An HS Daily Wire conversation with Walter Hamilton of the International Biometrics Industry Association (IBIA)

challenges. But we know that it is a unique and very useful tool, and we welcome the opportunity to share our reasoning and our expertise with those responsible for implementing these programs.

DW: Large claims are made for, among others, voice authentication techniques. Would it not be logical — not to say expedient and cost-effective — to identify the best biometric technology and favor it in all security applications?

Hamilton: This is a complex issue, and one that’s raised very frequently. It is not practical to assume that one biometric technology is the best. The key is matching the right biometric technology to a specific application requirement. All biometric technologies have a number of things in common. One is that they utilize biological or behavioral human characteristics that have a high degree of uniqueness. These are unique and measurable characteristics that, in theory at least, are not exactly the same in any other person. Or, if they are, it is a statistically insignificant recurrence.

Another desirable characteristic for a good biometric technology would be the relative ease with which a user can present the biometric sample to a sensor device. The device, and its related software algorithms, will convert that raw biometric sample to a usable digital form that can be accurately matched against a reference sample. The reference sample will have been previously stored in a data base or smart card, or wherever.

Speaker verification satisfies these desirable biometric traits, and thus might be the technology of choice for a specific application. But it might not be. If there happened to be a lot of background noise when the speech sample was collected, speaker verification might not meet performance expectations in terms of matching accuracy. Not that certain types of microphones and software couldn’t filter out the ambient noise - but this poses an additional challenge to accuracy when using speaker recognition biometrics.

As in any input-output process in the computing world, the quality of input is directly proportional to the quality of output. Should you capture any kind of biometric sample that is distorted, garbled, or interfered with by noise or ambient light or some other interfering occurrence — such as dirt, moisture, dryness of the skin surface — the quality of that sample would be affected. And your ability to accurately match it against a reference sample would be correspondingly degraded.

DW: Let us assume, however, that the sampling is sound. Does one