Veri Bilimi
Yazarlar: Nevzat OLGUN, İbrahim TÜRKOĞLU
Konular:Mühendislik
Anahtar Kelimeler:Gender Classification,Laser,LSTM,Sex Classification,Signal Processing
Özet: Gender classification can provide significant advantages in applications with access control, marketing activities and biometric verification processes. In cases where the entries to some areas are only male or female, advertising products according to the number of male and female in the store or reducing the database usage burden by primarily gender discrimination in biometric verification can be given as examples of gender classification practices. Gender classification is a binary classification problem as male or female. In traditional methods, gender classification has been made from facial images. One of the biggest difficulties in gender classification from facial images is that the person's face cannot be kept in a certain position, while other is the difficulties in the imaging stage. The desire of the person to hide herself from the cameras, differences in the face and lighting conditions can be given as examples of the difficulties of the image-based methods. In this study, we propose gender classification with low-power laser beams instead of traditional camera-based method of gender classification. In the experimental study conducted for this purpose, a low-powered laser beam is projected onto the subjects 'arm for a short period of time from a distance of 2 m, and laser signals reflected from the subjects' arm are recorded. Laser signals reflected from the arm of subjects are classified according to the LSTM deep learning architecture after data preparation, and the subjects' gender is determined. An average classification success rate of 76.20% was achieved as a result of the gender classification study in which 6 men and 6 women between the ages of 19 and 38 participated. The results show that gender classification can be performed with laser signals. Another advantage of this method is that the arm can be easily positioned at the desired location during the receiving signal from the person's arm.