Abstract
For a robot to interact with humans around its environment is necessary to achieve a communication that seems as natural as possible. Humans when talking to each other, look directly into their faces to indicate attention. A robot, on the other hand has to mimic this interaction by turning its point of attention towards the face of the human.
In this project a system for face detection on a mobile phone is presented. The sensor used for the input of the image is a mobile phone camera. The reason to use a mobile phone is the price and accessibility of the equipment. In addition, a mobile phone has integrated technologies such as Bluetooth and Java programming environment that are useful for the development of this project.
A research on the existing face detection algorithms is done. The bases for Java programming and description of the technologies to use are presented. Moreover, the implementation is proposed with three approaches: Detection of Dark Areas, Motion analysis and Skin color detection. Finally, the results are compared to various situations of luminance and position. With these implementations the algorithm can find faces with an accuracy rate of 87%.
The analysis of the image generates a coordinate (x, y) of the face location in the image. This coordinate may be sent to a robot and this turn towards the face. This results in an intuitive way of interaction between humans and robots. In other words, I am creating "the eyes2 of a robot for human-robot interaction.