Gender Recognition via Fused Silhouette Features Based on Visual Sensors

This paper investigates a novel method for gait gender recognition using a visual camera sensor. Numerous recent studies have focused on visual gait analysis involving a single gait energy image (GEI), yet these methods lose temporal movement information, which results in inferior accuracy under complex conditions. Therefore, in this paper, we introduce a new subGEI map by calculating from fewer frames than a gait cycle, extract the synthetic optical flow of multi subGEIs as temporal features, and utilize a two-stream CNN to combine use the GEI and the optical flow information for further gait analysis. In addition, this paper adopts three kinds of CNN networks, namely, VGG-16, Inception-V3, and a newly designed CNN and develops three types of subGEIs for comparison experiments. The experimental results on gait dataset CASIA show that: 1) the two-stream network produces better performance than that of the CNN-based methods; 2) the combination of temporal and spatial features significantly improves visual gait performance; and 3) the visual based approaches have great advantages than traditional sensors-based methods.