![]() | يوجد فقط 14 صفحة متاحة للعرض العام |
المستخلص This thesis aims to enhance the bag of features representation to improve the accuracy of human activity recognition. The k-means step in the bag of features is enhanced using multilevel clustering. This is achieved by two and three level clustering. The experimental results show that the multilevel clustering reduces the time, memory requirements and improves the recognition accuracy to reach 97.7 % on the KTH dataset and 98.9 % on the Weizmann dataset |