Gender Prediction by Gait Analysis Based on Time Series Variation of Joint Positions
Ryusuke Miyamoto, Risako Aoki
In this paper, a novel gender prediction scheme based on a
gait analysis is proposed. For the gait analysis, we propose a
novel feature extraction scheme that uses the time series vari-
ation in the joint positions directly. Here, normalization by
linear interpolation is adopted to set the number of samples
of a walking period as the same constant for all target hu-
mans. The classifier for gender prediction is constructed with
a support vector machine using the feature extraction scheme.
To evaluate our proposal, we carried out an experiment for
gender prediction using six male and six female humans who
are in their twenties. The experimental results show that the
classification accuracy is 99.12% when three-dimensional co-
ordinates are used directly for feature extraction and 99.12%
if two-dimensional features are used in the best case. Full Text
|