Speaker recognition

The speaker authentication makes use of the fact that a man's voice is as unique and distinctive as a fingerprint. This allows the access to protected areas are significantly more secure and more convenient than, for example, by a PIN or password entry.

Generally, one distinguishes between caller verification and identification. During verification, it is checked whether a caller is who he claims to be. To this end, the caller must first identify - so say for example its name or customer number. In a second step, the current voice profile is compared with a stored voice sample (1: 1 - Balance).

When identifying a voice sample from a large amount is recognized to voice profiles (1: n - Balance). The user need, therefore, to give not to recognize, however, he must first have been given a voice sample. The disadvantage of this method is that it requires a large computing power.

In modern authentication methods specific algorithms ensure that voice samples are neither rejected if they are from the same speaker still be accepted if they come from another speaker. For the first case, the systems are designed sufficiently tolerant, so they will still recognize a speaker, for example, where the vocal tract is altered by a cold.

In the second case, provide some randomly generated words that must repeat a spokesman, that the system, for example, can not be fooled by pre-recorded voices.

• Authentication technique
• Human voice