By Topic

Signal Processing, IET

Issue 4 • Date July 2009

Filter Results

Displaying Results 1 - 12 of 12
  • Special issue on biometric recognition (editorial)

    Page(s): 233 - 235
    Save to Project icon | PDF file iconPDF (83 KB)  
    Freely Available from IEEE
  • Speaker verification under mismatched data conditions

    Page(s): 236 - 246
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (705 KB)  

    This study presents investigations into the effectiveness of the state-of-the-art speaker verification techniques (i.e. GMM-UBM and GMM-SVM) in mismatched noise conditions. Based on experiments using white and real world noise, it is shown that the verification performance offered by these methods is severely affected when the level of degradation in the test material is different from that in the training utterances. To address this problem, a modified realisation of the parallel model combination (PMC) method is introduced and a new form of test normalisation (T-norm), termed condition adjusted T-norm, is proposed. It is experimentally demonstrated that the use of these techniques with GMM-UBM can significantly enhance the accuracy in mismatched noise conditions. Based on the experimental results, it is observed that the resultant relative improvement achieved for GMM-UBM (under the most severe mismatch condition considered) is in excess of 70%. Additionally, it is shown that the improvement in the verification accuracy achieved in this way is higher than that obtainable with the direct use of PMC with GMM-UBM. Moreover, it is found that while the accuracy performance of GMM-SVM can also considerably benefit from the use of these techniques, the extensive computational cost involved in this case severely limits the use of such a combined approach in practice. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Using jitter and shimmer in speaker verification

    Page(s): 247 - 257
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (360 KB)  

    Jitter and shimmer are measures of the fundamental frequency and amplitude cycle-to-cycle variations, respectively. Both features have been largely used for the description of pathological voices, and since they characterise some aspects concerning particular voices, they are expected to have a certain degree of speaker specificity. In the current work, jitter and shimmer are successfully used in a speaker verification experiment. Moreover, both measures are combined with spectral and prosodic features using several types of normalisation and fusion techniques in order to obtain better verification results. The overall speaker verification system is also improved by using histogram equalisation as a normalisation technique previous to fusing the features by support vector machines. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Phase congruency features for palm-print verification

    Page(s): 258 - 268
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (467 KB)  

    The existing palm-print verification schemes have demonstrated good verification performance when identity claims have to be verified based on palm-print images of adequate quality (e.g. acquired in controlled illumination conditions, free from distortions caused by the pressure applied to the surface of the scanner etc.). However, most of these schemes struggle with their verification performance when features have to be extracted from palm-print images of a poorer quality. In this study the authors present a novel palm-print feature extraction approach which deals with the presented problem by employing the two-dimensional phase congruency model for line-feature extraction. The proposed approach first computes a set of phase congruency features from a palm-print image and subsequently performs linear discriminant analysis on the computed features to represent them in a more compact manner. The approach was tested on two contrasting databases, namely, on the FE-LUKS and on the PolyU database. Encouraging results were achieved on both databases. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Finger-based personal authentication: a comparison of feature-extraction methods based on principal component analysis, most discriminant features and regularised-direct linear discriminant analysis

    Page(s): 269 - 281
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1120 KB)  

    In this study, feature-extraction methods based on principal component analysis, most discriminant features, and regularised-direct linear discriminant analysis (RD-LDA) are tested and compared in an experimental finger-based personal authentication system. The system is multimodal and based on features extracted from eight regions of the hand: four fingerprints (the prints of the finger tips) and four digitprints (the prints of the fingers between the first and third phalanges). All of the regions are extracted from one-shot grey-level images of the palmar surface of four fingers of the right hand. The identification and verification experiments were conducted on a database consisting of 1840 finger images (184 people). The experiments showed that the best results were obtained with the RD-LDA-based feature-extraction method -99.98% correct identification for 920 tests and an equal error rate of 0.01% for 64170 verification tests. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Face recognition from synchronised visible and near-infrared images

    Page(s): 282 - 288
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (224 KB)  

    The improvement in face recognition that can be obtained from the simultaneous availability of normal and near-infrared (NIR) images is quantitatively measured. The authors designed a camera that can perform a simultaneous acquisition of a NIR image and visible images (VI) and therefore built a face database with this camera aiming at comparing the performance of several algorithms on both types of images when illumination variations occur. The authors noticed the stability of the performance of all the tested algorithms on infrared images upon illumination variation, and the improvement in performance that results from the fusion of these two different types of images. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Complete face logs for video sequences using face quality measures

    Page(s): 289 - 300
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1007 KB)  

    Reducing a video sequence containing a human's face to just a few high-quality face images (face log) has a considerable importance in applications related to face processing. This face log can be considered as a concise representation of the video sequence. Producing such a complete face log is the focus of this paper. To decide about the presence of each face in the face log, their quality using four facial features is assessed. Relative quality scores are assigned to these features and then combined into one quality score for each face using a fuzzy inference engine. The authors introduce a method for choosing the M-best images for construction of the face logs. These best images are selected as local maxima in different temporal periods of the sequence. The introduced system has been evaluated using four different datasets including still images and video sequences under various conditions. Experimental results show the success of the system in finding the best face images and using these for producing complete face logs. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • On the use of quality measures in face and speaker identity verification based on video and audio streams

    Page(s): 301 - 309
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (503 KB)  

    This study addresses the advantage of adding quality information of the biometric signals into a multimedia-based (video and audio) identity verification system. The quality information of the biometric signals can be used in several ways and stages in the biometric system. In this study, the authors introduce quality-based decisions in two stages: score normalisation and frame selection. Quality-based score normalisation helps to handle quality dependent drifts in the scores distributions. We derive a necessary and sufficient condition for reducing error when introducing quality-based score normalisation and present a score normalising technique. Additionally, the number of frontal faces and speech vectors extracted from the video and audio streams allows quality-based selection of frames, both in training and test, to preserve quality in the statistical representation of the signals. For these two stages we defined some quality measures for speaker and frontal face signals and run experiments to show the reliability of the proposed techniques over the BANCA database. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Improving biometric verification with class-independent quality information

    Page(s): 310 - 321
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (634 KB)  

    Existing approaches to biometric classification with quality measures make a clear distinction between the single-modality applications and the multi-modal scenarios. This study bridges this gap with Q-stack, a stacking-based classifier ensemble, which uses the class-independent signal quality measures and baseline classifier scores in order to improve the accuracy of uni- and multi-modal biometric classification. The seemingly counterintuitive notion of using class-independent quality information for improving class separation by considering quality measures as conditionally relevant classification features. The authors present Q-stack as a generalised framework of classification with quality information is explained, and argue that existing methods of classification with quality measures are its special cases. The authors further demonstrate the application of Q-stack on the task of biometric identity verification using face and fingerprint modalities, and show that the use of the proposed technique allows a systematic reduction of the error rates below those of the baseline classifiers, in scenarios involving single and multiple biometric modalities. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Score bi-Gaussian equalisation for multimodal person verification

    Page(s): 322 - 332
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (599 KB)  

    Multimodal biometric fusion at score level can be performed by means of combinatory or classificatory techniques. In the first case, it is straightforward that the normalisation of the scores is a very important issue for the success of the fusion process. In the classificatory approach as, for instance, in support vector machine (SVM)-based systems, only simple normalisation methods are usually applied. In this work, histogram equalisation of biometric score distribution is successfully applied in a multimodal person verification system composed by prosodic, speech spectrum and face information. Furthermore, a new bi-Gaussian equalisation (BGEQ) is introduced, which takes into account the separate statistics of the genuine and impostor scores by using as a reference a sum of two Gaussian functions, whose standard deviations model the overlap between the genuine and impostor lobes of the original distributions. Multimodal verification experiments are shown, where prosodic and speech spectrum scores are provided by speech experts using the Switchboard-I database, and face scores are obtained by a face recognition expert using XM2VTS database. BGEQ in combination with an SVM fusion system with a polynomial kernel has obtained the best results and has outperformed in more than a 21.29% the results obtained by min%max normalisation. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • User authentication using keystroke dynamics for cellular phones

    Page(s): 333 - 341
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (581 KB)  

    A new approach for keystroke-based authentication when using a cellular phone keypad as input device is presented. In the proposed method, users are authenticated using keystroke dynamics acquired when typing fixed alphabetic strings on a mobile phone keypad. The employed statistical classifier is able to perform user verification with an average equal error rate of about 13%. The obtained experimental results suggest that, when using mobile devices, a strong secure authentication scheme cannot rely on the sole keystroke dynamics, which however can be a module of a more complex system including, as basic security, a password-based protocol eventually hardened by keystroke analysis. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Balancing performance factors in multisource biometric processing platforms

    Page(s): 342 - 351
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (204 KB)  

    It is generally recognised that no one biometric data source or processing platform is universally appropriate for optimising performance across all problem domains. Multibiometric processors, which combine identity information obtained from more than one biometric source are commonly promoted as optimal structures for maximising performance, and much research has been carried out to investigate appropriate strategies for combining the available information. However, the techniques of multiclassifier pattern recognition also offer opportunities to improve the performance of systems operating within a unimodal environment, yet such solutions have been less extensively investigated in the specific case of biometric applications. This study presents an empirical study of the relations between these two different approaches to enhancing the performance indicators delivered by biometric systems. In particular we are interested to increase our understanding of the relative merits of, on the one hand, multiclassifier/single modality systems and, on the other, full multibiometric configurations. We focus our study on three modalities, the fingerprint and hand geometry (two physiological biometrics) and the handwritten signature (a behavioural biometric). View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.

Aims & Scope

IET Signal Processing publishes novel contributions in signal processing.

Full Aims & Scope

Meet Our Editors

IET Research Journals
iet_spr@theiet.org