With so many distinctive protrusions, patterns and surfaces on the human body itself, it seems like the world doesn’t really need another way to tie a body part to an identity. However, voice-based product design fill a unique role in the universe of biometric security.

Each and every voice is unique because of the shape of a person’s vocal cavities, including the way and speed with which individuals move their mouths and tongues as they enunciate. Information for voice biometrics is measured via spectrogram, a quantitative measurement of different frequencies that are inherent to a voice (this contrasts with what many movies purport voice biometrics to look like, bouncing parabola-like oscilloscope waveforms).

To use a voice-based biometric system, all you have to do is speak a specific passphrase or in some cases say an extended sample, allowing the computer to identify your speech pattern as a whole. This can be done either in person or over the phone.

Voice recognition is especially useful because it allows users to authenticate their identities without being physically present. As mobile technology explodes in popularity, such operations could prove highly valuable for logging into one’s private accounts for work, school or banking.

However, over-the-phone biometric verification does have a few problems. People’s voices can be altered due to a heavily due to illness, mood and reception quality. In addition, a thief needs only to record a sample of someone’s spoken verification in order to break through a voice-based security system design on a mobile phone.

Researchers at the University of Colorado are gearing up to fix these problems, however. The task is accomplished by not transmitting voice data at all. Instead, the secure institution, such as a bank, would send two encrypted versions of each password or phrase to the mobile phone. One is the correct user’s voice, the other is a dummy data set spoken by an entirely different person. Software on the phone itself then compares the user’s voice against the two samples. Plotting the highs, lows and cadence, the client-side software partakes in a multiple-choice-like quiz to discern the samples.

The beauty of this new system is that a user’s recording of the sample words are never actually transmitted over the network, thereby preserving their privacy and limiting their availability to be hacked over the airwaves.

Though not as precise as in-person biometrics, voice biometrics will prove to be enormously useful in the coming years as an increasing amount of information is sent and received via mobile device.

To learn more about Pivot and our capabilities in biometric product design, development, and manufacturing, contact us at 877-206-5001.