This paper describes a large-scale gait database comprising the Treadmill Dataset. The dataset focuses on variations in walking conditions and includes 200 subjects with 25 views, 34 subjects with 9 speed variations from 2 km/h to 10 km/h with a 1 km/h interval, and 68 subjects with at most 32 clothes variations. The range of variations in these three factors is significantly larger than that of previous gait databases, and therefore, the Treadmill Dataset can be used in research on invariant gait recognition. Moreover, the dataset contains more diverse gender and ages than the existing databases and hence it enables us to evaluate gait-based gender and age group classification in more statistically reliable way.

We

We propose a method of speed-invariant gait recognition in a unified framework of model-and appearance-based approaches. When a person changes his/her walking speed, kinematic features (e.g., stride and joint angle) are changed while static features (e.g., thigh and shin lengths) are preserved. Based on the fact, firstly, static and kinematic features are decoupled from a gait silhouette sequence by fitting a human model. Secondly, a factorization-based stride transformation model for the kinematic features is created by using a training set for multiple non recognition-target persons on multiple speeds. This model can transform the kinematic features from a gallery speed to another arbitrary probe speed. Because only the kinematic features are insufficient to achieve a high recognition performance, we therefore finally synthesize a gait silhouette sequence by combining the preserved static features and the transformed kinematic features for matching. Experiments with the OU-ISIR Gait Database show the effectiveness of the proposed method.

Despite the advancement of information and transportation systems, inconvenient pedestrian crossing buttons remain common. In accordance with intelligent transportation systems (ITS), it is necessary to improve pedestrian crossing systems. Therefore, in this study, the proposed system adopts signal gaze, which is more natural compared to pressing a pedestrian crossing button, as a crossing request. A compact camera is inserted in a traffic light to view the other side of the crosswalk. The image data is analyzed in real time to identify all people who have a crossing request. An algorithm with three detectors using Haar-like feature quantities was developed and an evaluation experiment was conducted, considering three conditions: daytime, nighttime, and shade. The detection rate of crossing requests was 100% within 5 s. Although the detection rate was extremely high, there was a problem of incorrectly detecting non-humans. Therefore, in this research, we evaluated the system when detecting non-humans in order to determine the causes. As a result, it became clear that the detection rate changes rapidly depending on the waiting time for a traffic light and also when crossing the crosswalk; however, the system continues to detect the incorrectly detected background.