Multi Modal Gender Recognition for Gender-Based Marketing Using Depth Camera

Abstract

This research is conducted to prove that gender recognition by computer which can be done in real time by using depth camera. Gender recognition can be used on many industries, such as security, marketing, and other sectors. The purpose of this research is to detect gender by using images of user (RGB image) and voice. Furthermore, gender-based marketing is used for the implementation of this system. By using multi modalities, the result is more accurate than only using one factor. Image processing algorithm is used on processing facial image, which is Linear Discriminant Analysis (LDA) algorithm. Furthermore, gender can also be detected by special frequency of each gender speech. Autocorrelation is one of the methods that is able to detect pitch from detected audio. Kinect for Windows v2 was carried out as visual and audio sensor. This research proved that gender can be detected by using those modalities with right algorithm. Several problems are also found during the experiments, such as input data problem, not matching algorithm, and small percentage of accuracy. In conclusion, detecting gender can be done by computer (real time or not) and several ideal conditions must be made to get proper and high accuracy result, such as person distances from camera, lighting, image size.