利用图像距离的监督LLE头部姿态估计方法

局部线性嵌入(LLE)是一种经典流形学习方法,直接应用这种非监督的传统LLE估计图像中的头部姿态存在两点不足:未考虑图像像素空间信息和未利用样本标记信息.因此,本文结合图像欧式距离和偏置LLE流形学习方法,对头部姿态图像降维,并通过广义回归神经网络(GRNN)和多元线性回归的方法,估计头部图像的姿态.在FacePix头部姿态数据库的对比实验表明,本方法具有较好的头部姿态估计效果.

Head Pose Estimation Based on Supervised LLE with Image Euclidean Distance

The Locally Linear Embedding(LLE) is a classical manifold learning algorithm.This unsupervised traditional manifold learning algorithm can be introduced to head pose estimation,but there are two disadvantages: neither considering spatial information of image pixels nor using pose information of the face images.Biased manifold embedding is combined with Image Euclidean Distance(IMED) to compute embedding,when the embedded,the poses of test samples are estimated with general regression neural network(GRNN) and multi-variate liner regression.The comparative experiments on FacePix database shows that the proposed method gets better head pose estimation.