Audio-Visual Person Recognition for Security and · Web viewAudio-Visual Person Recognition for Security and Access Control Farzin Deravi University of Kent at Canterbury Table of

JTAPJISC Technology Applications Programme

Audio-Visual Person Recognition for Security

and Access ControlFarzin Deravi

University of Kent at Canterbury

Report: 38 JISC TechnologyApplications Programme

Joint Information Systems Committee

September 1999

JISC Technology Applications Programme

Audio-Visual Person Recognition for Security

and Access ControlFarzin Deravi

University of Kent at Canterbury

The JISC Technology Applications Programme is an initiative of the Joint Information Systems Committee of the Higher Education Funding Councils.

For more information contact:Tom FranklinJTAP Programme ManagerComputer BuildingUniversity of ManchesterManchesterM13 9PL

email: [email protected]: http://www.jtap.ac.uk/

JTAP 552 – Audio-Visual Person Recognition for Security and Access Control

Table of ContentsAbstract........................................................................................................................... 4

1 Introduction................................................................................................................. 41.1 Biometric Technologies...................................................................................................4

1.2 Scope of this report.........................................................................................................5

2 Biometric Systems........................................................................................................ 62.1 Types of Biometrics........................................................................................................6

2.2 Biometric System Components and Processes...............................................................7

2.3 Biometrics System Considerations.................................................................................8

2.4 Performance Assessment................................................................................................9

3 Audio-Visual Biometrics............................................................................................ 103.1 Motivation for Joint Audio-Visual Recognition...........................................................10

3.2 Audio-Visual Features..................................................................................................11

3.3 Feature Classification...................................................................................................12

3.4 Audio-Visual Fusion.....................................................................................................13

4 Survey of Commercial Systems..................................................................................15

5 Applicability in smart-card technologies....................................................................16

6 Potential Applications in Higher Education..............................................................166.1 Facilities Access Control Systems.................................................................................16

6.2 Workstation and Network Access................................................................................16

6.3 Attendance Monitoring.................................................................................................17

6.4 Monitored Student Assessments...................................................................................17

6.5 Security Applications....................................................................................................17

6.6 Payment Systems..........................................................................................................17

7 Standardisation.......................................................................................................... 177.1 Biometric Application Programmers Interfaces..........................................................17

8 Organisations............................................................................................................. 19

9 Conclusions & Recommendations.............................................................................20

10 Acknowledgements................................................................................................... 21

11 References & Resources........................................................................................... 21

Appendix A – Table of Commercial Systems................................................................24

Appendix B - Survey of Companies Providing Audio-Visual Biometric Solutions......28

3


Appendix C – Contact Address for Audio-Visual Biometric System Providers.............36

4


Audio-Visual Person Recognition for Security and Access Control

JTAP – 552 Report

Abstract

This report considers multimodal biometric systems and their applicability to access control, authentication and security applications in a Higher Education / University setting. In particular joint audio-visual recognition systems are reviewed. Alternative strategies for feature extraction and sensor fusion are considered and contrasted. A survey of new commercial systems claiming to offer multimodal biometrics is presented and the potential applications of such systems in a University environment are considered. Issues related to performance assessment, deployment and standardisation are discussed. Finally future directions of biometric systems development are explored.

1 Introduction

1.1 Biometric Technologies

Biometrics can be defined as measurable characteristics of the individual based on their physiological features or behavioural patterns that can be used to recognise or verify their identity. Biometric technologies attempt to automate the measurement and comparison of such characteristics for recognising individuals. Many different technologies have recently been developed for person recognition and identity authentication and some examples include measures based on information from handwriting (especially signatures), fingerprint, face, voice, retina, iris, hand or ear shape and gait data.

Biometric technologies were first proposed for high-security specialist applications but are now emerging as key elements in the developing electronic commerce and online systems revolution as well as for off-line and standalone security systems.

These technologies will provide important components in regulating and monitoring access and presence. Significant application areas include electronic commerce, security monitoring, database access, border control and immigration, forensic investigations and telemedicine.

The development of biometric technologies, beyond traditional high-security applications, has a compelling financial motivation. Transaction security is critical to e-commerce’s future development and there is serious concern over the adequacy of

5


current solutions. The problem with personal identification numbers, PINs, and identity tokens, such as cards, is that they do not guarantee the identity of the person who uses them. Credit card fraud has been reported to cost some 450 million dollars per year and ATM fraud has been reported as 3 billion dollars. Biometric systems have the advantage that they are tightly bound to the individual and can not be easily used by an impostor.

In addition to such verification applications, biometric systems can be used for the less constrained problem of automatic identification of individuals. In these applications the biometric system carries out a “one-to-many” search of its stored models of individuals’ identities. They can be therefore used in security applications for example to detect fraud or intrusion. This is a more difficult and time-consuming task than identity verification.

Until recently biometric machines have been relatively expensive costing thousands of pounds per unit. In addition they have lacked the required speed and accuracy except in special circumstances or with extensive user training. However, more recently the situation has improved with the introduction of machines with introductory prices below 1000 pounds per unit and improved performance. It is expected that these positive trends will accelerate. Falling prices should allow more biometric technologies to be employed into medium security environments.

While some commercial biometric products have recently become available, most of these technologies are still in a research and experimental stage. More research and development work is required on individual biometric modalities to improve their robustness and increase their performance for specific applications.

Different applications in which biometric systems may be used have different constraints and priorities. Consequently appropriate techniques and interfaces should be investigated, tailoring the biometric systems used to the needs specified by the user community. The procedural, psychological and legal aspects of biometrics in use will inform the development of appropriate technologies. Consideration of such human factors and user acceptability issues is crucial to the adoption of these technologies.

1.2 Scope of this report

This report presents an evaluation of new technologies for person identification using joint audio and video models of individuals for greater robustness and resistance to unauthorised access. The combined and simultaneous use of audio and video information provides a greater degree of security as tampering any one of these sources would not be enough for false access and authentication. Joint audio-visual models for individuals may be encrypted and placed on smart cards and used for authentication and controlled access to buildings and resources.

6


With the increasing availability of desktop video-conferencing and the decline in the price and availability of multimedia video capture and processing equipment the use of joint audio-visual processing for authentication and access control is becoming feasible and cost-effective. It is important to assess the state of the art and potential applicability of these technologies in the Higher Education context.

The report introduces the current techniques being explored for the extraction of biometrically significant audio and visual features. Next the various approaches for combining the information from the two sources is described. These include layered systems where the different modalities are considered in turn and fused systems where either the features or the decisions based on them are used simultaneously to come to a final decision on identity.

The report contains a section on potential application of such biometric authentication and recognition systems in a Higher Education setting. The report also contains a description of current commercial biometric systems that provide both audio and video modalities. Their ability to provide joint audio-visual functionality as well as issues relating to their cost and feasibility of deployment in a Higher Education setting is explored.

2 Biometric Systems

2.1 Types of Biometrics

Several different biometric modalities have emerged in recent years. The table below lists the more common biometric sources of identity information and key characteristics of some current systems; classified in broad terms:

Biometric Type Accuracy Ease of Use User AcceptanceFingerprint High Medium LowHand Geometry Medium High MediumVoice Medium High HighRetina High Low LowIris Medium Medium MediumSignature Medium Medium HighFace Low High High

It is important to note that some techniques, such as retinal scanning or finger print recognition, may offer high accuracy but may not be appropriate for some applications. This is due to the high level of co-operation required by the user or the social or psychological factors that may prove unacceptable to potential users.

7


Both voice and face recognition are considered to be easy to use and normally acceptable by potential users. However, their accuracy is currently less than some other biometric technologies, especially in unconstrained environments such as where the background sound and illumination is variable.

More information on the characteristics of specific biometric modalities can be found in [1-6].

2.2 Biometric System Components and Processes

There are two distinct phases of operation for biometric systems: enrolment and verification/identification. In the first phase identity information from users is added to the system. In the second phase live biometric information from users is compared with stored records. Typical biometric identification and recognition system may have the following components:

a) Capture: A sub-system for capturing samples of the biometric(s) to be used. This could be voice recordings or still facial images. Specific features will be extracted from the biometric samples to form templates for future comparisons. In the enrolment phase a number of such samples may be captured. A truly representative identity model may then be obtained from the features thus obtained. This enrolment process should ideally be simple and rapid, yet result in, good quality, representative templates. If the templates are of poor quality, this will affect the subsequent performance of the system. An elaborate and exhaustive enrolment process may be unacceptable.

b) Storage: The templates thus obtained will have to be stored for future comparison. This may be done at the biometric capture device or remotely in a central server accessible via a network. Another alternative is to store the template in a portable token such as a smart card. Each one of these options has its advantages and disadvantages (see [1]). In addition to template storage there is often a need for a secure audit trail for all the transactions of the system.

c) Comparison: If the biometric system is used in a verification setting, then the claimed user identity will have to be compared against the claimed reference template. The captured live biometric from the user will be compared with the claimed identity which may be provided by entering a pin, or presenting a card storing identity information. In a identification/recognition setting the live biometric will have to be compared with all the templates stored to see if there is a close match. In some systems it may be possible to automatically update the reference template after each valid match. This will make it possible for the system to adapt to gradual minor changes in user characteristics (e.g. due to ageing).

d) Interconnections: There is the need for interconnections between the capture device and the verification and storage components of the system. Often there are existing

8


access control and information systems into which the biometric system may have to be integrated. There is a need for generic networking and programming interfaces to allow easy interconnections for biometric systems. Security and efficiency will be key considerations.

2.3 Biometrics System Considerations

The following are some of the key issues that need to be considered in designing and applying biometric systems.

Robustness: It is important to consider how robust the system is to fraud and impersonation. Such fraud can occur at the enrolment stage as well as at the verification stage. Using more than one biometric modality can help combat fraud and increase robustness. Also the system should be robust to small variations to the users’ biometrics over time. For this, an adaptive system that gradually modifies the stored templates may be used.

Acceptability: The technology must be easy to use during both the enrolment and comparison phases. It must also be socially acceptable. The users would not accept a system that may threaten their privacy and confidentiality or that might appear to treat them as potential suspects and criminals. This accounts for the lower acceptability of fingerprint systems than voice or face recognition systems. A multimodal system is more capable to adapting to user’s requirements and capabilities.

Legal issues may also have to be considered in relation to biometric systems [7-11]. There may be concerns over potential intrusions into private lives by using biometric systems. The European Union’s comprehensive privacy legislation, the Directive on Data Protection, became effective on October 25, 1998 [8]. While it does not specifically mention biometrics, biometric identifiers are likely to fall within its legislative scope. The European Parliament has recently raised this issue in relation to European Community research efforts. Also, there is a growing lobby to limit and regulate the use of biometrics and surveillance technologies [10,11]. Legal issues must be considered for any potential application and appropriate measures must be taken. A clear public stance on the issue of privacy in relation to biometric technologies is required to ensure broad public acceptance.

Speed and Storage Requirements: The time required to enrol, verify or identify a person is of critical importance to the acceptance and applicability of the system. Ideally, the acceptable verification time should be of the order of one second or faster. The storage requirement for the templates is also an important issue, especially if the templates are to be stored in magnetic stripe or smart cards.

Integration: The hardware platform on which the system is to be implemented is a key concern. The software, hardware and networking requirements should ideally be

9


compatible with existing systems, allowing the biometric system to be integrated to the existing infrastructure. The system cost should be reasonable and the maintenance costs should be understood.

2.4 Performance Assessment

An important issue for the adoption of biometric technologies is to establish the performance of individual biometric modalities and overall systems in a credible and objective way.

For verification applications, a number of objective performance measures have been used to characterise the performance of biometric systems. In these applications a number of ‘clients’ are enrolled onto the system. An ‘impostor’ is defined as someone who is claiming to be someone else. The impostor may be someone who is not enrolled at all or someone who tries to claim the identity of someone else either intentionally or otherwise. When being verified the clients should be recognised as themselves and impostors should be rejected.

False Acceptance Rate (FAR) is defined as the ratio of impostors that were falsely accepted over the total number of impostors tested described as a percentage. This indicates the likelihood that an impostor may be falsely accepted and must be minimised in high-security applications.

False Reject Rate (FRR) is defined as the ratio of clients that are falsely rejected to the total number of clients tested described as a percentage. This indicates the probability that a valid user may be rejected by the system. Ideally this should also be minimised especially when the user community may be put-off from using the system if they are wrongly denied access.

The biometric verification process involves computing a distance between the stored template and the live sample. The decision to accept or reject is based on a pre-defined threshold. If the distance is less than this threshold then we can accept the sample. It is therefore clear that the performance of the system critically depends on the choice of this threshold and there is a trade-off between FRR and FAR. Vendors usually provide a means for controlling the threshold for their system in order to control the trade-off between FRR and FAR. The Equal Error Rate (EER) is the threshold level for which the FAR and the FRR are equal. Figure 1 shows a general example of the FRR and FAR curves. The EER is often quoted as a single figure to characterise the overall performance of biometric systems.

Another important performance parameter is the verification time defined as the average time taken for the verification process. This may include the time taken to present the live sample.

10


While some vendors quote the above performance parameters for their system under laboratory conditions there are seldom real world performance characteristics available for biometric systems. This is because it is almost impossible to account for the complexities of all possible real world conditions. For example the actual verification time will critically depend on user training, operating environment and psychological conditions such as stress. Vendor specifications should be seen only as rough guides to real world performance.

Error Rate%

DecisionThreshold

FARFRR

EER

EERthreshold

Figure 1 FRR and FAR curves.

The EU funded BIOTEST project is one initiative to provide objective performance characterisation of biometric products [12]. A National Biometric Test Centre has been established in the US and similar efforts are underway in other countries.

A number of databases have been developed for the evaluation of biometric systems [13]. For the testing of joint audio-visual systems a number of research databases have been gathered in recent years. The XM2VTS database is a European effort [14], while the BT DAVID database was gathered by BT and the University of Wales [15].

Developing new assessment strategies that allow meaningful comparisons between systems and solutions is an essential activity. This involves creating databases and putting together test procedures and systems for the online assessment of biometric technologies.

3 Audio-Visual Biometrics

3.1 Motivation for Joint Audio-Visual Recognition

11


The need to recognise individuals is vital to human life. The most natural way to do this is by recognising people’s faces and their voice. However, it is impossible to personally know everyone that an individual may have to interact with. Biometric devices and technologies automate the process of recognising individuals.

It is possible to base automated recognition systems on learned information (e.g. passwords, or PIN numbers) or tokens (e.g. identity cards) but by far the most natural and unique identifier is our physical characteristics. Fingerprints, shape of the hand, iris patterns, face and voice information have all been used for person identification and verification.

Amongst these techniques the least intrusive and most socially acceptable techniques are face and voice recognition. Face and voice recognition need minimal co-operation from the users and are not deemed to erode the privacy of the individuals as after all this is the basis of recognition by fellow humans.

Another advantage of voice and face recognition is that they can be done continuously, without the need for significant effort and active co-operation by the users. This makes them ideal for online applications or applications where continuous monitoring may be required.

The increasing availability and the low cost of audio and video sensors and the increasingly cost-effective computing power that is available makes it possible to consider deploying such systems for access control and monitoring.

One problem with using face or voice recognition is the robustness of these techniques to variable environmental conditions and to impersonation. It is possible to reduce the effect of these factors considerably by employing face and voice recognition concurrently and co-operatively. Such multimodal systems can be shown to be less sensitive to variations in speech patterns of a particular individual, to background noise, poor transmission conditions in remote applications and to determined attacks by impostors.

A comprehensive review of audio-visual recognition systems has been undertaken [16-17]. In this section a summary of the key technical issues is presented. These include the choice of features and methods for fusing these modalities.

3.2 Audio-Visual Features

3.2.1 Audio Features

In voice recognition the audio signal is sampled and quantised before feature extraction. A telephone quality system may be adequate for recognition purposes (8000 samples per

12


second, 8 bits quantisation) while systems with higher sampling rates and finer amplitude quantisation have also been developed. The sound signal must be segmented to isolate the speech waveform from non-speech sounds.

The features used are based on transforms of shot periods of speech signals called frames. These speech frames may be overlapping in time. Often a log frequency transform is used (the cepstrum) to transform each frame. The frames are typically 30ms long resulting a typical vector of 10 to 15 features per frame. Identity templates are formed using a combination of feature vectors obtained from a number of frames.

The variability of the sound capture systems and local acoustics and background noise and babble cause difficulties in acquiring reliable features. Enrolment can be a problem as several samples may be required to construct reliable templates. Much work has been done in this area and still continues to be done to overcome difficulties. For a tutorial introduction to speaker recognition techniques see [18].

3.2.2 Facial Features

Facial recognition has attracted a great deal of attention from researchers and continues to be an active research area. There are a number of problems associated with facial recognition. First the presence of a face or faces in a scene must be detected. Once the face has been detected it must be localised and a normalisation process may be required to bring the dimensions of the live facial sample and the one on which the template is based into alignment.

Some face recognition approaches use the whole face while others concentrate on facial components and/or regions such as lips or eyes. In some algorithms the grey-level or colour pixel information is used directly to produce features while in others the shapes of regions is first extracted and templates are based on the region shapes. The latter approach is more complex but may offer greater robustness against variations on illumination and occlusions.

A more recent approach is to rely on a sequence of face pictures instead of a still facial image. In this way dynamic features can be computed that relate to facial movements as opposed to static features that reflect the shape and texture of the face. This again produces more robustness while increasing the computational effort. More than one camera may be used to produce a three dimensional representation of the face and to protect against the use of photographs to gain unauthorised access.

For surveys of face recognition and related material see [19-22].

13


3.3 Feature Classification

Once the features are extracted from the voice or facial signals and templates representing each individual are constructed, there remains the task of classifying live samples.

In the case of identity verification the live sample features must be compared with the template associated with the claimed identity. The distance between these two points in the feature space will have to be compared with a threshold of acceptability. This threshold is set empirically to produce acceptable performance.

In the case of person identification the live sample will have to be compared with all the stored templates and a range of distances will be measured. The closest match with the smallest distance will then be chosen as the identity.

Various architectures have been used for performing such classifications. There is usually a training phase where the classifier is given valid feature vectors and their associated identity tokens. The success of the operational phase depends on the quality of this training phase.

3.4 Audio-Visual Fusion

Recognition/verification based on any one of these modalities alone may not be very robust whilst combining information from a number of different biometric modalities may well provide higher and more consistent performance levels. In addition to this, any one modality may not be acceptable by a particular user group or in a particular situation or instance. By combining modalities, greater robustness can be obtained while providing a measure of adaptability to given circumstances.

Several approaches can be adopted for combining the different modalities [23-27]. The two main approaches are called feature fusion and decision fusion; also called early and late fusion respectively. The term layered biometric is also used to describe forms of late or decision fusion. Each layer is one biometric modality and these can be combined to alter the performance parameters of the overall system (FRR, FAR).

A simple approach to decision fusion will be to treat the two modalities independently. For example, in an access control application, voice verification can be performed and if successful face verification can follow. If the latter is also successful then access can be granted. In such a sequential layer arrangement the latter layers will only be invoked if the earlier verification layers are successful. In this way the combined FAR of the system is the product of the FAR of each of the layers [27].

Alternatively, both biometric technologies can be invoked, possibly concurrently, in a parallel layered system. The system can be arranged so that if the any of the modalities produce an acceptance then the user is accepted and the other layers need not be invoked.

14


In this way the FRR is reduced to be the product of the FRR of all the layers [27]. It is also possible to have a logical operation performed at the final stage to combine the decisions at the parallel layers. In a layered approach several modalities of biometric information may be integrated.

AudioFeatures

Audio Models

DecisionLogic

FUSION

VideoFeatures

Video Models

DecisionLogic

DECISION

LOGIC

IDENTITY

Figure 2 An example of a decision fusion system.

AudioFeatures

FUSION

VideoFeatures

Audio / Video Models

DECISION

LOGIC

IDENTITY

Figure 3 An example of feature fusion system.

15


A more sophisticated version of decision fusion will hold information about the performance of individual classifiers; their strengths and weaknesses in identifying/verifying particular individuals or under particular circumstances. When it comes to combining the decisions from the different classifiers these additional performance information is combined in an optimal way to give appropriate weighting to the different biometric modalities. Figure 2 shows an example of this multimodal biometric configuration.

Alternatively in feature fusion the feature vectors obtained from the live samples are used together to train a combined classifier. This has the advantage that all the feature information is present at the classification stage; the disadvantage is that the classification stage becomes very sensitive to training data. Figure 3 shows an example of a feature fusion system.

The issue of efficient and effective combination of biometric modalities is still outstanding and attracts significant research attention.

4 Survey of Commercial Systems

A survey of commercial biometric products was conducted. A comparative table of products is provided as an appendix. Only a small number of companies are currently offering multimodal products. It has also proved very difficult to obtain performance information. This is to a large extent because the actual performance depends on the nature and size of the application environment. As these technologies are new and emerging there is not a large enough established body of experience and practice to form reliable performance indicators.

The market seems to include a relatively small number of primary technology developers and an increasing number of resellers and alliances between developers. The survey includes information from some primary developers and some resellers who provide support for multimodal use of audio-visual biometrics. Additional information regarding costs, system performance and deployment issues will be provided on the report’s web pages [28].

From the information available it appears that current face and voice verification devices are best suited to highly constrained environments where background sound and illumination variations can be minimised. Truly multimodal technologies that may overcome the current accuracy limitations of face and voice recognition systems alone appear to be still in their infancy. However much effort and interest seems to exist in the improvement of these biometric technologies because of their potential for continuous mode of operation and user acceptability.

16


The fact that audio and video capture devices (cameras and microphones) are increasingly bundled with personal computers should give an added cost advantage to face and voice recognition systems.

5 Applicability in smart-card technologies

A complementary authentication technology is the carrying of tokens such as magnetic strip and smart cards. Multipurpose, and possibly contact-less, smart cards that may be used for a number of different applications are likely to increase in popularity [29-30]. Smart card technologies allow a certain amount of processing to take place within the card.

Possible combinations of smart card and biometric technologies are likely to be a keen focus for research and development. It would be very desirable to be able to hold biometric templates as well as authentication routines within the smart card. In this way the biometric template can be kept within the card with no need for it ever to be communicated, thereby offering a higher level of security.

6 Potential Applications in Higher Education

Although biometric technologies are still in an early stage of development it is possible to envisage a number of key application areas where they may be of benefit in the higher education sector. Here some potential application areas are outlined.

6.1 Facilities Access Control Systems

Biometric technologies may provide added robustness in access control to high security facilities within higher education establishments. These may be hazardous research environments or research facilities that may be subject to attack by extremist groups. Such facilities are typically accessible by only a few staff members who may be happy to go through an elaborate enrolment and access process and are co-operative during the access authorisation phase.

A token-based system may suffer from loss of security through the theft or forgery of a token while this will not be the case for a biometric system. An audio-visual biometric system will in addition provide an audio-visual record that can act as an audit trail. Similar biometric systems may be used for regular access control to less secure facilities as required.

17


6.2 Workstation and Network Access

As the unit price for biometric devices continues to fall it is possible to employ these to replace the current pin numbers used for workstation and network access. These devices are likely to become a standard computer peripheral, built into future workstations.

6.3 Attendance Monitoring

The non-intrusive nature of some biometric systems may lend themselves to monitoring student attendance at lecture or laboratory sessions. As attendance may be compulsory for some sessions, an automatic system for monitoring attendance may be very desirable. Such a system may be installed at an entry port to the lecture or laboratory room or at workstations where the students will log in before the commencement of the session.

6.4 Monitored Student Assessments

Another possible application may be the authentication of student assessments. To combat the problem of plagiarism it is desirable to ensure that student assessments and tests that may have to be performed using a workstation are automatically monitored. This is to ensure that the student is completing the assessment himself without aid from other parties. An audio-visual biometric system, perhaps combined with the use of a smart card system, may be used for continuous monitoring of the student’s presence during the entire assessment or test.

6.5 Security Applications

A biometric system in its identification mode may be deployed to monitor surveillance cameras and/or the telephone system within the campus to identify known specific individuals who may have been excluded from parts or all of the facilities on campus. These could be known debtors, troublemakers, shoplifters etc. In this mode the system will have been supplied with template information for specific individuals and will continuously search for a match with the faces and voices that it detects.

A system similar to this has been reported to be in use for detecting known criminals by a local authority in East London (see Appendix B).

6.6 Payment Systems

To monitor the use of services and facilities offered free or at subsidised rates to specific user groups. An example of this would be the subsidised (VAT-free) use of university catering facilities which is only provided for registered students.

18


7 Standardisation

7.1 Biometric Application Programmers Interfaces

In recent years, a lot of effort has been put in the creation of standards for the interfacing and integration of biometric technologies in the form of Application Programmers Interfaces (APIs). The task of the API is to provide a generic interface between a software application that wants to use a biometric technology and the technology itself. The two most widely adopted APIs are the HA-API (SAFLINK Corporation) and the BAAPI (TrueTouch Technologies Inc.)

7.1.1 HAAPI – Human Authentication Application Programmers Interface

The Human Authentication API was first introduced at the Tenth US Biometric Consortium meeting in December 1997. The specification was originally developed by the National Registry Inc. under contract to an agency of the US Department of Defence. The API was then placed in the public domain in hopes that the adoption of a generalised biometric API would lead to the interchangeability of biometric technologies and encourage the widespread distribution of biometrics in general.

The HA-API provides a generalised interface between biometric applications and the biometric technologies themselves, thus eliminating the need for biometric application programmers to embed device specific code in their applications and, as a result, restrict themselves to a particular biometric type and vendor. If biometric technology vendors provide products that conform to the API specification and application developers also write their programs to the same specification then the problems of integration, substitution and future addition of multiple biometrics are greatly simplified. In addition, the API also allows biometric software developers to write programs without even selecting the particular biometric products that they are going to use.

The HA-API attempts to hide, to the degree possible, the unique features of individual biometric types and products by providing a toolbox of biometric functions, which is accessed via a standard interface. The HA-API supports the deployment of multiple and layered biometrics and both local and server based verification. Currently, HA-API is defined for the Microsoft Win-32 environment, with plans to expand it to other environments such as UNIX.

The HA-API specification has 11 function calls in 3 categories, as follows:

a) Biometric Technology Functions include the facility to list the biometric technologies installed and available on a system. There is also the facility to initialise or deactivate a particular biometric technology.

19


b) Biometric Authentication Functions include facilities to capture raw biometrics and perform enrolment and identity classification. In addition functions for performing continuous biometric monitoring are included.

c) Biometric Utility Functions include the interface for the application developer to get and set parameters specific to a biometric device and de-allocates memory used by a previous biometric function.

The HA-API specification and runtime software is in the public domain and can be downloaded free of charge [31].

7.1.2 BAAPI – Biometric Authentication Application Programmers Interface

TrueTouch Technologies is the developer of the Biometric Authentication API, which shares the same general philosophy with the HA-API.

TrueTouch technologies has formed agreements with a variety of software and hardware manufacturers and has created a Biometric Hardware Abstraction Layer by writing a set of DLLs that hide the actual device-dependant calls from the developer. As a result, a few lines of code allow the developer to call the BAAPI DLL and use a biometric device. Since the device dependant code is separate from the application, the developer need not even specify a device during the design of the application. When the BAAPI DLL loads it will determine what devices are available on the client system and validate accordingly.

The BAAPI SDK is available for sale and is licensed on a “per developer” basis. The cost is currently $200 per developer seat.

8 Organisations

A number of organisations have emerged to provide a focal point for exchange of information and co-ordinating of industrial, research and standardisation activities in the field of biometrics.

The Association for Biometrics is a UK based non-profit making organisation which aims to promote the awareness and development of biometric related technologies. It provides an international forum for research and development, system design and integration, application development, market development and other issues [32].

20


The Biometric Consortium serves as the US Government's focal point for research, development, test, evaluation, and application of biometric-based personal identification/verification technology [33].The International Biometric Industry Association (IBIA) is a trade association to advance and support the biometric industry. IBIA is controlled by biometric developers, manufacturers and integrators and provides information on all the different kinds of biometric technologies and the biometric industry as a whole [34].

The Commercial Biometric Developer Consortium (CBDC) has been established as a co-operative industry group designed to define and advance the common goals of the biometric vendor/integrator community. The CBDC is dedicated to advancing end-user confidence in biometric solutions by certifying biometric products, developing product guides, participating in industry working groups and publishing educational white papers [35].

9 Conclusions & Recommendations

Biometric technologies are potentially of significance in a range of security, access control and monitoring applications. The technologies are still new and rapidly evolving. It is clear that a number of biometric modalities working together can result in increased performance, reliability and ease of use. There is therefore considerable interest in developing multimodal and layered systems.

A number of on-line and networked applications can benefit from the use of continuous monitoring. In these applications audio-visual biometrics will be in a strong position. Also the non-intrusive nature of these biometrics gives them an advantage from a user-acceptance and ease of use point of view.

At this stage it is difficult to get performance figures and cost estimates. While laboratory test results are available from some vendors, actual performance depends on the application setting and environment and it is understandable that vendors may be reluctant to use lab figures for marketing purposes.

A number of possible applications can be envisaged within a higher-education setting. These include automated assessments, attendance monitoring and facility access control.

The interaction between different biometrically capable systems is an important issue. Protocols and standards may have to be developed for such interactions. Artificial intelligence techniques may need to be explored to manage the interaction between systems that are biometrically capable.

These technologies are potentially of great significance to the higher education sector and while their immediate deployment may not yet be realistic, they are going through a

21


phase of rapid development. It is recommended that the higher education sector should closely monitor developments in this sector.

The present report has focused only on audio-visual biometrics. There is a case for investigating in more depth the range of other biometric technologies available and their potential applications in the higher education sector. Additionally there is a case for conducting pilot projects to test the performance of some of the existing and soon-to-emerge technologies within a realistic higher-education setting.

10 Acknowledgements

This report draws upon the research work conducted with my co-workers, Dr John Mason of the University of Wales, Swansea and Dr Claude Chibelushi of the University of Staffordshire. Mr Stavros Paschalakis is gratefully acknowledged for his work on the survey of commercial biometric vendors included here. The help of Mr Peter Hawkes of the Association for Biometrics for suggesting various pointers for further information is gratefully acknowledged, as are many useful discussions on biometric technologies with Prof. Michael Fairhurst at the University of Kent.

11 References & Resources

[1] J Ashbourn, “The Biometric White Paper”, www.jsoft.freeserve.co.uk/whtpaper.htm, 1999

[2] Association for Biometrics and International Computer Security Association, “Glossary of Biometric Terms”, http://www.afb.org.uk/public/glossuk1.html, 1998.

[3] S G Davies, “Touching Big Brother – How biometric technology will fuse flesh and machine”, Information Technology and People, Vol. 7, No 4, 1994.

[4] B Millar, “Vital Signs of Identity”, IEEE Spectrum, pp 22-30, February 1994.

[5] C Jennings, “Biometrics – When the Person is the Key”, Sensor Review, Vol. 12, No. 3, pp 9-11, 1992.

[6] W Shen and R Khanna, “Scanning the Special Issue on Automated Biometrics”, Proceedings of IEEE, pp 1343-46, September 1997.

[7] J D Woodward, “Biometrics: Privacy’s Foe or Privacy’s Friend?”, Proceedings of the IEEE, Vol. 85, No. 9, September 1997.

[8] European Union, “Directive on Data Protection”, Official Journal of the European Communities, No L. 281 p. 31, 23 November 1995. http://www2.echo.lu/legal/en/dataprot/directiv/directiv.html

22


[9] P Agre and M Rotenberg, eds., Technology and Privacy: The New Landscape, MIT Press, 1997.

[10] Privacy International www.privacy.org

[11] Electronic Privacy Information Centre www.epic.org

[12] The Biotest Project, National Physical Laboratory, Teddington, Middlesex, http://www.npl.co.uk/npl/sections/this/biotest/index.html, 1999.

[13] C C Chibelushi, F Deravi, J S D Mason, "Survey of Audio-Visual Speech Databases", Speech and Image Processing Research Group, Department of Electrical and Electronic Engineering, University of Wales Swansea, http://www-ee.swan.ac.uk/SIPL/david/survey.html, 1996.

[14] K Messer, J Matas, J Kittler, J Luettin, G Maitre, “XM2VTS: The Extended M2VTS Database”, Proceedings 2nd Conference on Audio and Video-based Biometric Person Authentication AVBPA’99, Springer Verlag, New York, http://www.ee.surrey.ac.uk/Research/VSSP/xm2vtsdb, 1999.

[15] C C Chibelushi, S Gandon, J S D Mason, F Deravi, R D Johnston, “Design Issues for a Digital Audio-Visual Integrated Database”, IEE Colloquium on Integrated Audio-Visual Processing for Recognition, Synthesis and Communication(London - UK), Digest No: 1996/213, pp. 7/1 - 7/7, 1996.

[16] C C Chibelushi, F Deravi, J S D Mason, “A Review of Speech-Based Bimodal Recognition – Part 1: Foundations for Audio-Visual Fusion by Machine” Submitted for Publication in IEEE Transactions on Multimedia, 1999.

[17] C C Chibelushi, F Deravi, J S D Mason, “A Review of Speech-Based Bimodal Recognition – Part 2: Techniques, Performance, and Challenges” Submitted for Publication in IEEE Transactions on Multimedia, 1999.

[18] J P Campell, “Speaker Recognition: A Tutorial”, Proceedings of the IEEE, Vol 85, No 9, pp 1437-1462, September 1997.

[19] H Wechsler, et al (Eds.) “Face Recognition From Theory to Applications”, NATO ASI Series. SERS. F, Springer-Verlag, Berlin/Heidelberg, 1998.

[20] A Samal, and P A Iyengar, “Automatic recognition and analysis of human faces and facial expressions: a survey”, Pattern Recognition, 25, 65-77, 1992.

[21] D Valentin, H Abdi, and G W Cottrell, “Connectionist models of face processing: A survey”. Pattern Recognition, 27, 1209, 1994.

23


[22] R B Starkey, “The Human Face – A Unique Pattern?” Sensor Review, Vol. 12, No. 3, pp 16-18, 1992.

[23] C C Chibelushi, J S D Mason and F Deravi, “Audio-Visual Person Recognition: An Evaluation of Data Fusion Strategies”, Proceedings of the European Conference on Security, IEE, London, pp 26-30, 28-30 April 1997.

[24] P K Varshney, “Multisensor Data Fusion”, Electronics and Communications Engineering Journal, IEE, pp 245-253, December 1997.

[25] C C Chibelushi, J S D Mason and F Deravi, “Feature-level Data Fusion for Bimodal Person Recognition”, Sixth International Conference on Image Processing and its Applications, IEE, Trinity College, Dublin, Ireland, , pp 339-403, 14-17 July, 1997.

[26] C C. Chibelushi, F Deravi, J S D Mason, “Adaptive Classifier Integration for Robust Pattern Recognition”, Accepted for Publication, IEEE Transactions on Systems, Man and Cybernetics – Part B: Cybernetics, to appear December 1999.

[27] M Willems and P Forret, “Layered Biometric Verification”, Keyware Technologies, http://www.keyware.be, 1997

[28] F Deravi, “Audio-Visual Person Recognition for Security and Access Control”, JTAP project JTAP-552, local web site, URL: http://eleceng.ukc.ac.uk/~fd4/jtap.html, 1998-1999.

[29] L Burbridge, Experience with the use of a multi-purpose smart card, JTAP Report 019, JISC, March 1998.

[30] J R Parks, “Automated Personal Identification Methods for Use with Smart Cards”, Chapter 7 in Integrated Circuit Cards, Tags and Tokens edited by P Hawkes, D Davies and W Price, BSP, London, ISBN 0-632-01935-2, 1990

[31] Human Authentication API, http://www.saflink.com/h_downloads.html

[32] The Association for Biometrics, http://www.afb.org.uk

[33] The Biometric Consortium, http://www.biometrics.org

[34] International Biometric Industry Association (IBIA), http://www.ibia.org

[35] Commercial Biometric Developer Consortium, http://www.icsa.net/services/consortia/cbdc/cbdc_a.shtml

24


Appendix A – Table of Commercial Systems

Name Biometric URL

Face Voice Fingerprint Hand Iris/Retina

Handwriting

Advanced Biometrics Inc.

1 www.adv-bio.net

AEA Technology www.aeat.co.uk

American Biometric Company

www.abio.com

AuthenTec www.authentec.com

Betac TRS www.betac.com/trs

BioMet Partners Inc. 2 www.biomet.ch

Biometric Access Corporation

www.biometricaccess.com

Biometric Identification Inc.

www.biometricID.com

Biometric Security Corp.

www.fingerprintid.com

BioNetrix 3 www.biometricsciences.com

British Telecommunications

3 www.bt.co.uk

Cambridge Neurodynamics

www.neurodynamics.com

CHECKMATE None Available

Cogent Systems Inc. 4 www.cogentsystems.com

Communication Intelligence Corporation

None Available

Cyber-SIGN www.cybersign.com

Dermalog 5 www.dermalog.de

Dermo Trade Corporation

www.dermotrade.hu

25


Digital Persona www.digitalpersona.com

Eyedentify 6 None Available

FINGERSEC www.fingersec.com

Heimann Biometric Systems GmbH

www.hbs-jena.com

I/O Software Inc. www.iosoftware.com

Ideamation Inc. users.ids.net/~mikedn/idea/

Identicator Technology www.identicator.com

Identification Technologies International

www.iti-1on1.com

Identix www.identix.com

Integrated Visions www.integratedvisions.com

INTELITRAK www.intelitrak.com

INTELNET Inc. www.intelgate.com

IriScan www.iriscan.com

ITT Industries / Buytel Ltd.

www.speakerkey.com

Jasper Consulting Inc. www.jasperinc.com

Keyware Technologies www.keywareusa.com

KIA spg.kis.co.kr/Eng/index.htm

LEX Solutions Inc. www.lexsolutions.com

Miros www.miros.com

Mytec www.mytec.com

NEC www.nectech.com/afis

Neurotechnologija Ltd. www.neurotechnologija.com

26


NUANCE COMMUNICATIONS

www.nuance.com

Pantech pantech.co.kr

PenOp www.penop.com

Plettac Electronics www.plettac-electronics.com

PRINTRAK INTERNATIONAL Inc.

www.printrakinternational.com

PrintScan International Inc

www.printscan.com

Quintet www.quintetusa.com

Recognition Systems Inc.

5 www.recogsys.com

SAC Technologies Inc. www.sacman.com

SAFLINK www.saflink.com

SAGEM MORPHO Inc.

www.morpho.com

Sensar www.sensar.com

SENSE Technologies Inc.

www.senseme.com

SENSORY www.sensoryinc.com

SONETECH www.sonetechcorp.com

Sony www.world.sony.com

Startek www.startek.com.tw

TMS Inc. users.ids.net/~tms

T-NETIX www.t-netix.com

TrueTouch Technologies

www.truetouch.com

TRW www.trw.com

TSSI www.tssi.co.uk

27


Ultra-Scan www.Ultra-Scan.com

Unisys www.unisys.com

Ventura Identification Systems Inc.

www.letr.com/vis/index.html

Veridicom www.veridicom.com

VERITEL www.veritelcorp.com

VeriVoice www.verivoice.com

Viisage Technologies www.viisage.com

Visionics www.faceit.com

Vitrix www.vitrix.com

Who? Vision www.whovision.com

LEGEND

1 Veins in the hand2 Finger geometry3 Iris4 Palmprint5 Hand geometry6 Retina

28


Appendix B - Survey of Companies Providing Audio-Visual Biometric Solutions

B.1 T-NETIX - Visionics

Visionics Corporation and T-NETIX, Inc. offer their respective face recognition and speaker verification technologies as an integrated, turnkey solution. In particular, the companies will combine Visionics’ FaceIt technology with T-NETIX’s SpeakEZ Voice Print(sm) Speaker Verification, “providing customers with access via a single source to the two least intrusive, lowest cost biometric solutions available today.”

In lab tests the speaker recognition system has shown equal error rates of 1.7% in multimedia environments. In contrast, the company reports equal error rates of 1.7% and 6.5% respectively for landline and mobile telephony. A single 200 MHz IBM compatible PC running Microsoft Windows NT is capable of serving a large site with several thousand users performing roughly seven thousand verifications per day. Benchmarking on a 300 MHz IBM compatible PC the company reports an enrolment processing time of roughly 3-5 seconds and a verification time of roughly 0.15 seconds.

Visionics’ FaceIT product range offers software modules for head finding, tracking and face extraction irrespective of backgrounds. In addition there are libraries of enrolment, verification and identification functions. Turnkey database search engines can be integrated for identity detection. An automatic database can be built of time stamped facial records from live or recorded video. The technology works with still pictures as well as video feeds. Modules are available for continuous monitoring in workstation and application access where as soon as the authorised face moves from the field of view the application locks until the authorised face returns. The templates used are referred to as face prints. They are claimed to be resistant to changes in lighting, skin tone, eyeglasses, facial expression, hair and robust to pose variation, up to 35 degrees from all directions.

The size of the face print template is 600 to 3500 bytes depending on the application. The company claims a false accept and a false reject rate of less than 1%. A speed of 60000 matches per minute on a Pentium 300 machines running Microsoft Access is quoted.

The FaceIT technology has been used in a CCTV control room application as part of the anti-crime ‘Mandrake’ system recently installed in the borough of Newham, East London. This system automatically scans the faces of people passing 144 CCTV cameras located around Newham searching for matches to a watch list, in a video library of known criminals, stored in a local police station. There is no information available as to the performance of the system. The price of FaceIT NT for a single workstation is quoted as less than 100 dollars ($99.95).

29


No information was available at the time of writing this report on the joint performance characteristics of the SpeakEZ and FaceIT technologies.

http://www.faceit.comhttp://www.t-netix.com

B.2 Keyware

Keyware’s Layered Biometric Verification (LBV) server integrates multiple levels of biometric verification into a single solution. Some of the biometrics supported by the LBV server are face, voice, fingerprint, iris and retina recognition.

The LBV server allows an organisation to choose a single biometric or any combination of biometrics and deploy them in a multiple or layered configuration. In the multiple configuration there is no communication between the various verification procedures, whereas with the layered scheme the different biometric modules communicate with each other. The result of this communication process is that if one of the biometrics is slightly altered, the LBV server will require a higher confidence level from the other biometrics and, as a result, the overall security is enhanced.

The LBV server allows the administrator to optimise the organisation’s security policy for each customer and location so as to achieve the highest security. The LBV server also maintains a central database of biometric templates to provide for easy administration and access privileges from many different clients. Keyware also uses SmartCard technologies, which allow users to hold their own biometric data rather than have it stored in a central database.

The LBV server supports a full range of development environments including C/C++, Visual Basic, ActiveX and JAVA. Available clients are:

Basic Client: Integrated client application for biometric authentication purposes.

NT Login: Secured NT login procedure.

Telephony Client: Client with user definable scripts for biometric authentication on Dialogic or NMS telephony boards.

Physical Access: Client with RS232 communication to control door access.

30


The entry level system requirements for the LBV server are : Pentium 133 MHz or higher, Windows 95 stand alone, Windows NT Workstation or Server 4.0, ODBC compliant database, 32 MB RAM, 24 MB disk space, Network interface card.

The LBV server currently supports Keyware’s API (Application Programmer’s Interface) for the integration of different biometric modules into complete systems. However, Keyware is also involved in the BIOAPI, HA-API and UAS API groups and Keyware’s products will comply with these standards when they are published.

Keyware’s VoiceGuardian is a self contained voice verification product which can be integrated into the LBV server. FaceGuardian, a face recognition product which can also be integrated into the LBV server. In a partnership with STMicroelectronics, Keyware also develops integrated circuits implementing Layered Biometric Verification technology. The VoiceGuardian product is a password independent, phoneme-based voice verification system with a claimed EER of less than 1% for 2-3 seconds of speech.

http://www.keywareusa.comhttp://www.keyware.be

B.3 SAFLINK

SAFLINK provides identification and authentication products and services for enterprise-wide network security applications through its Secure Authentication Facility (SAF) family of multi-biometric products. Among the biometric technologies supported by SAFLIK are face, voice and fingerprint recognition. The individual biometric modules are provided by other vendors and integrated within the SAF framework.

SAFLINK’s SAF2000 Multi-Biometric Enterprise Security Suite is an integrated grouping of packaged biometric security applications. The entire suite is supported on Windows NT servers and utilises the HA-API standard for the interchangeability of biometric technologies. SAF2000 is network centric and allows a variety of client configurations to securely communicate with the SAFserver. The SAF architecture can concurrently support different networks employing a number of applications such as SAF/nt for Windows 95/98 and NT clients and SAF/iis for the Microsoft Internet Information Server.

Included with the SAF2000 suite are fingerprint, voice and face recognition biometric modules from other primary developers. Additional biometric modules can be purchased separately and added to the SAFserver and one or more client workstations. SAF2000 implements an administrator controlled trust model, i.e. the administrator can define

31


which users to trust, the credentials that are required to gain trust and when a trusted relationship is required.

http://www.saflink.com

B.4 Bionetrix BioNetrix is a personal identification software and services company, providing multimodal biometric identification systems and solutions. The core biometric technologies supplied by other vendors are integrated within the BioNetrix Authentication Suite, a suite of software products which facilitate the deployment and implementation of biometric technologies for authentication. The biometric technologies that are supported in the BioNetrix authentication suite are face, voice, fingerprint, iris and signature recognition. The BioNetrix Authentication Suite also allows for token based authentication (e.g. smartcards).

The BioNetrix Authentication Suite is made up of a number of different components which allow the effective management and operation of the system. The following is a brief description of each component.

BioServer: This is the core of the BioNetrix system which allows organisations to use any number of biometric devices in any combination. The BioServer also allows organisations to either centralise or distribute authentication nodes across an organisation. Some of the functions provided are : layered and/or multiple biometric authentication; remote enrolment and authentication; encrypted transport of biometric data and storage of users’ biometric profiles and templates.

BioClient: The BioClient provides the interface for users to be authenticated for access to applications and networks from any workstation in the BioServer domain. Some of the features provided include : layered and/or multiple biometric authentication; remote enrolment and authentication and encrypted transport of biometric data.

BioRemote Client: The BioRemote Client is a scaled down version of the BioClient, for use on a laptop or notebook. It is typically configured to support a single biometric.

BioRegistrar Station: The BioRegistrar Station allows the system administrator to manage user information. It is typically a workstation that has one of each of the biometric technologies used in the organisation. Some of the features provided are : support of multiple biometric devices; capture and organisation of biometric information; client workstation profiles and generation of user reports.

32


BioOperations Management Station: The BioOperations Management Station provides the controlling interface for the BioNetrix environment, enabling the operator to manage the devices used by BioClient, upgrade and manage software versions and generate status and security reports.

BioDomain Interface: The BioDomain application program interface (API) provides functions which allow the integration of the BioServer into applications.

BioWeb Interface: The BioWeb Interface facilitates biometric authentication over the Internet for users that access information via web browsers such as Netscape Navigator and Microsoft Internet Explorer. Currently, only voice and fingerprint technologies are supported in this module.

BioWorkstation is a separate product which offers biometric authentication for access to stand-alone Windows 95 or NT workstations. User authentication is performed using single or multiple biometrics such as face, voice, fingerprint and signature. The BioWorkstation is a self-contained product which provides all operational and administrative functions.

The following is a list of some of the environments and applications which BioNetrix products can be integrated with :

Workstations Windows 95/98/NT, UNIX (Sun, HP)Networks Novell, Windows NTDatabases Oracle, Sybase, MSSQLE-Mail Microsoft Exchange, Novell GroupWiseHuman Resources PeopleSoftFinance Oracle, PeopleSoftSecurity Single Sign-On, Digital Certificates

Although the individual biometric technologies that BioNetrix integrates into its systems are supplied by a number of different vendors, BioNetrix provides support for the systems that it integrates and allows for specific maintenance programs to be developed for each of its clients.

http://www.biometricsciences.com

33


B.5 Integrated Visions

Integrated Visions focuses on providing complete biometric security solutions to the healthcare industry for the protection of the confidentiality of computerised patient information.

The company’s Privacy Curtain family of products include biometric single sign-on security solutions to identify and authenticate individuals attempting to access confidential information. The authentication information can be biometrics or a combination of biometrics, user ID and passwords. Integrated Visions is also the developer of SAF-NT, a system which provides biometric log on to NT, tightly integrated with the NT operating system.

Integrated Vision’s biometric security solutions use a combination of fingerprint, face and voice recognition technologies which are provided by a number of biometric technology vendors. The individual technologies are integrated in the multimodal biometric system using the Human Authentication API (HA-API), which allows for the easy integration of any HA-API compliant device in the complete system and the easy upgrade of the system to new technology.

The biometric security systems provided by Integrated Visions also provide flexibility in choosing biometric modalities for particular sessions. The system speed depends heavily on the hardware being used and on the network speed, but for typical installations the training/enrolment phase to the system will take 5-10 minutes per user, while access into the system (recognition/authentication) typically takes less than 3 seconds. The company quotes an FAR of 0.0004% and an FRR of 0.566% for fingerprint verification. While no performance figures are available for voice and face recognition the company states that they are less accurate.

The server system that holds the authentication database is hosted on an NT system which integrates easily into existing networks. The database in use is the Microsoft SQL Server 6.5.

The overall authentication system is designed to be fault tolerant and scalable and can run on configurations as small as one server up to multiple clusters.

http://www.integartedvisions.com

34


B.6 TrueTouch

TrueTouch supports a very large and diverse collection of biometric systems available on the market, including fingerprint, face, voice, retina, iris, hand and palm recognition software and hardware. Like SAF it provides a framework for integrating biometric products from other vendors.

TrueTouch’s Biometric Software Security Suite or Bio Suite integrates with existing systems and applications, providing biometric security on stand-alone and networked PCs. The BioSuite is made up of the following components:

TrueBoot: Protects the boot sequence and denies unauthorised access to the PC.

TrueLock: TrueLock is a multi-purpose application which includes a screen saver that guards against intrusion. TrueLock also allows the administrator to define user profiles, giving rights to particular files and directories.

TrueEncrypt: TrueEncrypt facilitates the easy and quick encryption of files using popular encryption programs such as DES, PGP and RSA.

TrueClock: TrueClock is not a security product like the other components of the Bio Suite. TrueClock is a simple time-tracking application which offers the administrator the ability to monitor employee attendance.

TrueAdministrator: TrueAdministrator is the control centre of the Biometric Software Security Suite. The system administrator can easily set up the security features in the Bio Suite, create, edit and delete user profiles and select biometric devices.

The S.P.I.K.E. (Suspect and Prisoner Identification Key Evaluation System) is a complete turnkey system targeted at law enforcement agencies. S.P.I.K.E. records and verifies a suspect’s identity by combining face, voice and AFIS compatible fingerprint recognition.

http://www.truetouch.com

B.7 Miros

Miros does not produce multimodal biometric systems, but is included here because its face recognition products are used as components by the companies which provide complete multimodal biometric systems.

35


Miros is the developer of the TrueFace family of face recognition products. The following is a list of some of the biometric solutions that the TrueFace family offers:

TrueFace ID: Identifies a person’s face in a database of faces from either a surveillance video or computer file.

TrueFace Web: Browser Client-Server software that allows secure access to web pages using your face.

TrueFace Network: Network Client-Server software that allows secure access to server data using your face.

TrueFace PC: This product secures the screen saver of desktop PCs running Windows 95 or 98 with faces of up to five users.

TrueFace NT: This product secures the logon and screen saver of desktop PCs running Windows NT with the faces of up to five users.

The TrueFace products recognise the different faces by using neural networks trained on features extracted from the entire face image rather than distances and angles of the eyes, nose or mouth. Also, TrueFace uses stereoscopic data (i.e. two views of each face) for recognition, thus reducing the probability that the system may be fooled by a photograph.

The company quotes that TrueFace can verify one live face image per second and identify a best match in a database search at a rate of 500 frames per second on a Pentium 200 MHz PC.

http://www.miros.com

36


Appendix C – Contact Address for Audio-Visual Biometric System Providers

T-NETIX

URL : www.t-netix.come-mail : [email protected] : +1 303-790-9111Fax : +1 303-790-9540Address : T-NETIX Inc.

67 Inverness Drive EastEnglewood, CO 80112USA

Visionics

URL : www.faceit.come-mail : [email protected] : +1 201-332-9213Fax : +1 201-332-9313Address : Visionics Corporation

1 Exchange PlaceSuite 810Jersey CityNJ 07302USA

Keyware (headquarters in both the United States and Belgium)

URL : www.keywareusa.come-mail : [email protected] (USA) [email protected] (Belgium)Tel : +1 781-933-1311 (USA) +32 2-721-4574 (Belgium)Fax : +1 781-933-1554 (USA) +32 2-721-6949 (Belgium)Address : Keyware Technologies Inc. Keyware Technologies, NV

500 West Cummings park 28-30 ExcelsiorlaanSuite 3600 ZaventemWoburm, MA 01801 Belgium B-1930USA

37


SAFLINK

URL : www.saflink.come-mail : [email protected] : +1 813-636-0099Fax : +1 813-636-0422Address : The National Registry Inc.

SAFLINK Corporation2502 Rocky Point DriveSuite 100Tampa, FL 33607USA

Bionetrix

URL : www.biometricsciences.come-mail : [email protected] : +1 703-734-9200Fax : -Address : BioNetrix

8150 Leesburg Pike, Suite 1230Vienna, Virginia 22182USA

Integrated Visions

URL : www.integratedvisions.come-mail : [email protected] : +1 888 302 7200 ext. 7393Fax : -Address : -

TrueTouch

URL : www.truetouch.come-mail : -Tel : +1 407-877-6642Fax : +1 407-877-0989Address : TrueTouch Technologies, Inc.

38


123 S. Woodland St.Winter Garden, Florida 34787USA

Miros

URL : www.miros.come-mail : [email protected] : +1 781-235-0330Fax : +1 781-235-0720Address : Miros, Inc.

572 Washington St.Suite 18Wellesley, MA 02482USA

39

Documents

Audio-Visual Person Recognition for Security and · Web viewAudio-Visual Person Recognition for Security and Access Control Farzin Deravi University of Kent at Canterbury Table of