Solving the small sample size problem in face recognition using generalized discriminant analysis

The goal of face recognition is to distinguish persons via their facial images. Each person's images form a cluster, and a new image is recognized by assigning it to the correct cluster. Since the images are very high-dimensional, it is necessary to reduce their dimension. Linear discriminant analysis (LDA) has been shown to be effective at dimension reduction while preserving the cluster structure of the data. It is classically defined as an optimization problem involving covariance matrices that represent the scatter within and between clusters. The requirement that one of these matrices be nonsingular restricts its application to datasets in which the dimension of the data does not exceed the sample size. For face recognition, however, the dimension typically exceeds the number of images in the database, resulting in what is referred to as the small sample size problem. Recently, the applicability of LDA has been extended by using the generalized singular value decomposition (GSVD) to circumvent the nonsingularity requirement, thus making LDA directly applicable to face recognition data. Our experiments confirm that LDA/GSVD solves the small sample size problem very effectively as compared with other current methods.     

Pattern expressions and pattern automata

We define the pattern expressions as an extension of both regular expressions and patterns. We prove several properties of the new family of languages, similar to those of extended regex languages [Câmpeanu et al., Int. J. Found. Comput. Sci. 14 (6) (2003) 1007-1018]. We also define an automata system that recognizes these languages. Differences between regex and pattern expressions are also discussed.     

Experience, Context, and the Visual Perception of Human Movement.

Why are human observers particularly sensitive to human movement? Seven experiments examined the roles of visual experience and motor processes in human movement perception by comparing visual sensitivities to point-light displays of familiar, unusual, and impossible gaits across gait-speed and identity discrimination tasks. In both tasks, visual sensitivity to physically possible gaits was superior to visual sensitivity to physically impossible gaits, supporting perception-action coupling theories of human movement perception. Visual experience influenced walker-identity perception but not gait-speed discrimination. Thus, both motor experience and visual experience define visual sensitivity to human movement. An ecological perspective can be used to define the conditions necessary for experience-dependent sensitivity to human movement. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

基于ANN的声目标识别系统及其DSP实时实现

介绍了一种基于人工神经网络(ANN)的声目标识别系统,概述了用MATLAB专用工具箱对神经网络权值进行训练及仿真的过程,叙述了ANN目标识别系统的数字信号处理器(DSP)实时实现过程,并着重分析定点实现过程中程序变量的定标、非线性运算的实现、溢出的处理等关键步骤.对不同字长的识别结果进行比较表明,基于定点实时实现的系统数据保持很高的精度,可以得到与浮点处理相同的识别率.     

基于DTW算法的语音识别系统实现

动态时间归整(DTW)算法的实现简单有效,在孤立词语音识别系统中得到了广泛的应用.介绍了将DTW算法移植到TMS320VC5402上实现孤立词语音识别的原理、系统硬件组成和软件设计.研究结果表明,系统能满足实时性能要求,识别效果良好.     

高维空间下求解MDA的扰动算法及其在人脸识别中的应用

本文提出了一种在高维空间下直接求MDA(Multiple discriminant analysis)最佳解的扰动算法，并把这一算法应用于人脸的识别中。在传统的MDA求解算法中，一般要求训练样本的个数足够大，以至于其类内散射矩阵为非奇异矩阵，即所谓的“小样本”问题。但是，在入脸识别中，由于人脸空间的维数非常大，而训练样本的个数有限，造成类内散射矩阵为奇异矩阵，从而使得用传统的求解方法失效。为了能在高维空间中求出MDA的最佳解，本文采用扰动的思想，巧妙地避开了矩阵的奇异性问题并找到了最佳变换矩阵。此外，把这一算法用于人脸的识别中，对ORL人脸图像库的实验显示，采用本文提出的算法达到比较低的错误率，其错误率仅为特征脸(Eigenface)方法的49．9％，为Fisher脸(Fisherface)的7914％。     

人脸识别研究进展

人脸识别技术在各种应用的推动下发展很快。而近几年中，为避免各种实际应用环境的限制，非约束环境下的人脸识别也得到了很大的发展。本文概述了人脸识别的一些主要方法，讨论了影响人脸识别的几个主要因素及其处理方法以及基于视频的人脸识别技术，最后指出了鲁棒的人脸识别技术的研究方向和发展前景。     

Effects of stimulus emotionality and sentence generation on memory for words in adults with unilateral brain damage.

Does sentence generation and/or stimulus emotionality enhance verbal memory in patients with neurological impairment? This question was addressed by testing 40 patients with unilateral stroke (20 with left-brain and 20 with right-brain damage) and 20 healthy control participants for recall and recognition of 48 target words. During encoding, emotional and nonemotional words were either presented in sentences (read condition) or used to form sentences (generate condition). Both word emotionality and generative processing improved memory performance in all groups. The authors suggest that a similar influence (i.e., cognitive activation) underlies both of these memory-enhancing effects, although the putative origins of the 2 effects are quite different. Neuropsychological underpinnings and clinical implications of these phenomena are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Real-time control of video surveillance systems with program supervision techniques

In this article, we describe a knowledge-based controlled platform using program supervision techniques. This platform eases the creation and the configuration of video surveillance systems. Several issues need to be addressed to provide a correct system configuration: (1) to choose, among a library of programs, those which are best satisfying a given user request, (2) to assign a correct value for each program parameter, (3) to evaluate performances and to guarantee a performance rate which is satisfactory regarding end-user requirements. This platform is composed of three main components: the library of programs, the knowledge base and the control component. The knowledge is either given by experts or learnt by the system. The control is generic in the sense that it is independent of any application. To validate this platform, we have built and evaluated six video surveillance systems which are featured with three properties: adaptability, reliability and real-time processing.