Classical vs. biometric features in the 2013 speaker recognition evaluation in mobile environments

MFCC coefficients extracted from the power spectral density of speech as a whole, seems to have become the de facto standard in the area of speaker recognition, as demonstrated by its use in almost all systems submitted to the 2013 Speaker Recognition Evaluation (SRE) in Mobile Environment [1], thus relegating to background this component of the recognition systems. However, in this article we will show that selecting the adequate speaker characterization system is as important as the selection of the classifier. To accomplish this we will compare the recognition rates achieved by different recognition systems that relies on the same classifier (GMM-UBM) but connected with different feature extraction systems (based on both classical and biometric parameters). As a result we will show that a gender dependent biometric parameterization with a simple recognition system based on GMM- UBM paradigm provides very competitive or even better recognition rates when compared to more complex classification systems based on classical features

New methodology for the integration of biometric features in speaker recognition systems applied to security environments

La cuestión principal abordada en esta tesis doctoral es la mejora de los sistemas biométricos de reconocimiento de personas a partir de la voz, proponiendo el uso de una nueva parametrización, que hemos denominado parametrización biométrica extendida dependiente de género (GDEBP en sus siglas en inglés). No se propone una ruptura completa respecto a los parámetros clásicos sino una nueva forma de utilizarlos y complementarlos. En concreto, proponemos el uso de parámetros diferentes dependiendo del género del locutor, ya que como es bien sabido, la voz masculina y femenina presentan características diferentes que deberán modelarse, por tanto, de diferente manera. Además complementamos los parámetros clásicos utilizados (MFFC extraídos de la señal de voz), con un nuevo conjunto de parámetros extraídos a partir de la deconstrucción de la señal de voz en sus componentes de fuente glótica (más relacionada con el proceso y órganos de fonación y por tanto con características físicas del locutor) y de tracto vocal (más relacionada con la articulación acústica y por tanto con el mensaje emitido). Para verificar la validez de esta propuesta se plantean diversos escenarios, utilizando diferentes bases de datos, para validar que la GDEBP permite generar una descripción más precisa de los locutores que los parámetros MFCC clásicos independientes del género. En concreto se plantean diferentes escenarios de identificación sobre texto restringido y texto independiente utilizando las bases de datos de HESPERIA y ALBAYZIN. El trabajo también se completa con la participación en dos competiciones internacionales de reconocimiento de locutor, NIST SRE (2010 y 2012) y MOBIO 2013. En el primer caso debido a la naturaleza de las bases de datos utilizadas se obtuvieron resultados cercanos al estado del arte, mientras que en el segundo de los casos el sistema presentado obtuvo la mejor tasa de reconocimiento para locutores femeninos. A pesar de que el objetivo principal de esta tesis no es el estudio de sistemas de clasificación, sí ha sido necesario analizar el rendimiento de diferentes sistemas de clasificación, para ver el rendimiento de la parametrización propuesta. En concreto, se ha abordado el uso de sistemas de reconocimiento basados en el paradigma GMM-UBM, supervectores e i-vectors. Los resultados que se presentan confirman que la utilización de características que permitan describir los locutores de manera más precisa es en cierto modo más importante que la elección del sistema de clasificación utilizado por el sistema. En este sentido la parametrización propuesta supone un paso adelante en la mejora de los sistemas de reconocimiento biométrico de personas por la voz, ya que incluso con sistemas de clasificación relativamente simples se consiguen tasas de reconocimiento realmente competitivas. ABSTRACT The main question addressed in this thesis is the improvement of automatic speaker recognition systems, by the introduction of a new front-end module that we have called Gender Dependent Extended Biometric Parameterisation (GDEBP). This front-end do not constitute a complete break with respect to classical parameterisation techniques used in speaker recognition but a new way to obtain these parameters while introducing some complementary ones. Specifically, we propose a gender-dependent parameterisation, since as it is well known male and female voices have different characteristic, and therefore the use of different parameters to model these distinguishing characteristics should provide a better characterisation of speakers. Additionally, we propose the introduction of a new set of biometric parameters extracted from the components which result from the deconstruction of the voice into its glottal source estimate (close related to the phonation process and the involved organs, and therefore the physical characteristics of the speaker) and vocal tract estimate (close related to acoustic articulation and therefore to the spoken message). These biometric parameters constitute a complement to the classical MFCC extracted from the power spectral density of speech as a whole. In order to check the validity of this proposal we establish different practical scenarios, using different databases, so we can conclude that a GDEBP generates a more accurate description of speakers than classical approaches based on gender-independent MFCC. Specifically, we propose scenarios based on text-constrain and text-independent test using HESPERIA and ALBAYZIN databases. This work is also completed with the participation in two international speaker recognition evaluations: NIST SRE (2010 and 2012) and MOBIO 2013, with diverse results. In the first case, due to the nature of the NIST databases, we obtain results closed to state-of-the-art although confirming our hypothesis, whereas in the MOBIO SRE we obtain the best simple system performance for female speakers. Although the study of classification systems is beyond the scope of this thesis, we found it necessary to analise the performance of different classification systems, in order to verify the effect of them on the propose parameterisation. In particular, we have addressed the use of speaker recognition systems based on the GMM-UBM paradigm, supervectors and i-vectors. The presented results confirm that the selection of a set of parameters that allows for a more accurate description of the speakers is as important as the selection of the classification method used by the biometric system. In this sense, the proposed parameterisation constitutes a step forward in improving speaker recognition systems, since even when using relatively simple classification systems, really competitive recognition rates are achieved.

Biometric template security using higher order spectra

A method of improving the security of biometric templates which satisfies desirable properties such as (a) irreversibility of the template, (b) revocability and assignment of a new template to the same biometric input, (c) matching in the secure transformed domain is presented. It makes use of an iterative procedure based on the bispectrum that serves as an irreversible transformation for biometric features because signal phase is discarded each iteration. Unlike the usual hash function, this transformation preserves closeness in the transformed domain for similar biometric inputs. A number of such templates can be generated from the same input. These properties are illustrated using synthetic data and applied to images from the FRGC 3D database with Gabor features. Verification can be successfully performed using these secure templates with an EER of 5.85%

Ocular refractive and biometric characteristics in patients with tilted disc syndrome

BACKGROUND Tilted disc syndrome (TDS) is associated with characteristic ocular findings. The purpose of this study was to evaluate the ocular, refractive, and biometric characteristics in patients with TDS. METHODS This case-control study included 41 eyes of 25 patients who had established TDS and 40 eyes of 20 healthy control subjects. All participants underwent a complete ocular examination, including refraction and analysis using Fourier transformation, slit lamp biomicroscopy, pachymetry, keratometry, and ocular biometry. Corneal topography examinations were performed in the syndrome group only. RESULTS There were no significant differences in spherical equivalent (P = 0.13) and total astigmatism (P = 0.37) between groups. However, mean best spectacle-corrected visual acuity (Log Mar) was significantly worse in TDS patients (P = 0.003). The lenticular astigmatism was greater in the syndrome group, whereas the corneal component was greater in controls (P = 0.059 and P = 0.028, respectively). The measured biometric features were the same in both groups, except for the lens thickness and lens-axial length factor, which were greater in the TDS group (P = 0.007 and P = 0.055, respectively). CONCLUSIONS Clinically significant lenticular astigmatism, more oblique corneal astigmatism, and thicker lenses were characteristic findings in patients with TDS.

Ocular refractive and biometric characteristics in patients with tilted disc syndrome

Purpose: To evaluate the ocular refractive and biometric characteristics in patients with tilted disc syndrome (TDS). Methods: This case-control study comprised 41 eyes of 25 patients with established TDS and forty eyes of 20 age- and sex-matched healthy control subjects. All had a complete ocular examination including refraction and analysis using Fourier transformation, slit lamp biomicroscopy, pachymetry keratometry, and ocular biometry. Corneal topography examinations were performed in the syndrome group only. Results: There were no significant differences in spherical equivalent (p = 0.334) and total astigmatism (p= 0.246) between groups. However, mean best spectacular corrected visual acuity was significantly worse in TDS patients (P < 0.001). The lenticular astigmatism was significantly greater in the syndrome group, while the corneal component was greater in the controls (p = 0.004 and p = 0.002, respectively). The measured biometric features were the same in both groups, except for the lens thickness, relative lens position, and lens-axial length factor which were greater in the TDS group (p = 0.002, p = 0.015, and p = 0.025, respectively). Conclusions: Clinically significant lenticular astigmatism, more oblique corneal astigmatism, and thicker lens were characteristic findings in patients with TDS.

IRIS BIOMETRIC RECOGNITION SYSTEM EMPLOYING CANNY OPERATOR

Biometrics has become important in security applications. In comparison with many other biometric features, iris recognition has very high recognition accuracy because it depends on iris which is located in a place that still stable throughout human life and the probability to find two identical iris's is close to zero. The identification system consists of several stages including segmentation stage which is the most serious and critical one. The current segmentation methods still have limitation in localizing the iris due to circular shape consideration of the pupil. In this research, Daugman method is done to investigate the segmentation techniques. Eyelid detection is another step that has been included in this study as a part of segmentation stage to localize the iris accurately and remove unwanted area that might be included. The obtained iris region is encoded using haar wavelets to construct the iris code, which contains the most discriminating feature in the iris pattern. Hamming distance is used for comparison of iris templates in the recognition stage. The dataset which is used for the study is UBIRIS database. A comparative study of different edge detector operator is performed. It is observed that canny operator is best suited to extract most of the edges to generate the iris code for comparison. Recognition rate of 89% and rejection rate of 95% is achieved

Semiautomatic dental recognition using a graph-based segmentation algorithm and teeth shapes features

Dental recognition is very important for forensic human identification, mainly regarding the mass disasters, which have frequently happened due to tsunamis, airplanes crashes, etc. Algorithms for automatic, precise, and robust teeth segmentation from radiograph images are crucial for dental recognition. In this work we propose the use of a graph-based algorithm to extract the teeth contours from panoramic dental radiographs that are used as dental features. In order to assess our proposal, we have carried out experiments using a database of 1126 tooth images, obtained from 40 panoramic dental radiograph images from 20 individuals. The results of the graph-based algorithm was qualitatively assessed by a human expert who reported excellent scores. For dental recognition we propose the use of the teeth shapes as biometric features, by the means of BAS (Bean Angle Statistics) and Shape Context descriptors. The BAS descriptors showed, on the same database, a better performance (EER 14%) than the Shape Context (EER 20%). © 2012 IEEE.

Classification of handwriting patterns in patients with Parkinson´s disease, using a biometric sensor

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Non-intrusive continuous user behaviour analysis using computerized systems

There are different ways to authenticate humans, which is an essential prerequisite for access control. The authentication process can be subdivided into three categories that rely on something someone i) knows (e.g. password), and/or ii) has (e.g. smart card), and/or iii) is (biometric features). Besides classical attacks on password solutions and the risk that identity-related objects can be stolen, traditional biometric solutions have their own disadvantages such as the requirement of expensive devices, risk of stolen bio-templates etc. Moreover, existing approaches provide the authentication process usually performed only once initially. Non-intrusive and continuous monitoring of user activities emerges as promising solution in hardening authentication process: iii-2) how so. behaves. In recent years various keystroke dynamic behavior-based approaches were published that are able to authenticate humans based on their typing behavior. The majority focuses on so-called static text approaches, where users are requested to type a previously defined text. Relatively few techniques are based on free text approaches that allow a transparent monitoring of user activities and provide continuous verification. Unfortunately only few solutions are deployable in application environments under realistic conditions. Unsolved problems are for instance scalability problems, high response times and error rates. The aim of this work is the development of behavioral-based verification solutions. Our main requirement is to deploy these solutions under realistic conditions within existing environments in order to enable a transparent and free text based continuous verification of active users with low error rates and response times.

Robust image hash functions using higher order spectra

Robust hashing is an emerging field that can be used to hash certain data types in applications unsuitable for traditional cryptographic hashing methods. Traditional hashing functions have been used extensively for data/message integrity, data/message authentication, efficient file identification and password verification. These applications are possible because the hashing process is compressive, allowing for efficient comparisons in the hash domain but non-invertible meaning hashes can be used without revealing the original data. These techniques were developed with deterministic (non-changing) inputs such as files and passwords. For such data types a 1-bit or one character change can be significant, as a result the hashing process is sensitive to any change in the input. Unfortunately, there are certain applications where input data are not perfectly deterministic and minor changes cannot be avoided. Digital images and biometric features are two types of data where such changes exist but do not alter the meaning or appearance of the input. For such data types cryptographic hash functions cannot be usefully applied. In light of this, robust hashing has been developed as an alternative to cryptographic hashing and is designed to be robust to minor changes in the input. Although similar in name, robust hashing is fundamentally different from cryptographic hashing. Current robust hashing techniques are not based on cryptographic methods, but instead on pattern recognition techniques. Modern robust hashing algorithms consist of feature extraction followed by a randomization stage that introduces non-invertibility and compression, followed by quantization and binary encoding to produce a binary hash output. In order to preserve robustness of the extracted features, most randomization methods are linear and this is detrimental to the security aspects required of hash functions. Furthermore, the quantization and encoding stages used to binarize real-valued features requires the learning of appropriate quantization thresholds. How these thresholds are learnt has an important effect on hashing accuracy and the mere presence of such thresholds are a source of information leakage that can reduce hashing security. This dissertation outlines a systematic investigation of the quantization and encoding stages of robust hash functions. While existing literature has focused on the importance of quantization scheme, this research is the first to emphasise the importance of the quantizer training on both hashing accuracy and hashing security. The quantizer training process is presented in a statistical framework which allows a theoretical analysis of the effects of quantizer training on hashing performance. This is experimentally verified using a number of baseline robust image hashing algorithms over a large database of real world images. This dissertation also proposes a new randomization method for robust image hashing based on Higher Order Spectra (HOS) and Radon projections. The method is non-linear and this is an essential requirement for non-invertibility. The method is also designed to produce features more suited for quantization and encoding. The system can operate without the need for quantizer training, is more easily encoded and displays improved hashing performance when compared to existing robust image hashing algorithms. The dissertation also shows how the HOS method can be adapted to work with biometric features obtained from 2D and 3D face images.

Implementação de verificações biométricas processadas em cartões inteligentes a multi-aplicações.

As biometrias vêm sendo utilizadas como solução de controle de acesso a diversos sistemas há anos, mas o simples uso da biometria não pode ser considerado como solução final e perfeita. Muitos riscos existem e não devem ser ignorados. A maioria dos problemas está relacionada ao caminho de transmissão entre o local onde os usuários requerem seus acessos e os servidores onde são guardados os dados biométricos capturados em seu cadastro. Vários tipos de ataques podem ser efetuados por impostores que desejam usar o sistema indevidamente. Além dos aspectos técnicos, existe o aspecto social. É crescente a preocupação do usuário tanto com o armazenamento quanto o uso indevido de suas biometrias, pois é um identificador único e, por ser invariável no tempo, pode ser perdido para sempre caso seja comprometido. O fato de que várias empresas com seus diferentes servidores guardarem as biometrias está causando incomodo aos usuários, pois as torna mais suscetíveis à ataques. Nesta dissertação, o uso de cartões inteligentes é adotado como possível solução para os problemas supracitados. Os cartões inteligentes preparados para multi-aplicações são usados para realizar as comparações biométricas internamente. Dessa forma, não seria mais necessário utilizar diversos servidores, pois as características biométricas estarão sempre em um único cartão em posse do dono. Foram desenvolvidas e implementadas três diferentes algoritmos de identificação biométrica utilizando diferentes características: impressão digital, impressão da palma da mão e íris. Considerando a memória utilizada, tempo médio de execução e acurácia, a biometria da impressão da palma da mão obteve os melhores resultados, alcançando taxas de erro mínimas e tempos de execução inferiores a meio segundo.

Análise de desempenho do Criptossistema Fuzzy Vault em aplicações reais

Pós-graduação em Ciência da Computação - IBILCE

Reconhecimento semiautomático de dentes para a identificação humana forense

Pós-graduação em Ciência da Computação - IBILCE

An optical study of human ocular dimensions

Many workers have studied the ocular components which occur in eyes exhibiting differing amounts of central refractive error but few have ever considered the additional information that could be derived from a study of peripheral refraction. Before now, peripheral refraction has either been measured in real eyes or has otherwise been modelled in schematic eyes of varying levels of sophistication. Several differences occur between measured and modelled results which, if accounted for, could give rise to more information regarding the nature of the optical and retinal surfaces and their asymmetries. Measurements of ocular components and peripheral refraction, however, have never been made in the same sample of eyes. In this study, ocular component and peripheral refractive measurements were made in a sample of young near-emmetropic, myopic and hyperopic eyes. The data for each refractive group was averaged. A computer program was written to construct spherical surfaced schematic eyes from this data. More sophisticated eye models were developed making use of linear algebraic ray tracing program. This method allowed rays to be traced through toroidal aspheric surfaces which were translated or rotated with respect to each other. For simplicity, the gradient index optical nature of the crystalline lens was neglected. Various alterations were made in these eye models to reproduce the measured peripheral refractive patterns. Excellent agreement was found between the modelled and measured peripheral refractive values over the central 70o of the visual field. This implied that the additional biometric features incorporated in each eye model were representative of those which were present in the measured eyes. As some of these features are not otherwise obtainable using in vivo techniques, it is proposed that the variation of refraction in the periphery offers a very useful optical method for studying human ocular component dimensions.

Speaker identification using higher order spectral phase features and their effectiveness vis-a-vis Mel-Cepstral features

The effectiveness of higher-order spectral (HOS) phase features in speaker recognition is investigated by comparison with Mel Cepstral features on the same speech data. HOS phase features retain phase information from the Fourier spectrum unlikeMel–frequency Cepstral coefficients (MFCC). Gaussian mixture models are constructed from Mel– Cepstral features and HOS features, respectively, for the same data from various speakers in the Switchboard telephone Speech Corpus. Feature clusters, model parameters and classification performance are analyzed. HOS phase features on their own provide a correct identification rate of about 97% on the chosen subset of the corpus. This is the same level of accuracy as provided by MFCCs. Cluster plots and model parameters are compared to show that HOS phase features can provide complementary information to better discriminate between speakers.