It is possible to execute a copy of each human voice according to characteristic changes of sounds. If the phone call is without extraneous noises (wind, operating technique) and from the same telephone set is possible to carry out recognition. The speech fragment of 6 seconds is distinguished with an accuracy of 90 – 95% (depends on the voice).
It has a meaning in several aspects. For example, there is no need for the operator to identify you on passport data while request for phone balance. As you call from your own phone, then to tell the required information which is noncritical for you it is admissible. If the swindler decides to copy your voice (a timbre, individual intonations) for carrying out fraud, he won’t manage to reproduce physical parameters of the speech. In such way the system of voice identification works great. Factually, when the bank’s call-center receives the call the person’s voice compares with known swindlers’ voices. It’s possible to find out your call that has been done from the other telephone number analyzing all communications. At the same time several tens of records, and long pieces of a conversation will be listened with the existing accuracy and will give only hundreds of options. For more information, see Move Up.