressions. This is commonly known in biology (Beaupré, 2005), but never tests on computer systems.

In this paper, we use human face elements to build a recognition model for recognizing the human emotions based on facial feature considering various ethnic groups and gender using backpropagation neural network. We also study the effect degree of the ethnic group and gender on accuracy of emotion recognition models.

First some related works that build some models of human emotion recognition are presented. We then describe our proposed model and discuss how to implement each stage of our model. We present and explain the experiments afterwards. Next we present the results and we evaluate our model. Finally, the conclusion is in the last section.