Please wait a minute...
Front. Inform. Technol. Electron. Eng.  2016, Vol. 17 Issue (6): 516-526    DOI: 10.1631/FITEE.1500235
    
Unseen head pose prediction using dense multivariate label distribution
Gao-li Sang, Hu Chen, Ge Huang, Qi-jun Zhao
State Key Laboratory of Fundamental Science on Synthetic Vision, College of Computer Science, Sichuan University, Chengdu 610064, China; College of Mathematics and Information Engineering, Jiaxing University, Jiaxing 314001, China
Download:   PDF(0KB)
Export: BibTeX | EndNote (RIS)      

Abstract  Accurate head poses are useful for many face-related tasks such as face recognition, gaze estimation, and emotion analysis. Most existing methods estimate head poses that are included in the training data (i.e., previously seen head poses). To predict head poses that are not seen in the training data, some regression-based methods have been proposed. However, they focus on estimating continuous head pose angles, and thus do not systematically evaluate the performance on predicting unseen head poses. In this paper, we use a dense multivariate label distribution (MLD) to represent the pose angle of a face image. By incorporating both seen and unseen pose angles into MLD, the head pose predictor can estimate unseen head poses with an accuracy comparable to that of estimating seen head poses. On the Pointing’04 database, the mean absolute errors of results for yaw and pitch are 4.01° and 2.13°, respectively. In addition, experiments on the CAS-PEAL and CMU Multi-PIE databases show that the proposed dense MLD-based head pose estimation method can obtain the state-of-the-art performance when compared to some existing methods.

Key wordsHead pose estimation      Dense multivariate label distribution      Sampling intervals      Inconsistent labels     
Received: 23 July 2015      Published: 06 June 2016
CLC:  TP391.4  
Cite this article:

Gao-li Sang, Hu Chen, Ge Huang, Qi-jun Zhao. Unseen head pose prediction using dense multivariate label distribution. Front. Inform. Technol. Electron. Eng., 2016, 17(6): 516-526.

URL:

http://www.zjujournals.com/xueshu/fitee/10.1631/FITEE.1500235     OR     http://www.zjujournals.com/xueshu/fitee/Y2016/V17/I6/516


基于稠密多变量标签的“连续”头部姿态估计方法

目的:精确的头部姿态估计对于人脸相关的应用,如人脸识别、视线估计、情感分析等具有重要意义。大多数现有的人脸姿态估计方法仅能对训练数据库包含姿态的情况进行估计。为实现对训练数据库不包含姿态的情况进行预测,有学者提出了基于回归的头部姿态估计方法。然而,这些基于回归的方法虽然可以预测连续的姿态,但是却很少有相关的系统性性能评估。
方法:针对训练数据库不包含姿态的估计问题,本文提出使用稠密多变量标签分布表示人脸姿态。通过给样本分配稠密化的多变量标签,可以实现对数据库不包含姿态的情况进行较为准确的估计。
结论:本文方法在Pointing’04数据库上的yaw和pitch方向分别取得了平均绝对误差4.01°和2.13°。此外,在CAL-PEAL,Multi-PIE等公开库上的实验表明,本文方法在训练数据库包含姿态上的预测性能也优于其他比较先进的方法。

关键词: 头部姿态估计,  稠密多变量标签分布,  角度间隔,  不一致性标签 
[1] Ying Cai, Meng-long Yang, Jun Li. Multiclass classification based on a deep convolutional network for head pose estimation[J]. Front. Inform. Technol. Electron. Eng., 2015, 16(11): 930-939.