Poland\'s first system that allows the identification of a person based on sound and accentuation of spoken words has been developed by researchers at AGH in Kraków.
"The human voice is so unique that it can be used to identify any of us" - explained Dr. Jakub Gałka of AGH University of Science and Technology in Kraków. Therefore, scientists all over the world have been working for over a decade on biometric systems that allow recognition of a person only on the basis of voice.
These systems are based on the characteristics associated with the physics of speech, resulting from the voice frequency, the anatomy of the speech organs, such as the size and shape of the larynx. "It all determines the sound, which we call speech. These elements distinctive of every human can be measured, although not all of them are consciously discernible" – emphasised Dr. Jakub Gałka.
Other unique features are associated more with how we speak, what we say, and how we emphasize how fast we speak, how we choose words, how often we take a breath and pause between words. "These features are somewhat more controllable by the speaker, but are not able to model all of them at the same time while speaking" - noted the researcher.
Together with a team of scientists and engineers from AGH University of Science and Technology and the Kraków company Unico Software, Dr. Gałka developed the voice-based biometric identity verification technology - VoicePass. The user, who undergoes such verification, should give voice password. "The password is shared by all users of the system, because not the content of the password is the verification factor, but the biometric features hidden in the expression" - explained Dr. Gałka.
The password should contain at least a few words. The longer the statement, the more sound material can be analysed and the greater the effectiveness of verification. The password should also be natural to say, it should not be a tongue twister, such as "she sells sea-shells on the sea-shore".
According to Dr. Gałka, VoicePass can be useful in all call centres in banks, telecom operators, insurance companies, brokerage houses, or in any institution that requires verification of the identity of the caller.
This type of verification can also be used in smartphones. "Imagine a banking application, which requires identity confirmation prior to a transfer. Use of voice biometrics would be a much more convenient solution than typing a PIN or password. It can also be used on websites, where instead of typing a password you could say it. One of the elements of the project is also building an electronic door lock that can be opened by voice" - the researcher told PAP.
He assured that, contrary to common belief, this method to protect access to our money or data is also safe. "It is much easier to impersonate someone by stealing such Social Security number or mother\'s maiden name. Acquiring such data is much easier than imitating a person’s voice and biometric features" - noted Dr. Gałka.
In addition, voice identification is convenient, because it does not require any additional sensors. "All other biometric verification methods, such as the identification of the iris, require the installation of additional scanning device. Everyone carries a voice biometrics device in their pockets, it is a mobile phone" - said the researcher.
In recent years the effectiveness of identification methods greatly increased. Currently, they are imtroduced primarily in the United States, Canada, and in several European countries. In Poland, scientists are just beginning to test them. The system developed at AGH is the first Polish solution of this type with commercialisation potential.
As part of the project, the scientists have developed Surikate technology, used to identify speakers at various meetings and conferences. "The system recognizes who speaks during the meeting and when. Speaker’s name can be displayed on a screen. However, the more people the system has to recognize, the lower the effectiveness of identification will be" - described Dr. Gałka.
VoicePass system is now being tested in several institutions. It was created as a result of the work of the consortium of the AGH University of Science and Technology in Kraków and Unico Software as part of the Applied Research Programme of the National Research and Development Centre - project "Biometric Voice Verification and Identification". "I hope that will soon be available to customers" - said the scientist.
Ewelina Krajczyńska (PAP)
ekr/ agt/ mag/
tr. RL
