Product

Machine presentation

Tactim is proud to present the contribution of dermo-science to a new secure personal electronic identification system. In this section you will get a quick presentation of this system.

The human skin	The pacinian corpuscule

Dermo-Science is relatively new, but its researches have shown that the skin of the human hand is densely populated by different kinds of sensory receptors. One set of these, the Pacinian Corpuscules in the middle-fingertip, has the ability to help produce a uniquely personal sound wave feature specific to one individual.

Technical point of view.

When the person asked to provide a signature rubs his or her middle-fingertip on a specific sensor, the resulting electronic wave-signal can be recorded as a small sound file. The nerve impulses transmitting the original neuro-physiological data produced by the natural touching process to the brain provide a second opportunity for a unique signature. This discovery has been noted and detailed in a patented method (US and European), which enables the recording of this sound signal to be used for authentication purposes.

The following diagram explains the process used for authentication:

Recording

A specific sensor is used to record the sound produced by the rubbing of the middle-fingertip of a person on a wooden plate. The analysis of the signal shows that frequencies greater than 2500 Hz are not pertinent for the system. So we integrate in the sensor a Salen-Key structured filter to eliminate the unwanted frequencies. A database of 20 samples per person is made to obtain a good accuracy during the recognition step.

Extraction

Autoregressive models parameterized to predict the signal in order to minimize the total variance of the signal are used. The result is a set of coefficients (17 seems to be a good compromise between performances and computing time). The coefficients of the Fourier transform (FT) are also computed. Wavelet transform have also been investigated.

Learning

The coefficients of the AR models and of the Fourier transform are simply stored. More complicated, the 20 sets of AR models coefficients are used to produce one hidden Markov model (HMM) with a classical learning algorithm (Baum-Welch). The same operation is done with the sets of the FT coefficients.

Recognition

We investigate many techniques going from Euclidian distance to hidden Markov models (HMM). In a classical way, the recognition consists in the comparison between the signature we want to test with all others signatures contained in the database. The unknown signature is recognized to belong to the person owning the data the closest to a known sample. The final version of the classifier will use data fusion to take account from all of the parameters extracted from signatures.

Practical point of view

The current identifying system provides the main functionalities : recording, learning and recognition. The current software runs under the matlab environment. The sensor is connected to the computer. The software uses the integrated soundcard to get the signal. This one is displayed in the upper window, named ”acoustic signal”. It is then analysed and a graphical view of the signature is displayed in the bottom window. We can add this signature to the database, or test it to know if its owner is registered or not. A screenshot of the software is represented on the right.

The first prototype of the sensor is shown here. The principle of the sensor is registered to a US patent, number 2-482-773. Then the new system presented here, describing the global process from the use of the sensor to the processing of the acquired data is patented too (EP # 0775323). The last prototype built was analysed and tested in October, 2001. Since this date, new developments were made, increasing the accuracy of the recogniser.

Applications

Now, imagine you are in front of a cash dispenser. You want money, you have a credit card but no secret code. This one is in you, in your fingertip. You insert your credit card, ask for money and validate by rubbing your middle-fingertip on the sensor.

Another situation : feeling sick, you go to the doctor. This one authenticates and signs its prescription using this system. Miniaturised in a chip, it could be used as a password system to access to your mobile phone, or your computer or your car.

The researches are not yet finished on this subject. Two points are more interesting. The first one comes from neuro-physiology : this signal could reflect emotional state. In other words and in theory, we could know if somebody signs from its own will or feeling afraid. Another particular thing we note during the evaluation of the system : sometime people from the same family are not well discerned. Probably the signature has characteristics in common for people with a family link.

Credits

This quick presentation of our product was built with the help of Dr. Thierry Brouard, from the university of Tours (France).

Contact information
Computer Science Laboratory  
University of Tours  
64 av. Jean Portalis, 37000 Tours
France 
+(33)247361425