| Popis: |
Gait recognition is a behavioral biometric system that has attracted a great deal of interest on the part of researchers recently. Several technologies such as CCTV camera and Kinect sensors have been used to recognize or to identify persons via their behavioral activities. However, using different image processing techniques and selecting various features to achieve high accuracy remains a challenge. This paper focuses on using a Kinect sensor for extracting gait signatures to be used for the process of gait identification. A set of triangles are created based on 20 joint points, which are provided by Kinect sensor. The uniqueness of this paper is the extracted feature based on the generated triangles using the triangle area and the three angles of each triangle, entitled the triangular grid feature (TGF). Furthermore, the extracted feature vector is a high dimensional feature set. Therefore, to reduce the dimensions of the feature vector, principal component analysis (PCA) followed by linear discriminant analysis (LDA) are applied as dimension reduction methods. Finally, Linear Discriminant Classifier (LDC) and k-Nearest Neighbors (k-NN) are used separately to test the performance of the proposed methods. The results obtained from the set of experiments shows that the proposed method achieves a significant result, and provides 90% and 87.3% in terms of recognition accuracy using k-NN and LDC, respectively. |
