WO2018187950A1 - 基于核判别分析的人脸识别方法 - Google Patents

基于核判别分析的人脸识别方法 Download PDF

Info

Publication number
WO2018187950A1
WO2018187950A1 PCT/CN2017/080174 CN2017080174W WO2018187950A1 WO 2018187950 A1 WO2018187950 A1 WO 2018187950A1 CN 2017080174 W CN2017080174 W CN 2017080174W WO 2018187950 A1 WO2018187950 A1 WO 2018187950A1
Authority
WO
WIPO (PCT)
Prior art keywords
kernel
discriminant analysis
method based
feature extraction
recognition method
Prior art date
Application number
PCT/CN2017/080174
Other languages
English (en)
French (fr)
Inventor
邹霞
Original Assignee
邹霞
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 邹霞 filed Critical 邹霞
Priority to PCT/CN2017/080174 priority Critical patent/WO2018187950A1/zh
Publication of WO2018187950A1 publication Critical patent/WO2018187950A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition

Definitions

  • the present invention relates to a face recognition method based on kernel discriminant analysis, which belongs to the field of face recognition.
  • Face recognition is a computer technology that achieves the purpose of identity identification by analyzing human facial visual features.
  • the academic community gives a specific definition of face recognition in both broad and narrow sense.
  • Generalized face recognition includes face detection, face representation, face identification, and face expression analysis.
  • narrow face recognition is defined as a technology or system that enables identity confirmation, identity comparison, and identity lookup through facial features.
  • biometrics mainly come from the following aspects: face, retina, iris, palmprint, fingerprint, voice, body shape, habits, etc. Therefore, based on the above, research has focused on identifying faces, retinas, and irises.
  • the advantage of face recognition lies in its natural and friendly characteristics.
  • the so-called natural nature means that human beings also identify and confirm the identity of each other by observing and comparing human facial features.
  • speech recognition and body shape recognition also have natural features, while humans or other creatures usually do not pass fingerprints.
  • Features such as the iris distinguish individuals, so the above feature recognition does not have natural features.
  • the so-called friendliness means that the identification method does not increase the psychological burden of the authenticated person due to special treatment, and thus it is easier to obtain direct and authentic feature information.
  • Fingerprint or iris recognition needs to use special techniques such as electronic pressure sensor or infrared to collect information.
  • the above special collection technology is easy to be discovered, which greatly increases the possibility that the authenticated person avoids identity identification and reduces the efficiency of identity authentication.
  • face recognition can directly obtain the face information of the authenticated person through simple image or video technology. This information collection method is not easy to be perceived, and the authenticity and reliability of the information are increased.
  • an object of the present invention is to provide a face recognition method based on kernel discriminant analysis, comprising the following steps:
  • Step 1 Perform feature extraction, map the sample data to the high-dimensional kernel space, and perform the first feature extraction by using the principal component analysis method;
  • Step 2 Construct a centralization matrix H to solve the characteristic equation.
  • Step three calculating a vector
  • Step 4 extracting the main component, completing the first facial feature extraction, and obtaining the sample Y retained after the principal component analysis of the face data;
  • Step 5 performing secondary feature extraction on Y by linear discriminant analysis
  • Step 6 For the test data set X', project it into the feature subspace of the training set, and obtain the test data set ⁇ ' after the feature extraction;
  • Step 7 The sample Z′ is classified and identified by the nearest neighbor classifier.
  • KPCA is used for feature extraction.
  • a given pair of training set data ⁇ [ ⁇ ⁇ , ⁇ 2 , . . . , ⁇ ⁇ ], calculate a kernel matrix ⁇ , among them
  • the vector calculated in step three above is
  • the first k principal components whose total contribution rate is 90% or more are extracted.
  • the above step 5 is specifically calculating the optimal projection direction W according to the criterion function, and projecting Y into the LD.
  • the face recognition method based on kernel discriminant analysis provided by the present invention considers the system structure of the entire node macroscopically, and performs overall architecture design, which has miniaturization, low power consumption, high stability, Easy to expand and low cost, low complexity design features.
  • Embodiments of the invention provides a face recognition method based on nuclear discriminant analysis, and the present invention will be further described in detail in the following embodiments in order to clarify and clarify the objects, technical solutions and effects of the present invention. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
  • the kernel method is a kernel-based pattern recognition method, which is generated along with the support vector machine theory.
  • the idea of the kernel method is to make some linear methods have the ability to process nonlinear information through nonlinear mapping, so that the nonlinear relationship in the image can be better analyzed.
  • the face recognition method for kernel discriminant analysis provided by this embodiment includes the following steps:
  • KPCA is used for feature extraction
  • the sample data is mapped to the high-dimensional kernel space
  • the first feature extraction is performed by principal component analysis.
  • X [ X l , X 2 , ..., x M ]
  • test data set X' For the test data set X', project it into the feature subspace of the training set to obtain a test data set Z' after feature extraction.
  • the face recognition method based on kernel principal component analysis provided by the invention can greatly shorten the identification of the daytime, and by using the nuclear method, the principal component analysis method and the linear discriminant analysis method can not be made up.
  • the use of nonlinear information in the data is a shortcoming.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Image Analysis (AREA)

Abstract

一种基于核判别分析的人脸识别方法,包括:进行特征提取,将样本数据向高维核空间映射,并利用主成分分析法进行第一次特征提取;构造中心化矩阵H,求解特征方程;计算向量;提取主成分,完成第一次人脸特征提取,得到人脸数据主成分分析后保留的样本Y;利用线性判别分析法对Y进行二次特征提取;对于测试数据集X',将其投影至训练集的特征子空间中,得到特征提取后的测试数据集Z';通过最近邻分类器,将样本Z'进行分类识别。上述基于核主成分分析的人脸识别方法,能够大幅缩短识别时间,通过运用核方法,弥补了主成分分析法和线性判别分析法不能利用数据中非线性信息的缺憾。

Description

说明书 发明名称:基于核判别分析的人脸识别方法 技术领域
[0001] 本发明涉及一种基于核判别分析的人脸识别方法, 属于人脸识别领域。
背景技术
[0002] 人脸识别是通过分析人类脸部视觉特征来达到身份鉴别目的的一种计算机技术 。 学术界对人脸识别给出了广义和狭义两方面的具体定义。 广义的人脸识别包 括人脸检测 (face detection) 、 人脸表征 (face representation) 、 人脸鉴别 ( face identification ) 、 表情分析 ( face expression analysis )
以及物理分类 (physical classification) 等一系列相关技术; 而狭义的人脸识别则 被定义为一种技术或系统, 这一技术或系统能够通过人脸的特征进行身份确认 、 身份比较和身份査找。
[0003] 目前, 由于人脸识别技术能够通过生物体 (一般特指人) 本身的生物特征来区 分个体, 提高了生物体识别的精度, 因此, 该技术得到了广泛关注和推崇, 使 该领域也成为了生物识别特征研究中的热点。 以人类为例, 生物特征主要来自 于以下方面: 脸、 视网膜、 虹膜、 手掌纹、 指纹、 语音、 体形、 习惯等, 因而 基于上述内容, 研究则被重点放在了识别人脸、 视网膜、 虹膜、 手掌纹、 指纹 、 语音、 体形、 键盘敲击、 签字等相应特征的计算机识别技术上, 并取得了具 有重要意义的成果。
[0004] 人脸识别的优势在于其自然性和友好性的特点。 所谓自然性, 是指人类本身也 是通过观察和比较人类脸部特征来辨别和确认对方身份的, 如语音识别、 体形 识别等也同样具有自然性的特征, 而人类或其他生物通常不通过指纹、 虹膜等 特征区别个体, 因此上述特征识别就不具有自然性的特征。
[0005] 所谓友好性, 是指该识别方法不因特殊对待而增加被鉴别人的心理负担, 并且 也因此而更容易获取直接和真实的特征信息。 指纹或者虹膜识别需要利用电子 压力传感器或红外线等特殊技术手段采集信息, 上述特殊的采集技术易被人发 现, 大大增加了被鉴别人躲避身份鉴别的可能性, 降低了身份鉴别的效率。 [0006] 然而, 人脸识别却可通过简单的图像或视频技术直接获取被鉴别人的人脸信息 , 这种信息采集方式不易于被人察觉, 增加了信息的真实性和可靠性。
技术问题
[0007] 虽然人脸识别技术具有上述优点, 但该技术的实现却并不容易。 主要受人脸的 生物特性所限制, 具体表现在:
[0008] 第一, 由于同种类型的人脸的结构都具有较高的相似性。 该特点可以用于人脸 定位, 但是却大大增加了利用人脸特征鉴别个体的难度。
[0009] 第二, 受年齢、 情绪、 温度光照条件、 遮盖物等因素的限制, 人脸的外形很不 稳定, 甚至在不同观察角度, 人脸的图像特征也存在显著的差异, 增加了人脸 识别技术应用的复杂性。
[0010] 为使人脸识别技术更好的服务于所需领域, 则需要对上述两项限制进行研究寻 求突破。
问题的解决方案
技术解决方案
[0011] 鉴于上述现有技术的不足之处, 本发明的目的在于提供一种基于核判别分析的 人脸识别方法, 包括以下步骤:
[0012] 步骤一、 进行特征提取, 将样本数据向高维核空间映射, 并利用主成分分析法 进行第一次特征提取;
[0013] 步骤二、 构造中心化矩阵 H, 求解特征方程。
[0014] 步骤三、 计算向量;
[0015] 步骤四、 提取主成分, 完成第一次人脸特征提取, 得到人脸数据主成分分析后 保留的样本 Y;
[0016] 步骤五、 利用线性判别分析法对 Y进行二次特征提取;
[0017] 步骤六、 对于测试数据集 X', 将其投影至训练集的特征子空间中, 得到特征提 取后的测试数据集 ζ' ;
[0018] 步骤七、 通过最近邻分类器, 将样本 Z'进行分类识别。
[0019] 优选的, 上述步骤一利用 KPCA进行特征提取。
[0020] 优选的, 上述步骤一对给定的 Μ个训练集数据 Χ=[Χ Ι2,...,Χ μ], 计算核矩阵 Κ , 其中
[数]
:【 :·(^: '
[0021] 优选的, 上述步骤三计算的向量为
[0022] [数]
Ι 舅::
Figure imgf000004_0001
Figure imgf000004_0002
[0023] 优选的, 上述步骤四提取的是总贡献率达到 90%以上的前 k个主成分。
[0024] 优选的, 上述步骤五具体为根据准则函数计算最佳投影方向 W, 将 Y投影到 LD
A的最佳投影方向 W
[0025] Z ;=W Ty ;
[0026]
[0027] [0026]进而完成了 KFDA特征提取, 形特征子空间, 并得到训练样本数据集 Z 发明的有益效果
有益效果
[0028] 相比现有技术, 本发明提供的基于核判别分析的人脸识别方法, 从宏观上考虑 整个节点的系统结构, 进行整体架构设计, 具有微型化、 低功耗、 高稳定性、 易扩展性以及低成本、 低复杂度的设计特点。
本发明的实施方式 [0029] 本发明提供一种基于核判别分析的人脸识别方法, 为使本发明的目的、 技术方 案及效果更加清楚、 明确, 以下举实施例对本发明进一步详细说明。 应当理解 , 此处所描述的具体实施例仅用以解释本发明, 并不用于限定本发明。
[0030] 核方法是一种基于核的模式识别方法, 它伴随着支持向量机理论的提出而产生 。 核方法的思想是通过非线性映射, 使得一些线性方法具有处理非线性信息的 能力, 进而对图像中的非线性关系能够更好的进行分析处理。
[0031] 本实施例提供的核判别分析的人脸识别方法, 包括以下步骤:
[0032] (1)首先利用 KPCA进行特征提取, 将样本数据向高维核空间映射, 并利用主成 分分析法进行第一次特征提取。 对给定的 M个训练集数据 X=[X l,X 2,...,x M], 计算 核矩阵 K, 其中
[数]
(1^》、
[0033] (2)构造中心化矩阵 Η, 求解特征方程。
[0034] 3)计算向量
Figure imgf000005_0001
[数]
Figure imgf000005_0002
[0035] (4)提取总贡献率达到 90%以上的前 k个主成分, 完成第一次人脸特征提取, 得 到人脸数据主成分分析后保留的样本 Y。
[0036] (5)其次是利用线性判别分析法对 Υ进行二次特征提取。 根据准则函数计算最 佳投影方向 W, 将 Y投影到 LDA的最佳投影方向 W
[0037] Z ;=W Ty ;
[0038] [0039] 进而完成了 KFDA特征提取, 形特征子空间, 并得到训练样本数据集∑。
[0040] (6)对于测试数据集 X', 将其投影至训练集的特征子空间中, 得到特征提取后 的测试数据集 Z'。
[0041] (7)通过最近邻分类器, 将样本 Z'进行分类识别。
[0042]
[0043] 相比现有技术, 本发明提供的基于核主成分分析的人脸识别方法, 能够大幅缩 短识别吋间, 通过运用核方法, 巧妙的弥补了主成分分析法和线性判别分析法 不能利用数据中非线性信息的缺憾。
[0044]
[0045] 可以理解的是, 对本领域普通技术人员来说, 可以根据本发明的技术方案及其 发明构思加以等同替换或改变, 而所有这些改变或替换都应属于本发明所附的 权利要求的保护范围。

Claims

权利要求书
一种基于核判别分析的人脸识别方法, 其特征在于: 所述识别方法包 括以下步骤:
步骤一、 进行特征提取, 将样本数据向高维核空间映射, 并利用主成 分分析法进行第一次特征提取;
步骤二、 构造中心化矩阵 H, 求解特征方程;
步骤三、 计算向量;
步骤四、 提取主成分, 完成第一次人脸特征提取, 得到人脸数据主成 分分析后保留的样本 Y;
步骤五、 利用线性判别分析法对 Y进行二次特征提取;
步骤六、 对于测试数据集 X', 将其投影至训练集的特征子空间中, 得 到特征提取后的测试数据集 Z';
步骤七、 通过最近邻分类器, 将样本 Z'进行分类识别。
如权利要求 1所述的基于核判别分析的人脸识别方法, 其特征在于: 所述步骤一利用 KPCA进行特征提取。
如权利要求 1所述的基于核判别分析的人脸识别方法, 其特征在于: 所述步骤一对给定的 M个训练集数据 Χ=[χ ,,χ 2,...,χ Μ], 计算核矩阵 Κ , 其中
[数] 俩 如权利要求 1所述的基于核判别分析的人脸识别方法, 其特征在于: 所述步骤三计算的向量为
数]
Figure imgf000007_0001
和 [数]
Figure imgf000008_0001
[权利要求 5] 如权利要求 1所述的基于核判别分析的人脸识别方法, 其特征在于: 所述步骤四提取的是总贡献率达到 90%以上的前 k个主成分。
[权利要求 6] 如权利要求 1所述的基于核判别分析的人脸识别方法, 其特征在于: 所述步骤五具体为根据准则函数计算最佳投影方向 W, 将 Y投影到 LD
A的最佳投影方向 W
Z ;=W Ty ; 进而完成了 KFDA特征提取, 形特征子空间, 并得到训练样本数据集 Z。
PCT/CN2017/080174 2017-04-12 2017-04-12 基于核判别分析的人脸识别方法 WO2018187950A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/080174 WO2018187950A1 (zh) 2017-04-12 2017-04-12 基于核判别分析的人脸识别方法

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/080174 WO2018187950A1 (zh) 2017-04-12 2017-04-12 基于核判别分析的人脸识别方法

Publications (1)

Publication Number Publication Date
WO2018187950A1 true WO2018187950A1 (zh) 2018-10-18

Family

ID=63792187

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/080174 WO2018187950A1 (zh) 2017-04-12 2017-04-12 基于核判别分析的人脸识别方法

Country Status (1)

Country Link
WO (1) WO2018187950A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020168796A1 (zh) * 2019-02-19 2020-08-27 深圳先进技术研究院 一种基于高维空间采样的数据增强方法
CN111611909A (zh) * 2020-05-18 2020-09-01 桂林电子科技大学 多子空间域自适应人脸识别方法
CN113158812A (zh) * 2021-03-25 2021-07-23 南京工程学院 基于混合扩展块字典稀疏表示的单样本人脸识别方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102142082A (zh) * 2011-04-08 2011-08-03 南京邮电大学 用于人脸识别的基于虚拟样本的核鉴别方法
CN102609693A (zh) * 2012-02-14 2012-07-25 南昌航空大学 基于模糊二维核主成分分析的人脸识别方法
CN105844291A (zh) * 2016-03-18 2016-08-10 常州大学 一种基于核典型相关分析的特征融合方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102142082A (zh) * 2011-04-08 2011-08-03 南京邮电大学 用于人脸识别的基于虚拟样本的核鉴别方法
CN102609693A (zh) * 2012-02-14 2012-07-25 南昌航空大学 基于模糊二维核主成分分析的人脸识别方法
CN105844291A (zh) * 2016-03-18 2016-08-10 常州大学 一种基于核典型相关分析的特征融合方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WANG JIAN ET AL.: "Fast Kernel subspace face recognition algorithm based on neural network", COMPUTER SCIENCE, vol. 42, no. 11A, 30 November 2015 (2015-11-30), pages 175 - 178, ISSN: 1002-137X *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020168796A1 (zh) * 2019-02-19 2020-08-27 深圳先进技术研究院 一种基于高维空间采样的数据增强方法
CN111611909A (zh) * 2020-05-18 2020-09-01 桂林电子科技大学 多子空间域自适应人脸识别方法
CN113158812A (zh) * 2021-03-25 2021-07-23 南京工程学院 基于混合扩展块字典稀疏表示的单样本人脸识别方法

Similar Documents

Publication Publication Date Title
Galbally et al. A high performance fingerprint liveness detection method based on quality related features
Sahoo et al. Multimodal biometric person authentication: A review
Ashraf et al. Multi-biometric sustainable approach for human appellative
Meraoumia et al. An automated ear identification system using Gabor filter responses
WO2018187953A1 (zh) 基于神经网络的人脸识别方法
Tiwari et al. A review of advancements in biometric systems
Kaur A study of biometric identification and verification system
WO2018187950A1 (zh) 基于核判别分析的人脸识别方法
Hossain et al. Human identity verification by using physiological and behavioural biometric traits
Deriche Trends and challenges in mono and multi biometrics
Ghosh et al. Symptoms-based biometric pattern detection and recognition
Jadhav et al. Review on Multimodal Biometric Recognition System Using Machine Learning
Arbab‐Zavar et al. On forensic use of biometrics
Monwar A Multimodal Biometric System based on Rank Level fusion.
Kalsoom et al. Iris Recognition: Existing Methods and Open Issues
Ennaama et al. Comparative and analysis study of biometric systems
WO2018187951A1 (zh) 基于核主成分分析的人脸识别方法
WO2018187952A1 (zh) 基于神经网络的核判别分析逼近方法
Lu et al. Multimodal biometric identification approach based on face and palmprint
Sundaresan et al. Monozygotic twin face recognition: An in-depth analysis and plausible improvements
Cheng et al. Multiple-sample fusion of matching scores in biometric systems
Rahim et al. Biometric authentication using face recognition algorithms for a class attendance system
Abdulla et al. Exploring Human Biometrics: A Focus on Security Concerns and Deep Neural Networks
Amrouni et al. Identity Recognition based on Palmprints: The Preliminary Results
Salah Machine learning for biometrics

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17905494

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 21.02.2020)

122 Ep: pct application non-entry in european phase

Ref document number: 17905494

Country of ref document: EP

Kind code of ref document: A1