An Appearance-Based Gender Classification Using Radon Features
Recognizing human gender automatically by a computer is a challenging problem, which attracts research attention due to its vast real-life application. Gender classification has been playing a wide role in security and surveillance system. The proposed sy
- PDF / 572,671 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 6 Downloads / 197 Views
1 Introduction Face is a characteristic feature of the human beings which contains identity, age, and emotions. Gender classification from a person’s face could play an important role in computer vision such as security surveillance systems, search engine, demographic studies, marketing research and performance enhancement (face recognition, smart human–computer interface). In real-world scenario, due to natural reasons, images might be occluded naturally like injury, wearing scarf, or sunglasses because of weather conditions [1] and thus, it becomes difficult to classify gender from such a partial occluded face. In this work, wavelet and Radon transforms are combined together for extracting features to classify male or female from facial information. The basic gender classification system used for our work is shown in Fig. 1 contains mainly three modules, i.e., preprocessing, features extraction, and classifier. In a pre-processing, basically relevant features are taken out, which are the most potential segment of an image. Feature extraction is done by combing wavelet and Radon transforms. The obtained features are then passed to a powerful supervised learning algorithm using SVM to discriminate male and female. This paper is assembled as follows: A review of the past research work in the area of gender classification is given in Sects. 2 and 3 describes our gender classification technique using wavelet and Radon transforms as features extraction and SVM as classifier. Experimental analysis and conclusion are shown in Sects. 4 and 5, respectively.
R. K. Sangha (B) · P. Rai Gyan Ganga Institute of Technology and Science, Jabalpur, India e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2019 R. K. Shukla et al. (eds.), Data, Engineering and Applications, https://doi.org/10.1007/978-981-13-6347-4_15
159
160 Fig. 1 Gender classification system
R. K. Sangha and P. Rai
Face images
Pre -processing
Extraction of features
Classification
Male or Female
2 Review of Literature This chapter presents a review of available and relevant literature. In gender classification, the two main approaches to detect the features of the face are appearance based [2] and geometrical based [3]. The chapter investigates various techniques developed under these methods and finally summarizes the literature review, with the identification of the shortcomings and opportunities provided by the past research. In the prior works [4–7], raw face image had been used for classification, and high classification rate was obtained. Later, new techniques were developed for the extraction of features and classification of gender. Sun et al. [8] used local binary patterns (LBP) histograms along with Adaboost for classification under constrained environment to obtain 95.75% classification rate. Makinen and Raisamo [9] used alignment and nonalignment approaches to compare different gender classification schemes. Caifeng [10] used discriminative LBP-Histogram (LBPH) on LFW to achieve 94.81% classification rate. Berbar [11] use
Data Loading...
