摘要
在表情识别中,阈值都是以经验给出的,没有一种定量的标准来计算得出,而且阈值一旦给出就不能改变,不能随着情况的不同而变化。为了解决这一问题,提出了一种阈值的自适应调整机制。根据表情样本训练时的错分值和训练识别时的错分值来得到阈值自适应调整函数,阈值依据此函数来自动的调整。实验结果表明,本文提出的阈值自适应调整方法在表情识别中取得了很好的效果,降低了错误分类,提高了识别的可靠性。
The threshold in the facial expression recognition depends on operators experience and ability. Once the threshold is selected, the value is hard to change with different cases. In order to solve this problem, a method for self-adaptive threshold regulation is proposed. The regulation function is obtained based on two main parts. One is the value of the wrong classification in sample training, and other one is the value of the wrong classification in recognition training. So the threshold can be adjusted by the regulation function. Experiments show that the method is effective. The data indicate that the proposed method achieves good performance by recognition rates and recognition reliability.
出处
《数据采集与处理》
CSCD
北大核心
2009年第4期531-535,共5页
Journal of Data Acquisition and Processing
基金
国家自然科学基金(60674092)资助项目
江苏省高科技研究(BG2006010)资助项目
关键词
阈值
自适应调整
表情识别
可靠性
threshold
self-adaptive regulation
expression recognition
reliability
