Some considerations of age estimation method for “Augmented TV” based on posture of gripping tablet PC

Abstract

Augmented TV is an augmented reality system for making TV video to appear to come out of the screen. With Augmented TV, a television program is viewed through the camera of a tablet PC, and related content such as three-dimensional computer graphics (3DCG) is overlaid and displayed on the tablet PC. To provide users with age-targeted content, we have developed a height-estimation method based on the posture estimation method of Augmented TV, in which the height of the user is estimated from the elevation of the tablet PC off the ground. The estimated user height is then used to estimate his/her age. A public testing was carried out with around 90 elementary school students, and the data acquired was used to formulate a relationship between the elevation of the tablet PC and the height of the user. In the age estimation method proposed, the most statistically likely age for a given height was used. Therefore, we created a table containing the most statistically likely age for a given height, and the age of the user was estimated using the table. We evaluated the accuracy of both the height and age-estimations based the results of the public testing. The error in the age estimation based on the height-estimation was about 2 years of age. A rate of the age estimation in the case of using the age estimation method was about 60%. In addition, developing a prototype application, we confirmed the effectiveness of both methods.