1.1 Background of the Study
Gender recognition is one of basic face analysis tasks. Automatic classification of gender from frontal face images which was taken under artificial conditions has been well studied with remarkable results (Lemley et al., 2016). It is a basic task for human beings because most common functions depend on the correct gender discernment. Automatic gender classification is employed in applications like intelligent user interface, visual surveillance, aggregating demographic statistics for promotions etc. Human faces provide necessary visual information for gender discernment.
Gender determination from facial images has received much research attention in the last twenty years (Shan, 2011).
Gender classification is one of the biometrics methods to identify individuals by the features of the face. The gender of an individual is classified by visual observation of images which is difficult with computer vision (Jaswante et al., 2014). Biometrics can be used as a behavioural trait in identifying and verifying an individual. However, the biometric solutions to used depends on user acceptance, security, accuracy, cost and implementation time.
Because biometric attributes of an individual are unique and it is not transferrable, this gives it edge over traditional authentication techniques. (Jaswante et al., 2014).
Gender classification is a binomial classification problem that is used to determine if an image is a man or a woman. While it is not challenging for humans to determine the gender of an image, it is a big task for computers (Fida and Abu, 2013). Performance of applications like face recognition, age classification can be improved with the output of gender classification.
Gender recognition is a pattern recognition problem. There are two broad classes of pattern recognition they are, one stage pattern recognition and two stage pattern recognition systems. In One stage pattern recognition system input data is classified directly, while two stage pattern recognition systems first perform feature extractor, followed by classification (Ardakani, 2016). To learn a gender, a feature extractor is used to extract features and then a classifier is applied to classify the gender.
The part of the human body that humans use to recognize themselves is the face; this makes facial images the most common biometric characteristics being used for human verification and identification. Facial images have gained high relevance because of its use in several places such as airports, criminal identification systems and security systems. The pre-processing step for classification of gender is face detection. Classification is an effective data mining technique that can be used to identify hidden affinities and distinctive pattern in large dataset, other data mining techniques that can be used for same purpose are clustering and regression. The detection of regularities and affinities in different parameters of a dataset serves as a useful tool for decision making and drawing predictions (Muhammad et al., 2013).
1.2 Statement of the Problem
Research has shown that the difference between a masculine and feminine face can be used to improve performances of facial recognition applications in biometrics, human-computer interaction (HCI), surveillance and computer vision. How to deal with the effect of things like illumination, facial expression, pose etc. is the challenge in a real – world environment, which is the same challenge with the development of a face-based gender classification system that has high classification accuracy. Hence, there is need to develop a gender determination system using support vector machine (SVM) and K-Nearest Neighbour to distinguish between male and female.
1.3 Aim and Objectives
The aim of this project is to develop a facial gender recognition system using support vector machine (SVM).
The specific objectives are as follows:
i. design a gender recognition system using support vector machine and K-Nearest Neighbour
ii. simulate the developed system in (i) using MATLAB
iv. performance evaluation of the systems using recognition accuracy, true positive (TP), true negative (TN), false positive (FP), false negative (FN), sensitivity and specificity
v. evaluate the level of significance of the algorithms using one-way ANOVA
1.4 Scope of the Study
This scope of this work is to evaluate existing gender recognition systems and improve on it by using images in an unconstrained element using support vector machine classifier while accuracy, sensitivity and precision will be used as the performance evaluation metrics.
1.5 Significance of the Study
Gender plays a significant role in human interactions and also with computers. Actually, gender classification is a binary classification task which predicts an image whether it belongs to a man or woman. Gender classification output can improve the performance of many applications such as face recognition, age classification and so on.
To achieve the listed objectives, the following research approaches were taken:
i. An in-depth study of various gender classification techniques
ii. Improvement on existing systems using an uncontrolled environment
iii. Highlighting and reporting of the weakness of the approaches
iv. Evaluating the approach using standard metrics
2.1 Introduction to Face Recognition
One of the most used human biometric feature for identification is the face. Every human has his/her own unique face. Face recognition is easy for humans but it is a complex task for computers (Jawad, 2007). Face recognition is a research area that is still being explored because of its wide range of relevance, it can be used in areas like surveillance and security, telecommunication and digital libraries, human – computer intelligent interaction, and smart environment.
Faces comprise of the nose, mouth, eyes, etc. and recognition of individuals happens when there is a need to discriminate between the same basic configurations. Every individual must therefore be distinguishable from others because of the manner in which the basic element configuration varies (Young and Bruce, 1991).
Face analysis is a wide research area even when approached only from the computer vision point of view. Face detection and tracking, face recognition, gender classification, facial expression and gesture recognition, age classification, and ethnicity classification belongs to automatic face analysis. There are still many unsolved problems in the topics listed above despite the progress made within the last few years (Yulia Gizatdinova, 2009).
Face recognition works by comparing a face image to other face images saved in a database and then identifying the person in the input face image. Face detection is relevant to recognition because the images must be analyzed and then identified before recognition can happen.