Abstract: In this paper, we applied an Elliptic Fourier analysis to bust slice of human body. First, the Elliptic Fourier description of closed planar contour were introduced, as well as the normalization of position, scale and orientation of the contour; Then bust slices of 459 females and 107 males were extracted from 3D scan data and fitted using Elliptic Fourier. The fitting errors with different maximum harmonic number were also analyzed. A new method was proposed to build symmetric bust slice, and the average shapes of bust slice under different bust girth were calculated. By applying principal component analysis to Elliptic Fourier descriptors of bust slice, the main modes of bust slice variation were discussed in detail. Finally, a BP neural network was built to recognize the gender of a bust slice with Elliptic Fourier descriptors. The test results showed high identification rates.

Key words: Elliptic Fourier, bust slice, shape analysis, gender recognition, BP neural network