The recognition of persons by their biometric features is a rapidly growing field of research. Many systems arise using face, voice, fingerprint or eye features. All of them have different advantages and disadvantages. By combining some of those methods, recognition accuracy can be improved dramatically. BioID is a multimodal biometric system, i.e. it is designed to allow the usage of several biometric techniques. At present, BioID offers face recognition, voice recognition, and human lip movement recognition. Each of these techniques can be used solely, but by combining them, it represents a true multimodal system. This multimodal biometric system therefore not only yields highly accurate results but by including dynamic features (lip movement analysis) it is also more robust against frauds than systems using static features like the face or finger print.

Introduction

The identification of humans for e.g. financial transactions, access control, or computer access has almost always been conducted by using ID numbers, such as a PIN or a password. The main problem with those numbers is that, let aside the fact that they sometimes can be cracked quite easily, they can be stolen and used by an unauthorized person without detection.

Biometric identification systems use personal features of the user itself to check the identity. If, for example, biometric features stored on a chip card are stolen, they cannot be used, because the impostor’s biometric features do not match the features stored on the card. For this reason, the interest in biometric systems has risen very much lately.

There are many features of a person that can be useful for biometric feature extraction. The face, the voice, the eye, the movement, amongst many others, can be used to distinguish a human person from another. Many research activities have proven the principle usefulness of those features.

A single feature, however, sometimes fails to be exact enough for identification. Consider twins, for example. The face feature alone will not be able to distinguish them. If additional features, like voice or movement, are taken into account, the system can distinguish those individuals more accurately. Another disadvantage of using only one feature is the "readability" of the chosen biometric feature. For example, 5% of the human beings do not have fingerprints that can be recorded.

Therefore, HumanScan developed BioID, a multimodal system, currently using face, voice and lip movement detection for person identification. By using those three modalities, a much higher accuracy can be achieved. Even when one modality is somehow disturbed, e.g. in a noisy environment, the other two modalities still lead to an accurate decision. Nevertheless, BioID offers the full flexibility to choose any of the modalities in any combination: applications like a face recognition system, or a voice identifier, or a combined face/voice recognizer can therefore easily be created.