Ethnicity Classification Based on a Hierarchical Fusion

In this paper, we propose a cascaded multimodal biometrics system involving a fusion of frontal face and lateral gait, for the specific problem of ethnicity classification. This system performs human ethnicity classification first from the cues of gait recorded by a long-distance camera and requires next classification using facial images captured by a short-distance camera only when gait based ethnicity identification fails. For gait, we use Gait Energy Image (GEI), a spatio-temporal compact representation of gait in video, to characterize human walking properties. For face, we extract the well-known Gabor feature to render the effective facial appearance information. Experimental results obtained from a database of 22 subjects containing 12 East-Asian and 10 South-American shows that this cascaded system is capable of providing competitive discriminative power on ethnicity with a correct classification rate over 95%.