基于信号质量动态加权的多模生物特征识别研究

doi:10.11959/j.issn.2096-109x.2018022

摘要/Abstract

摘要：

将人脸识别和说话人识别进行决策层级的融合，为应对外界环境对识别结果的影响，引入图像质量和声音质量评价方法，通过对信息质量进行评估，去除信息质量较差的特征，根据信息质量动态调整模块的权重比例，并对单模特征识别匹配度低的个体做拒绝处理，然后根据D-S理论将各个证据合并成为一个新的证据体，实现对用户身份识别。实验结果显示，这种考虑特征信号质量的融合方法可以有效提升识别的准确率和安全性。

关键词: 人脸识别, 说话人识别, 特征信号质量评估, D-S证据理论, 多生物特征识别融合

Abstract:

Face recognition and speaker recognition were integrated at the decision-making level.In order to cope with the influence of the external environment on the recognition result,image quality and sound quality assessment methods were introduced.By evaluating the quality of information,features with poor information quality were removed.Information quality dynamically adjusted the weight proportion of the module,and rejected individuals with low matching degree of single-mode feature recognition.Then,according to DS theory,each evidence was merged into a new body of evidence to realize user identification.The experimental results show that the fusion method which takes into account the characteristic signal quality can effectively improve the recognition accuracy and security.

Key words: face recognition, speaker recognition, characterization of signal quality, D-S evidence theory

中图分类号:

TP391.41

张闻彬,刘培顺,薛峰会. 基于信号质量动态加权的多模生物特征识别研究[J]. 网络与信息安全学报, 2018, 4(3): 59-67.

Wenbing ZHANG,Peishun LIU,Fenghui XUE. Multi-biometrics fusion based on dynamic weighting of characteristic signal quality[J]. Chinese Journal of Network and Information Security, 2018, 4(3): 59-67.

图/表 13

图1

图2

图3

表1

图4

图5

图6

表2

图7

图8

图9

表3

表4

参考文献 38

[1]	JAIN A , BOLLE R , PANKANTI S . Biometrics Personal Identification in Networked Society[M]. Boston: Kluwer Academic PublishersPress, 1999.
[2]	绕过三星 Galaxy S8 的虹膜识别？黑客表示非常简单[EB/OL]. .
	To bypass the iris recognition of samsung Galaxy S8? Hackers say it's very simple[EB/OL]. .
[3]	疯狂的身份证:一张身份证如何攻破人脸识别技术[EB/OL]. .
	Crazy ID card:how does a card breach face recognition technolo-gy[EB/OL]. .
[4]	ROSS A , JAIN A . Information fusion in biometrics[J]. Pattern Recognition Letters, 2003,24(13): 2115-2125.
[5]	ROSS A , JAIN A . Multimodal biometrics:an overview[C]// European Signal Processing Conference. 2015: 1221-1224.
[6]	MAURER D E , BAKER J P . Fusing multimodal biometrics with quality estimates via a Bayesian belief network[J]. Pattern Recognition, 2008,41(3): 821-832.
[7]	NANDAKUMAR K , CHEN Y , JAIN A ,et al. Quality-based score level fusion in multibiometric systems[C]// The International Conference on Pattern Recognition. 2006: 473-476.
[8]	KHIARI-HILI N , MONTAGNE C , LELANDAIS S ,et al. Quality dependent multimodal fusion of face and iris biometrics[C]// The International Conference on Image Processing Theory TOOLS and Applications. 2017: 1-6.
[9]	杨静波 . 融合说话人识别和人脸识别的身份认证[D]. 兰州:西北师范大学, 2016.
	YANG J B . Research on authentication by fusing with speaker recognition and face recognition[D]. Lanzhou:Northwest Normal University, 2016.
[10]	陈倩 . 多生物特征融合身份识别研究[D]. 杭州:浙江大学, 2007.
	CHEN Q . Multimodel biometric personal identification based on data fusion[D]. Hangzhou:Zhejiang University, 2007.
[11]	LEE S Y , HAM Y K , PARK R H . Recognition of human front faces using knowledge-based feature extraction and neuro fuzzy algorithm[J]. Pattern Recognition, 1996,29(11): 1,863-1,876
[12]	KIRBY M , SIROVICH L . Application of the karhunen-loeve procedure for the characterization of human faces[J]. IEEE Transactions on Pattern Analysis ＆ Machine Intelligence, 2002,12(1): 103-108
[13]	TAIGMAN Y , YANG M , RANZATO M ,et al. DeepFace:closing the gap to human-level performance in face verification[C]// IEEE Computer Vision and Pattern Recognition. 2014: 1701-1708.
[14]	SUN Y , WANG X , TANG X . Deep learning face representation from predicting 10,000 classes[J]. Computer Vision ＆ Pattern Recognition, 2014: 1891-1898.
[15]	SUN Y , WANG X , TANG X . Deep learning face REPRESENTA TION by joint identification-verification[J]. Advances in Neural Information Processing Systems, 2014,27: 1988-1996.
[16]	SUN Y , WANG X , TANG X . Deeply learned face representations are sparse,selective,and robust[J]. Computer Vision ＆ Pattern Recognition, 2015: 2892-2900.
[17]	SUN Y , LIANG D , WANG X ,et al. DeepID3:face recognition with very deep neural networks[J]. Computer Science, 2015:
[18]	SCHROFF F , KALENICHENKO D , PHILBIN J . FaceNet:a unified embedding for face recognition and clustering[J]. Computer Vision ＆ Pattern Recognition, 2015: 815-823.
[19]	DOUGLAS A , REYNOLDS , QUATIERI T F ,et al. Speaker verification using adapted Gaussian mixture models[J]. Digital Signal Processing, 2000,10(1-3): 19-41.
[20]	KENNY P , BOULIANNE G , OUELLET P ,et al. Speaker and session variability in GMM-based speaker verification[J]. IEEE Transactions on Audio Speech ＆ Language Processing, 2007,15(4): 1448-1460.
[21]	DEHAK N , DEHAK R , KENNY P ,et al. Support vector machines versus fast scoring in the low-dimensional total variability space for speaker verification[C]// Conference of the International Speech Communication Association. 2009: 1559-1562.
[22]	KRYSZCZUK K , DRYGAJLO A . What do quality measures predict in biometrics?[C]// European Signal Processing Conference. 2015: 1-5.
[23]	HSU R L V , SHAH J , MARTIN B . Quality assessment of facial Images[C]// Biometric Consortium Conference. 2006: 1-6.
[24]	CAPPELLI R , FERRARA M , MALTONI D . The quality of fingerprint scanners and its impact on the accuracy of fingerprint recognition algorithms[C]// The International Conference on Multimedia Content Representation. 2006: 10-16.
[25]	ABAZA A , HARRISON M A , BOURLAI T ,et al. Design and evaluation of photometric image quality measures for effective face recognition[J]. Biometrics Iet, 2014,3(4): 314-324.
[26]	BARTOS A L , NELSON D J . Enabling improved speaker recognition by voice quality estimation[C]// Signals,Systems ＆ Computers. 2012: 595-599.
[27]	SELLAHEWA H , JASSIM S A . Image-quality-based adaptive face recognition[J]. IEEE Transactions on Instrumentation ＆ Measurement, 2010,59(4): 805-813.
[28]	高修峰, 张培仁, 李子青 . 人脸图像质量评估标准[J]. 小型微型计算机系统, 2009,30(1): 95-99
	GAO X F , ZHANG P R , LI Z Q . Standardization of face image sample quality[J]. Journal of Chinese Computer Systems, 2009,30(1): 95-99.
[29]	ABBOUD A J , SELLAHEWA H , JASSIM S A . Quality based approach for adaptive face recognition[C]// The International Society for Optics and Photonics. 2009.
[30]	WANG Z , BOVIK A C . A universal image quality index[J]. IEEE Signal Processing Letters, 2002,9(3): 81-84
[31]	FERZLI R , KARAM L J . A no-reference objective image sharpness metric based on the notion of just noticeable blur (JNB)[J]. IEEE Transactions on Image Processing, 2009,18(4): 717-28.
[32]	NARVEKAR N D , KARAM L J . A no-reference image blur metric based on the cumulative probability of blur detection (CPBD)[J]. IEEE Transactions on Image Processing, 2011,20(9): 2678-2683.
[33]	肖宾杰 . 基于图像质量加权的 D-S 证据理论多生物特征融合识别[J]. 计算机应用, 2012,32(1): 264-268.
	XIAO B J . Muti-biometric feature fusion identification based on D-S evident theory of image quality with different weights[J]. Journal of Computer Applications, 2012,32(1): 264-268
[34]	蒋晔 . 基于短语音和信道变化的说话人识别研究[D]. 南京:南京理工大学, 2013.
	JIANG Y . Research on speaker recognition over short utterance and varying channels[D]. Nanjing:Nanjing University of Science and Technology, 2013.
[35]	BLATNIK R , KANDUS G . Influence of the speech quality in telephony on the automated speaker recognition[C]// WSEAS International Conference on Circuits. 2011: 115-120
[36]	RIX A W , BEERENDS J G , HOLLIER M P ,et al. Perceptual evaluation of speech quality (pesq)[J]. Journal of the Audio Engineering Society, 2001,50(10): 755-764.
[37]	JAIN A , NANDAKUMAR , ROSS K A . Score normalization in multimodal biometric systems[J]. Pattern Recognition, 2004,38(12): 2270-2285.
[38]	王延平, 袁杰, 廖原 ,等. 利用信息融合技术的缺损目标识别方法[J]. 中国图象图形学报, 2000,5(3): 237-240.
	WANG Y P , YUAN J , LIAO Y ,et al. Pattern recognition of oc-cluded objects using information fusion[J]. Journal of Image and Graphics, 2000,5(3): 237-240.

图像序号	质量指标
图3(a)	0.79
图3(b)	0.58
图3(c)	0.30
图3(d)	0.18

人脸\声音	归一化后匹配分数α_i	特征信息质量Q_i	结果可信系数R_i	赋值概率函数	目标身份概率
待测身份1	0.21\0.80	0.35\0.96	0.08\0.92	0.11\0.92	0.57
待测身份2	0.06\0.69	0.53\0.78	0.05\0.94	0.03\0.91	0.25
待测身份3	0.85\0.10	0.67\0.48	0.96\0.04	0.95\0.03	0.38
待测身份4	0.08\0.17	—	—	—	—

AUC值	本文方法	不考虑特征质量融合	说话人识别	人脸识别
Q_F∈[0.5,0.7]、Q_S<0.4	0.947 4	0.855 7	0.762 3	0.914 3
Q_S∈[0.5,0.7]、Q_F<0.4	0.932 7	0.819 2	0.876 1	0.737 3
Q_S<0.3、Q_F<0.3	0.874 4	0.723 3	0.627 8	0.731 3

识别方式	时间/s
本文方法	16.81
说话人识别	16.58
人脸识别	15.37