Financial-credit organization "Priorbank" today announced the launch of the service identity verification by voice "Voice biometrics". The project aims to reduce the time of customer service. Now when you call the call centre on identification of the customer takes up to 40 seconds: during this time the operator asks about five questions. In the same month the service handles up to 40 thousand calls, two-thirds of them relate to personal data and require identity verification.

Read more...

Financial-credit organization "Priorbank" today announced the launch of the service identity verification by voice "Voice biometrics". The project aims to reduce the time of customer service. Now when you call the call centre on identification of the customer takes up to 40 seconds: during this time the operator asks about five questions. In the same month the service handles up to 40 thousand calls, two-thirds of them relate to personal data and require identity verification.

When you use voice biometrics (VoiceKey) the time of identity verification by telephone will be reduced by approximately 4-5 times, will also decrease the likelihood of illegal access to personal information of customers. System efficiency is 98-99%. As noted by the representative of the company-the resident of "SKOLKOVO" "speech technology Center" Vladislav Filippov, the voice is the same unique identifier, like a fingerprint or iris is.

the Algorithm is as follows: the customer calls the contact center of the Bank and asks the operator a question. The Bank's employee determines whether the client identification and, if necessary, finds it in the database. At the same time creates a voice model-the standard digital file, 74 taking into account the characteristics of the voice. In subsequent requests, the system again creates a model of the voice and compares it with the benchmark. The verification result is displayed CRM. To confirm the identity requires one answer to the question.

Embedded system loyal to the source of the voice: it successfully processes data received through digital and analogue communication line, coal and digital microphones. The noise reduction system is able to identify and analyze voice in the presence of extraneous sounds, but in some circumstances it will not work. Also takes into account the variable characteristics of the voice (changed due to lung disease or, for example, if the customer was drunk).

Still in the testing phase the system tried to cheat in various ways, attracted even parodist. However, the artist, say the developers, was unable to forge physiological features of the "original". Similarly the situation is with a computer-generated voice. The answer to the question, is it possible to apply for an advanced cheating system that uses a database of words spoken by the client, representatives of the developer made a joke. In addition, communication occurs with a contact centre operator who can smell a rat.