Int J Soc Robot (2011) 3:233–241 DOI 10.1007/s12369-010-0089-0 Determination of Age and Gender Based on Features of Human Motion Using AdaBoost Algorithms S. Handri · S. Nomura · K. Nakamura Accepted: 28 December 2010 / Published online: 11 January 2011 © Springer Science & Business Media BV 2011 Abstract Automated human identification by their walk- ing behavior is a challenge attracting much interest among machine vision researchers. However, practical systems for such identification remain to be developed. In this study, a machine learning approach to understand human behav- ior based on motion imagery was proposed as the basis for developing pedestrian safety information systems. At the front end, image and video processing was performed to separate foreground from background images. Shape-width was then analyzed using 2D discrete wavelet transformation and 2D fast Fourier transformation to extract human mo- tion features. Finally, an adaptive boosting (AdaBoost) al- gorithm was performed to classify human gender and age into its class based on spatiotemporal information. The re- sults demonstrated the capability of the proposed systems to classify gender and age highly accurately. Keywords Human motion · Gender and age · AdaBoost · Shape feature 1 Introduction Recently, many researchers have been interested in using computer vision to analyze images involving humans. That S. Handri ( ) · S. Nomura Top Runner Incubation Center for Academia-Industry Fusion, Nagaoka University of Technology, Nagaoka, Niigata, Japan e-mail: handri.santoso@kjs.nagaokaut.ac.jp S. Nomura e-mail: nomura@kjs.nagaokaut.ac.jp K. Nakamura Department of Management and Information Systems Science, Nagaoka University of Technology, Nagaoka, Niigata, Japan e-mail: nakamura@kjs.nagaokaut.ac.jp application is well-known as “Looking at people”. The abil- ity to recognize humans and their activities by vision is the key for a machine to interact intelligently and effort- lessly with human-inhabited environment. One of the gen- eral goals of artificial intelligence has been to design ma- chines which act more intelligently or human-like [1]. How- ever, current developments provide only a partial solution for machine to be truly intelligent and useful. Thus, the abil- ity to perceive and to extract the information independently was required for machine, rather than rely on information supplied to it externally such as by keyboard or other man- ual input. The machine which has the capability to retrieve infor- mation, perceive information and interpret information will make interaction for human easier. Such systems are reli- able, however, only if the systems are able to detect and determine pedestrian behavior. Automatically monitoring pedestrian behavior via computers could conceivably work in continuously monitoring and operating the surveillance systems. Similar to the uniqueness of fingerprints or faces, ways of walking are believed to be unique to an individual. This argument was based on considerable evidence in biome- chanics, psychology and literature research that people can be recognized by the way they walk or, in short, gait. There were some attempts to recognize person based on the way of their walk [2–25]. In general, gait recognition approaches can be classified into two main classes, i.e., appearance-based approaches and model-based approaches. Appearance-based approach was derived from the spa- tiotemporal pattern of a walking person, while model-based approaches aims to explicitly model human body of motion, and they usually perform model matching in each frame of a walking sequence so that the parameters are measured on the model [26]. One of prospective application by using