傳統(tǒng)的身份驗(yàn)證策略（如密碼和智能卡）因?yàn)樗鼈兛梢员还蚕�、遺忘、復(fù)制、操縱或偽造,其安全性存在隱患。與傳統(tǒng)方法不同,生物識(shí)別是基于人的生理或行為特征進(jìn)行身份識(shí)別的科學(xué),已成為確定個(gè)體身份的合法方法。如今,生物識(shí)別技術(shù)已不再局限于刑事執(zhí)法,更多企業(yè)使用生物識(shí)別技術(shù)來管理對(duì)建筑物和信息的訪問。然而,大多數(shù)單模態(tài)生物特征識(shí)別受到諸如噪聲數(shù)據(jù),非普遍性和欺騙攻擊之類的限制,使得它無法達(dá)到現(xiàn)實(shí)世界應(yīng)用的性能要求。為了克服單模態(tài)方法的這些缺點(diǎn),本文提出了一種新穎的安全多模態(tài)生物識(shí)別方法,使用不同的融合方法將心電圖（ECG）與指紋相結(jié)合。該方法克服了單一方法的局限性,提高了整體方法的性能并增強(qiáng)了安全性,對(duì)欺騙攻擊具有更好的魯棒性。與其他多模態(tài)生物識(shí)別方法（例如,面部,耳朵和基于指紋的多模態(tài)生物識(shí)別系統(tǒng)）相比,ECG信號(hào)可以容易地從手指獲取,這使得系統(tǒng)非常方便和有效。此外,ECG信號(hào)只能從活人身體獲取,因此還可以據(jù)此進(jìn)行活體檢測(cè),使系統(tǒng)具有更強(qiáng)的抗攻擊能力。本文的第一部分研究了心電圖和指紋作為單模態(tài)生物識(shí)別的方法。為此,我們首先提出一種改進(jìn)的生物哈希和矩陣運(yùn)算方法,為生物識(shí)別系統(tǒng)生成了一種新的可取消的心... 

【文章來源】：哈爾濱工業(yè)大學(xué)黑龍江省 211工程院校 985工程院校

【文章頁(yè)數(shù)】：143 頁(yè)

【學(xué)位級(jí)別】：博士

【文章目錄】：
摘要
Abstract
Abbreviations
Chapter1 Introduction
    1.1 Motivation for Multi-biometrics
        1.1.1 Multimodal biometric systems
        1.1.2 Fusion in Multimodal Biometrics
    1.2 Motivation for Applying ECG and Fingerprint Multi-Biometrics
    1.3 Research Goals and Contributions
    1.4 Outline of Thesis
Chapter2 Related Work and Existing Databases
    2.1 Recent ECG Biometric Methods
        2.1.1 Conventional Machine Learning Approaches
        2.1.2 Deep Learning-Based Approaches
    2.2 Recent Fingerprint Biometric Methods
        2.2.1 Conventional Machine Learning Approaches
        2.2.2 Deep Learning-Based Approaches
    2.3 Recent Multimodal Biometric Methods based on ECG and Fingerprints.
    2.4 Existing ECG Databases
    2.5 Existing Fingerprint Databases
    2.6 The Multimodal ECG/Fingerprint Databases
    2.7 Summary
Chapter3 ECG and Fingerprint Biometric Authentication Method
    3.1 Introduction
    3.2 Proposed Cancelable Biometric Method based on ECG
        3.2.1 ECG Feature Extraction Algorithm
        3.2.2 ECG Feature Template Protection Methods
            3.2.2.1 Cancelable ECG using improved Bio-Hashing method
            3.2.2.2 Build cancelable ECG using an input feature matrix
        3.2.3 Feed Forward Neural Network(FFNN)
    3.3 Fingerprint Classification Algorithm
        3.3.1 Fingerprint Preprocessing and Feature Extraction
        3.3.2 QG-MSVM Classifier
    3.4 Experimental Results
        3.4.1 Performance of the Proposed Cancelable Method
            3.4.1.1 Evaluation of improved Bio-Hash algorithm requirement
            3.4.1.2 Evaluation of Matrix operation algorithm requirement
            3.4.1.3 Results of the proposed cancelable method
        3.4.2 Performance of Fingerprint Classification Algorithm
            3.4.2.1 Results of the proposed classifier
            3.4.2.2 Discussion
    3.5 Summary
Chapter4 Multimodal Biometric Authentication Methods Using Convolution Neural Network based on Different Level Fusion of ECG and Fingerprint
    4.1 Introduction
    4.2 The Proposed Hierarchical Multimodal Method using CNN based on Decision Level Fusion
        4.2.1 First Stage of Hierarchical Multimodal Method(ECG Authentication)
        4.2.2 Second Stage of Hierarchical Multimodal Method(Fingerprint Authentication)
        4.2.3 Extracting the Features of ECG and Fingerprint(The Proposed CNN model)
        4.2.4 Updating of ECG and Fingerprint Templates(Cancelable Method)
        4.2.5 Classification of ECG and Fingerprint
        4.2.6 Internal Fusion of ECG and Fingerprint
        4.2.7 Data Augmentation
        4.2.8 Decision Fusion of ECG and Fingerprint
    4.3 The Proposed Parallel Multimodal Method using CNN based on Feature Level Fusion
    4.4 Evaluation and Results
        4.4.1 Multimodal Datasets
        4.4.2 Experimental Setup
        4.4.3 Fusion of ECG and Fingerprint(the first method)
        4.4.4 Fusion of ECG and Fingerprint(the second method)
    4.5 Discussion
        4.5.1 Effect of Dataset Augmentation
        4.5.2 Effect of Template Protection Method
        4.5.3 Computational Costs
    4.6 Summary
Chapter5 Parallel Score Fusion of ECG and Fingerprint for Human Authentication based on Convolution Neural Network
    5.1 Introduction
    5.2 The Proposed Multimodal Method using CNN based on Parallel Score Fusion
        5.2.1 First Step of Multimodal Method(ECG Authentication)
        5.2.2 Second Step of Multimodal Method(Fingerprint Authentication)
        5.2.3 ECG and Fingerprint Templates protection(Matrix Operation method) ..
        5.2.4 Classification of ECG and Fingerprint
        5.2.5 Parallel Score Fusion
        5.2.6 Data Augmentation
    5.3 Experimental Setup and Results
        5.3.1 Multimodal Dataset
        5.3.2 Experimental Setup
        5.3.3 Fusion of ECG and fingerprint
        5.3.4 Discussion
    5.4 Summary
Conclusions
References
List of Publications
Acknowledgements
Resume


【參考文獻(xiàn)】：
期刊論文
[1]Fingerprint Liveness Detection Based on Multi-Scale LPQ and PCA[J]. Chengsheng Yuan,Xingming Sun,Rui Lv.  中國(guó)通信. 2016(07)



本文編號(hào)：3606854