International Journal of Biometrics (9 papers in press)

Regular Issues

• Face image database: A test-bed for evaluation and certification of facial recognition systems  
  by Frank Wu, Qinghan Xiao, Tien Vo 
  Abstract: Recently there has been a growing interest in facial recognition technology because of its great potential usage in tracking and finding persons of interest in public spaces. However, there is no government agency or organization in Canada with the capability to test or certify facial recognition algorithms, products, or systems. To fill this gap, we carried out a Public Security Technical Program (PSTP) project to create a database that consists of face images commonly used by biometric communities. The objective is to establish a test-bed for government agencies to evaluate the performance of facial recognition algorithms and products. Different test scenarios were studied, and the experimental results illustrated the utility, validity and flexibility of the database.
  Keywords: facial recognition; face image database; benchmark test-bed; performance evaluation.
   
• A Novel Approach for Recognizing People of the Same Family  
  by Mohammad Ghahramani, Wei-Yun Yau, Eam-Khwang Teoh 
  Abstract: Family members have certain facial resemblances due to genetic similarities. Such resemblance allows us to guess the family relationship which has received little attention to date. In this paper, we propose to recognize immediate members of the same family known as family verification through identification of members facial resemblance features. At first, the proposed operator extracts detailed facial information comparing to the state-of-the-art face descriptors. A novel redundant feature set removal is incorporated to reduce feature dimension. Experiments are conducted to compare performance of various features employing the proposed method and the state-of-the-art face recognition as well as the proposed redundant feature removal. The experiments also cover various scenarios where different members of the family are absent from training but present in the testing applied for missing family member verification. Results obtained show that the proposed approach is feasible even in case of missing member verification.
  Keywords: Family verification, Feature reduction, kinship, LBP
   
• Random Vector Quantization Modelling in Automatic Speaker Verification  
  by Hayet Djellali, Mohamed Tayeb LASKRI 
  Abstract: Automatic speaker verification (ASV) is a binary classification task, it consists of accepting or rejecting the claimed identity. ASV system has to decide whether a claimed speaker uttered a sentence. This paper proposes an algorithm called Impostor Random vector quantization (IRVQ) based on multiples random codebook. IRVQ algorithm represents the impostor model also called universal background model (UBM) and compare it to the second algorithm called Partial impostor VQ (IVQ) for vector quantization (VQ) in speaker verification. The present study demonstrates that the several random selected codebooks representing impostor models give better results and less half total error HTER than impostor IVQ method and baseline system. The performance of these models is evaluated on Arabic speaker verification dataset. However, this improvement also depends on the codebook size. The assumption concerning partitioning impostor acoustic spaces and choose the best subspace to represent impostor model specific to each speaker is an efficient approach.
  Keywords: Automatic Speaker Verification, Vector Quantization (VQ), codebook, Impostor Random VQ model, Universal Background Model.
   
• A Multi-modal Dataset, Protocol and Tools for Adaptive Biometric Systems: A Benchmarking Study  
  by Ajita Rattani, Gian Luca Marcialis, Fabio Roli 
  Abstract: Adaptive biometric systems have received a recent spurt in biometric community. These systems have the additional capability to adapt themselves using biometric data available during the system's operation. Although several studies have been proposed in this field, no conclusive evidences can be drawn about the expected performance gain on making the biometric system adaptive. This is due to the adoption of different and inappropriate databases, protocols and tools for evaluating adaptive biometric systems. This paper presents a benchmarking study to facilitate fair comparison and independent replication of the results from different research groups. To this aim, this paper describes DIEE multi-modal database consisting of face and fingerprint biometrics and a protocol tailored for adapting as well as evaluating adaptive biometric systems. In addition, several tools for evaluating and visualizing the performance gain on making the biometric system adaptive are provided as well. To the best of our knowledge, this is the first attempt to benchmark database, protocol and tools for evaluating adaptive biometric systems operating in verification mode.
  Keywords: Adaptive biometrics, Biometric template update, Self- update, Co-update, Benchmarking, Dataset, Protocol, Tools
   
• Using Brain Waves as Transparent Biometrics for On-Demand Driver Authentication  
  by Isao Nakanishi, Sadanao Baba, Koutaro Ozaki, Shigang Li 
  Abstract: Conventional biometric systems mainly assume one-time-only authentication. However, this technique is not used with user management applications. If a user is replaced by an imposter after the authentication has occurred, the systems cannot detect such a replacement. One solution to this problem is on-demand authentication, in which users are authenticated on a regular or nonregular schedule, as determined by the system. However, the on-demand-authentication technique requires that we present biometric data without regard to do so. In this paper, we focus on the use of brain waves as transparent biometric signals. In particular, we assume driver authentication and measure the brain waves of drivers when they are performing mental tasks such as tracing routes or using a simplified driving simulator as an actual task. We propose to extract the power spectrum in the alpha-beta band as an individual feature and propose two verification methods based on the similarity of the features. In addition, we propose to divide the alpha-beta band into either four or six partitions and to fuse the similarity scores from all the partitions. We evaluate the verification performance using 23 subjects and obtain an equal error rate of 20-25 %.
  Keywords: brain wave; EEG; on-demand authentication; transparent biometrics; driver authentication; route trace; simplified driving simulator.
   
• A method for static CCTV image analysis to improve biometric security systems  
  by Jerzy Mikulik 
  Abstract: In this paper a concept of an access-control system based on mantrap and an analysis of static images from surveillance cameras is introduced. The results of this concept have been obtained by the analysis of images from surveillance cameras or digital cameras. These devises are not actual elements of mantraps. The described method can be used not only to count the number of people inside a mantrap, but also to detect threats such as a person with/without luggage, left luggage, an unaccompanied child, etc. The motivation of this study is the high importance of the ability to detect such events in a protected area for the safety of people and data. The method can be used as a supplement to a system of biometric identification of persons. The paper presents the structure of a threat detection system and subsequent operations on recorded images made by an application specially developed for this purpose. The paper discusses methods of collecting and recording data as well as methods of contextual and non-contextual processing of images that have a decisive impact on the final form of the material used for comparison with a database of reference images. Moreover, basic information on the Hamming neural network used for object classification is presented.
  Keywords: access control; CCTV; mantrap; image processing; people recognition; threats in security system.
   
• User Verification Based on the Support Vector Machine Using Intra-Body Propagation Signals  
  by Isao Nakanishi, Yuuta Sodani, Shigang Li 
  Abstract: Use of intra-body propagation signals has been proposed for biometric authentication. However, verification performance of the conventional method is low. To overcome this limitation, this study introduces the support vector machine (SVM) into the verification process, which improves the verification rate to approximately 83%. However, the correct acceptance rate of genuine users using only SVM is 49%, which is too low for practical applications. Thus, we introduce the concept of one versus one (1vs1) SVM. Each 1vs1~SVM distinguishes a genuine (authorized) user from another (unauthorized) user. Verification is achieved on the basis of a majority rule using plural 1vs1~SVMs related to a genuine user. The correct acceptance rate is greatly improved to 84% while maintaining equivalent verification performance. As a result, it is further confirmed that an intra-body propagation signal is a potential new biometric trait.
  Keywords: biometrics; intra-body propagation signal; support vector machine; 1vs1 SVM; correct acceptance rate
   
• Compensating for Pose and Illumination in Unconstrained Periocular Biometrics  
  by Hugo Proença, Chandrashekhar Padole 
  Abstract: In the context of less constrained biometrics recognition, the use of information from the vicinity of the eyes (periocular) is considered with high potential and motivated several recent proposals. In this paper we focus on two factors that are known to degrade the performance of periocular recognition: varying illumination conditions and subjects pose. Hence, this paper has three major purposes: 1) describe the decreases in performance due to varying illumination and subjects poses; 2) propose two techniques to improve the robustness to these factors; and 3) announce the availability of an annotated data set of periocular data (UBIPosePr), where poses vary in regular intervals, turning it specially suitable to assess the effects of misalignments between camera and subjects in periocular recognition.
  Keywords: Unconstrained Biometrics, Periocular Recognition, Illumination Compensation, Pose Compensation, Pose Estimation.
   
• New Algorithms for Improved Speaker Identification  
  by Eric Fang, John Gowdy 
  Abstract: This paper focuses on developing new algorithms for improving speaker identification accuracy. Forty nine male speakers from the DARPA resource management continuous speech database were used for training and testing. Mel-frequency cepstral coefficients (MFCC) components were used for training and testing. Vector quantization (VQ) was used for classification. In addition to presenting identification results, this paper shows the error reduction rate relative to a baseline system.
  Keywords: speaker identification, mel-frequency cepstral coefficients (MFCC), vector quantization (VQ)