Peking University at TRECVID 2008: High Level Feature Extraction

We participated in one task of TRECVID 2008, that is, the high-level feature extraction (HLFE). This paper presents our approaches and results on the HLFE task. We mainly focus on exploring the data imbalance learning in this year, and propose two methods for this problem: (1) adaptive borderline-SMOTE and under-sampling SVM (ABUSVM), and (2) concept category. Our approach can be divided into two phases: feature representation and data imbalance learning. In feature representation phase, four low-level visual features namely color moment grid (CMG), local binary pattern (LBP), Gabor wavelet texture (Gabor), and edge histogram layout (EHL) are combined together in an "early fusion" manner. In data imbalance learning phase, ABUSVM and concept category are employed jointly to handle the data imbalance problem. In addition, we also investigate the fusion of 2005 and 2008 training data to improve the performance. The experimental results show our four visual features, ABUSVM, and...

• Type: Conference
• Year: 2008
• Where: TRECVID
• Authors: Yuxin Peng, Zhiguo Yang, Jian Yi, Lei Cao, Hao Li, Jia Yao
• Keywords: Concept Category, Data Imbalance, Feature Representation, Internet Technology, TRECVID 2008