Gender recognition method based on silhouette, footstep, and foot pressure measurements for counting customers

Conventional methods of gender recognition cannot be applied to counting store customers in terms of both data acquisition conditions and processing time. To solve this problem, the authors proposed a gender recognition method based on human silhouette shape characteristics. The current study demonstrated for the first time how efficient the footstep characteristic and foot pressure characteristic are for gender recognition. It discussed the recognition technique based on joining of the mentioned characteristics with the silhouette characteristic. These were joined by the one-dimensional compression based on the Fischer discrimination of characteristic vectors arranged in the order of characteristic rates for these three types. Based on the results of clothing-related research aimed at department store customers, a population group consisting of 45 male and 45 female participants was gathered to simulate distribution of various types of clothing and footwear. The authors conducted recognition tests with this population group and examined distributions by types of clothes and footwear in the characteristic space. The results showed that (1) for each separate characteristic rate the recognition rate which used to be 80% increased up to 90% due to use of joint characteristics, (2) there were numerous samples that featured correct recognition by the footstep characteristic or foot pressure characteristic even with erroneous recognition by the silhouette characteristic, which demonstrated effectiveness of the proposed combination of the three types of characteristic rates. © 2002 Wiley Periodicals, Inc. Electron Comm Jpn Pt 2, 85(8): 54–64, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjb.10066