基于嘴巴区域的空间光滑子空间特征微笑检测算法

Smile detection based on spatially smooth subspace features on the mouth region

微笑检测作为一种通过人脸识别人类情绪的重要方法已广泛应用在生活中,大量微笑检测算法被提出以提高微笑检测的准确率。然而在实际应用中计算资源通常有限,减少算法所需的计算时间与存储空间同样重要。不同于现有的利用人脸区域的微笑检测算法,提出一种基于嘴巴区域的微笑检测方法来减少算法运行的时间。使用流形降维算法在嘴巴区域提取具有判别性能的特征,同时结合图像是二维形式的本质特点引入空间光滑子空间学习来提高分类准确率。实验及分析表明基于嘴巴区域的微笑检测算法减少了算法运行所需的时间与特征所需的存储空间,取得了较高的准确率。

As an important method to recognize human emotion,smile detection has been widely used in daily life.A large number of smile detection algorithms have been proposed to improve the accuracy of smile detection.However,in practical application,computing resources are usually limited,so it is important to reduce the computing time and the storage space.Different from the existing smile detection algorithm based on face region,we propose a smile detection method based on mouth region to reduce the computing times.A manifold dimensionality reduction algorithm combining spatial smooth subspace learning is proposed to extract discriminant features for classification.Experiments and analysis show that the proposed algorithm based on mouth region in this paper reduces the computational time and the storage space,and achieves high accuracy.