Gender Classification Based on Support Vector Machine with Automatic Confidence

In this paper, we propose a support vector machine with automatic confidence (SVMAC) for gender classification based on facial images. Namely, we explore how to incorporate confidence values introduced in each primitive training sample and compute these values automatically into machine learning. In the proposed SVMAC, we use both the labels of training samples and the label confidence which projected by a monotonic function as input. The main contribution of SVMAC is that the confidence value of each training sample is calculated by some common algorithms, such as SVM, Neural Network and so on, for gender classification. Experimental results demonstrate that SVMAC can improve classification accuracy dramatically.

• Type: Journal
• Year: 2009
• Where: ICONIP
• Authors: Zheng Ji, Bao-Liang Lu
• Keywords: Confidence Value, Gender Classification, ICONIP 2009, Information Technology, Training Sample