Multimodal biometric authentication based on score level fusion using support vector machine

Fusion of multiple biometrics for human authentication performance improvement has received considerable attention. This paper presents a novel multimodal biometric authentication method integrating face and iris based on score level fusion. For score level fusion, support vector machine (SVM) based fusion rule is applied to combine two matching scores, respectively from Laplacianface based face verifier and phase information based iris verifier, to generate a single scalar score which is used to make the final decision. Experimental results show that the performance of the proposed method can bring obvious improvement comparing to the unimodal biometric identification methods and the previous fused face-iris methods.     

A data hiding approach for the self-security of iris recognition

Attacks to biometric data are the primary danger to the self-security of biometrics. To improve the iris feature template data security, a data hiding approach based on bit streams is proposed, in which an iris feature template is embedded into a face image. The proposed approach is applicable to present dominant techniques of iris recognition. With the low computation cost and the zero decoding-error-rate, this data hiding approach, embedding target biometrie data into other biometrie data for improving the security of target data in identity recognition, data storage and transmission, can deceive attackers more effectively. Furthermore, it does not degrade the iris recognition performances. Experimental results prove that the proposed approach can be used to protect iris feature templates and enhance the security of the iris recognition system itself.     

Multimodal biometric authentication based on score level fusion of finger biometrics

In this article, we propose a novel finger multimodal biometric authentication that combines finger vein, fingerprint, finger shape and finger knuckle print features of a single human finger. The proposed multimodal biometrics provides score-level fusion approach based on triangular norm with four finger biometric traits, instead of two or three ones combined in the previous approaches. The experimental evaluations and analysis are conducted on a merged multimodal biometrics database. The results show that the proposed score-level fusion approach using triangular norm obtains a larger distance between genuine and imposter score distribution as well as achieves lower error rates. Moreover, the comparison results suggest that the proposed score level fusion of finger biometrics using triangular norm outperforms the state-of-the-art approaches.     

Wood species identification using feature-level fusion scheme

In this paper, we propose a robust wood species identification scheme by using a feature-level fusion scheme. First, a novel wood feature acquirement system is devised, which can get the curve of 1D wood spectral reflectance ratio and the 2D wood surface color image. Second, the 4 wood color features, the 4 principal texture features, the 4 secondary texture features and the 4 spectral features are established, respectively. Third, a fuzzy BP neural network is proposed to perform the classification work, which consists of 4 sub-networks based on the color feature, texture feature and spectral feature. We have experimentally proved that this scheme improves the mean recognition accuracy to approximately 90% for 5 wood species. Moreover, our feature-level fusion scheme is superior to the recognition schemes which use color feature and texture feature.     

一种新的自组织聚类网络及其在数据融合目标分类中的应用

介绍了一种新的自组织聚类人工神经网络(DIGNET)模型和网络的非监督学习算法。针对数据融合和目标识别的特点,提出了基于DIGNET的决策层数据融合目标分类方法。利用仿真数据研究了DIGNET和自组织特征映射网络(SOFM)的聚类性能以及基于DIGNET的决策层数据融合结构,实验结果表明DIGNET较SOFM正确分类率高、抗噪能力好;基于DIGNET的决策层数据融合能够有效地实现融合识别。将该数据融合方法应用于前视红外(FLIR)和可见光摄像机目标跟踪系统,结果表明该方法是可行的。     

Multi-Stream Convolution-Recurrent Neural Networks Based on Attention Mechanism Fusion for Speech Emotion Recognition

The quality of feature extraction plays a significant role in the performance of speech emotion recognition. In order to extract discriminative, affect-salient features from speech signals and then improve the performance of speech emotion recognition, in this paper, a multi-stream convolution-recurrent neural network based on attention mechanism (MSCRNN-A) is proposed. Firstly, a multi-stream sub-branches full convolution network (MSFCN) based on AlexNet is presented to limit the loss of emotional information. In MSFCN, sub-branches are added behind each pooling layer to retain the features of different resolutions, different features from which are fused by adding. Secondly, the MSFCN and Bi-LSTM network are combined to form a hybrid network to extract speech emotion features for the purpose of supplying the temporal structure information of emotional features. Finally, a feature fusion model based on a multi-head attention mechanism is developed to achieve the best fusion features. The proposed method uses an attention mechanism to calculate the contribution degree of different network features, and thereafter realizes the adaptive fusion of different network features by weighting different network features. Aiming to restrain the gradient divergence of the network, different network features and fusion features are connected through shortcut connection to obtain fusion features for recognition. The experimental results on three conventional SER corpora, CASIA, EMODB, and SAVEE, show that our proposed method significantly improves the network recognition performance, with a recognition rate superior to most of the existing state-of-the-art methods.     

尺度不变特征与几何特征融合的人耳识别方法

要提高人耳的识别率,关键是特征的提取与表达.尺度不变特征变换(SIFT)技术是局部点特征提取算法,在尺度空间寻找极值点,提取对图像的尺度和旋转变化具有不变性,对光照变化和图像变形具有较强的适应性的特征向量.尝试用SIFT技术来提取外耳图像的结构特征点以形成稳定的特征描述子,为了克服一幅图像中有多个局部描述子相似的问题,在SIFT特征描述子中融入一个耳廓几何特征.最后采用特征向量的欧氏距离作为两幅图像相似性度量标准进行人耳识别.在耳图像库七进行实验.结果表明,该方法不仅可以有效地提取人耳特征,通过少量特征可获得较高的识别率,而且对耳图像刚体变化具有较强的稳健性.     

基于I-BGLAM纹理和光谱融合的高光谱显微成像木材树种分类

为了提高木材树种分类的正确率,提出了一种基于I-BGLAM纹理特征和光谱特征融合的高光谱图像的木材树种分类方法。实验数据是利用SOC710VP高光谱成像仪获取的可见光/近红外（372.53~1 038.57 nm）范围内的高光谱图像。首先,利用基于OIF的特征波段选择方法降低高光谱图像的维数,选择出含有信息量大的波段。其次,对选择出的波段图像使用NSCT及NSCT逆变换得到融合图像,对得到的融合图像使用I-BGLAM提取其纹理特征。与此同时,对高光谱图像的全波段求取平均光谱并进行S-G（Savitzky-Golay）平滑得到光谱特征。最后,将得到的纹理特征和光谱特征融合后送进极限学习机（ELM）中进行分类。此外,还和基于灰度共生矩阵（GLCM）的木材识别的传统方法以及近几年木材树种识别领域内被提出的主流方法进行了比较。该研究主要创新点有两个：一是将强纹理提取器I-BGLAM用于高光谱图像中提取其纹理特征;二是提出一种新的特征融合的模型用于高光谱图像的分类。针对8个树种的实验结果表明,单独使用I-BGLAM提取的纹理特征来进行分类的正确率最高可到达88.54%,而使用GLCM提取纹理特征的传统方法正确率最高只有76.04%,该结果可以得出本文使用I-BGLAM在纹理特征提取方面要优于GLCM,这为后面建立的融合模型打下很好的基础,单独使用平均光谱特征来分类的正确率最高可以达到92.71%,使用所提出的特征融合方法所得到的分类正确率最高可达到100%,这说明使用所提出的融合模型来分类要比以前单独使用某一种特征的分类模型要好。此外,使用所提出的方法得到的分类正确率要高于本领域内其他两种主流的识别方法。因此,所提出的基于I-BGLAM纹理特征和光谱特征融合的方法能够提高木材树种分类的正确率,该方法在木材树种分类方面有着一定的利用价值。     

Improving iris recognition performance via multi-instance fusion at the score level

Fusion of multiple instances within a modality for biometric verification performance improvement has received considerable attention. In this letter, we present an iris recognition method based on multiinstance fusion, which combines the left and right irises of an individual at the matching score level. When fusing, a novel fusion strategy using minimax probability machine （MPM） is applied to generate a fused score for the final decision. The experimental results on CASIA and UBIRIS databases show that the proposed method can bring obvious performance improvement compared with the single-instance method. The comparison among different fusion strategies demonstrates the superiority of the fusion strategy based on MPM.     

语音情感识别中的特征选择方法

为了解决传统卷积神经网络在识别中文语音时预测错误率较高、泛化性能弱的问题,首先以深度卷积神经网络(DCNN)-连接时序分类(CTC)为研究对象,深入分析了不同卷积层、池化层以及全连接层的组合对其性能的影响;其次,在上述模型的基础上,提出了多路卷积神经网络(MCNN)-连接时序分类(CTC),并联合SENet提出了深度SE-MCNN-CTC声学模型,该模型融合了MCNN与SENet的优势,既能加强卷积神经网络的深层信息的传递、避免梯度问题,又可以对提取的特征图进行自适应重标定。最终实验结果表明:SE-MCNN-CTC相较于DCNN-CTC错误率相对降低13.51%,模型最终的错误率达22.21%;算法改进后的声学模型可以有效地提升泛化性能。     

Object Selection as a Biometric

The use of eye movement as a biometric is a new biometric technology that is now in competition with many other technologies such as the fingerprint, face recognition, ear recognition and many others. Problems encountered with these authentication methods such as passwords and tokens have led to the emergence of biometric authentication techniques. Biometric authentication involves the use of physical or behavioral characteristics to identify people. In biometric authentication, feature extraction is a very vital stage, although some of the extracted features that are not very useful may lead to the degradation of the biometric system performance. Object selection using eye movement as a technique for biometric authentication was proposed for this study. To achieve this, an experiment for collecting eye movement data for biometric purposes was conducted. Eye movement data were measured from twenty participants during choosing and finding of still objects. The eye-tracking equipment used was able to measure eye-movement data. The model proposed in this paper aimed to create a template from these observations that tried to assign a unique binary signature for each enrolled user. Error correction is used in authenticating a user who submits an eye movement sample for enrollment. The XORed Biometric template is further secured by multiplication with an identity matrix of size (n × n). These results show positive feedback on this model as individuals can be uniquely identified by their eye movement features. The use of hamming distance as additional verification helper increased model performance significantly. The proposed scheme has a 37% FRR and a 27% FAR based on the 400 trials, which are very promising results for future improvements.     

水下无源声呐目标听觉域张量特征提取方法

特征提取是水下无源声呐目标分类识别的关键步骤,提出了一种基于听觉Patterson-Holdsworth耳蜗模型的听觉域张量特征提取方法。将耳蜗模型的滤波器冲激响应视为信号分解的基函数,根据听觉模型非线性尺度或常规线性尺度确定不同通道的中心频率,然后计算出相应通道的增益和带宽,并量化冲激响应的阶数和相位参数,得到信号分解基,再根据信号分解原理得到通道数×阶数×相位数的三阶张量特征,并通过计算测试样本张量特征与训练样本张量特征间的相似性实现了水下无源声呐目标的分类识别。海上实录无源声呐目标的分类识别实验表明,提取的张量特征具有比较好的分类识别性能,听觉模型等效矩形带宽尺度优于线性尺度划分中心频率,能够提高无源声呐的目标指示能力。     

基于BLCD分解与ACO-DBN网络的光纤振动信号识别

针对光纤振动信号受噪声干扰严重、特征提取单一和识别时间长的问题,提出了改进的局部特征尺度分解和蚁群算法优化深度置信网络的识别方法.首先,采用三次B样条函数插值拟合均值曲线改进局部特征尺度分解算法,并对原始信号进行分解得到一系列内禀尺度分量之和.其次,利用峭度因子和能谱系数构成融合指标筛选有效分量.然后,分别提取有效分量...     

Optics based biohashing using joint transform correlator

Most authentication process uses password and personal identification numbers (PIN) for security purposes. In order to remove the problem of hacking or stealing of the password and PIN numbers, there has been an increased interest in the utilization of specific biometric feature of the user. Recently, biohashing systems have been introduced for automatic biometric recognition. In a biohashing system, biohash codes are generated using the feature of the biometric. A basic biohashing system involves two steps. First is the extraction of the feature from the input biometric image and second is the discretisation of the obtained feature vector by using ortho-normalized random numbers. In this paper, a new biohashing system has been proposed in which joint transform correlator (JTC) has been used for extraction of the specific feature of the biometric. In the enrolment process, a biohash code has been generated by using a single face image and then stored. In the verification process, this biohash code is matched with the verification codes for recognition purpose. The main advantage of the proposed biohashing method is the possibility of the optical implementation of the feature extraction of the face image. Experimental as well as simulation results have been given to validate the proposed technique. Normalized Hamming distance has been calculated to discriminate the genuine and impostor face images. By varying the dimension of the feature matrix, the study of the variation of the normalized Hamming distance with the density of the population has been undertaken. For the performance evaluation of the proposed technique the false rejection ratio (FRR) and false acceptance ratio (FAR) have also been calculated.     

采用特征融合的紫檀属内相似树种识别方法研究

紫檀属中的木材有很多属于名贵木材,不同树种之间十分相似。传统的木材识别方法多以木材解剖学为主,通过观察木材的切片结构特征对木材的树种进行判断,这类方法虽有较高的识别精度,但是其识别工艺较为复杂而且技术难度也相对较高。与木材解剖学相对应的是利用图像信息或光谱信息的木材树种识别方法,该类方法虽具有较为简单的识别工艺,但是在对同属相似木材树种进行识别时,往往不能够取得较好的识别效果。提出了一种基于木材切面光谱特征和纹理特征相融合的木材树种识别方法,该方法不仅识别工艺简单、自动化程度高,而且具有较高的识别精度。首先通过数码相机和光谱仪采集木材切面的图像信息和光谱信息,然后分别使用纹理特征提取方法和光谱特征提取方法提取两类特征的特征向量,接下来使用基于典型相关分析的特征级融合方法将这两个特征向量进行融合,最后使用支持向量机对融合后的特征向量进行分类识别。为了验证方法的有效性,以市场中常见的5种紫檀属树种的三个切面为研究对象,对这些木材树种进行了识别。实验结果显示,单独使用纹理特征的识别正确率最高为80.00%,单独使用光谱特征的识别正确率最高为94.40%,使用融合的特征最高的识别正确率可达99...     

改进的对向传播网络及其在多传感器目标识别中的应用

针对多传感器数据融合和目标识别的特点,提出了改进的对向传播网络(MCPN),并与Dempster-Shafer(D-S)证据推理相结合,实现了决策层数据融合目标识别.文中利用仿真数据对所提出的网络训练算法和融合结构进行了实验研究.结果表明:改进后的对向传播网络识别性能优于传统的对向传播网络(CPN),融合后的目标识别率较单传感器明显提高.最后,将该方法应用于前视红外(FLIR)和可见光摄像机目标跟踪系统对算法和融合结构进行验证,结果表明文中提出的方法是可行的.     

Tensor feature extraction of underwater passive sonar target based on auditory model

Feature extraction is a key step for underwater passive sonar target classification and recognition.A kind of tensor feature extraction method based on auditory PattersonHoldsworth cochlear model is proposed.First,the filter impulse response of the cochlear model is regarded as the basis function of signal decomposition,and the center frequency of different channels is determined according to the nonlinear scale or conventional linear scale of the auditory model.Then,the gain and bandwidth of th...     

Biometric identification using local iterated function

Biometric identification protocol has been received an increasing interest recently. It is a process that determines person identity by making use of their biometric features. A new biometric identification method is presented in this paper based on partial self-similarity that used to identify features within fingerprint images. This approach is already used in Fractal Image Compression (FIC) due to their ability to represent the images by a limited number of affine transformations, and its variation of scale, translation or rotation. These features give the recognition process high impact and good performance. To process data in a fingerprint image, it first converted into digital format using Optical Fingerprint Reader (OFR). The verification process is done by comparing these data with the server data. The system analysis shows that the proposed method is efficient in terms of memory and time complexity.     

Fusion of deep shallow features and models for speaker recognition

In order to further improve the performance of speaker recognition, features fusion and models fusion are proposed. The features fusion method is to fuse deep and shallow features. The fused feature describes speaker characteristics more comprehensively than a single feature because of the complementarity between different levels of features. The models fusion method is to fuse i-vectors extracted from different speaker recognition systems. The fused model can combine advantages of different speaker recognition systems. Experimental results show the effectiveness of the proposed methods. Compared with the state-of-the-art system on CASIA North and South dialect corpus,the proposed features fusion system and models fusion system achieved about 54.8% and 69.5% relative improvement on the equal error rate(EER),respectively.     

基于多光谱图像融合的掌纹识别方法

生物特征识别在信息安全领域发挥着重要作用,掌纹识别作为一种新型生物特征识别方式,具有低失真、非侵入性和高唯一性等优势。传统掌纹研究大多使用自然光成像系统以灰度格式获取,识别精度很难进一步提升。为了获得更多的身份鉴别信息,提出利用多光谱掌纹图像代替自然光掌纹图像。针对现有掌纹识别算法由于没有考虑到不同光谱的特性而导致纹理细节丢失,识别精准率低的问题,提出了一种基于多光谱图像融合的掌纹识别算法。该方法通过对不同光谱下的掌纹图像进行快速自适应二维经验模式分解（FABEMD）,将多光谱掌纹图像分解成一系列频率由高到低的二维固有模态函数(BIMF)和一个残余分量,残余分量可被视为该光谱图像低频信息的初步估计。图像采集过程中光照条件很难保持稳定,而近红外光谱图像在进行FABEMD分解时对光照变换敏感,容易导致分解后的BIMF背景信息过于冗余;因此对分解后的近红外掌纹图像进行背景重建及特征细化,在对背景冗余信息进行平滑处理的同时可以有效增强高频信息的特征表达。为避免直接融合处理后引发的图像过度曝光问题,提出对近红外特征压缩后再融合。此外,提出了一种结合了注意力机制的改进残差网络（IRCANet）,用于融合后的掌纹图像分类,在网络中引入分阶段残差结构,缓解了网络的退化问题,在学习过程中有效地减少信息丢失,对于融合后的多光谱掌纹图像,分阶段残差结构能够稳定地将图像信息在网络间传输,但对图像中的高低频信息区分效果不够显著,为了使网络关注更多区分性特征,利用特征通道间的相互依赖性,在分阶段残差结构中结合了通道注意力(Channel Attention)机制。最终,在香港理工大学（PolyU）多光谱掌纹数据集上进行的综合实验表明,该方法可以取得良好的效果,算法识别准确率能达到99.67%且具有良好的实时性。