Researchers have come up with a new algorithm for establishing speech identity more quickly.
Everyone’s voice
is different. But when it comes to computerized voice identification, the flash
of recognition often comes too late to satisfy business and government demands.
Shaving precious seconds from previous benchmarks, researchers at North Carolina State University in Raleigh
have come up with a new algorithm for establishing speech identity.
“Our work makes
the process more efficient, so the latency in recognition is shortened,” says Dr.
Robert Rodman, professor of Computer Science at N.C. State. Typical improvement
is between 8 and 15 percent faster recognition. “It might reduce 5- or 6-second
latency by 1 second,” Rodman explains, adding, “Every little bit counts. If
you’re a user, every second of latency has an effect.”
The research is described in a paper, “Joint Frame and
Gaussian Selection for Text Independent Speaker Verification,” that will be
presented at the International Conference on Acoustics, Speech, and Signal
Processing (ICASSP) in Dallas
from March 14-19.
“It’s highly
mathematical—we found a method for eliminating non-matches,” Rodman states. The
researchers hit upon a novel Gaussian selecton method that builds on previous
voice identification work, according to Rodman. “It moves us closer to making
this technology a practical, secure tool," he says.
Current computer
models that are used to compare acoustic profiles by evaluating whether a
speaker is who he says he is may take several seconds or more to process the
information, which is still too long for the technology to gain widespread
acceptance. "In order for this technology to gain traction among users, the
response time needs to improve without increasing the error rate," states
Rodman.
Speech
authentication is of particular interest in this age of heightened security and
mobile electronics. "Potential users of this technology include
government, financial, health care and telecommunications industries,"
Rodman says, "for applications ranging from preventing ID theft and fraud
to data protection."
Rodman has been
working in this field for five years. The latest research was funded, in
part, by the Centre for International Mobility. The research paper was
co-authored by N.C. State’s Rodman; Rahim Saeidi, Tomi Kinnunen and Pasi Franti
of the University of Joensuu in Finland; and Hamid Reza Sadegh Mohammadi of the
Iranian Academic Center for Education, Culture, and Research.
New way to drive customer service solutions?Posted on: 03-16-10 | By: Anil G.The old notion of using personal informtion to identify individuals that call their bank or credit card companies is totally outmoded. Too often this information can be compromised. Using the unique "voiceprint" rather than mother's maiden name seems far more useful and SECURE!
New way to drive customer service solutions?Posted on: 03-16-10 | By: Anil G.The old notion of using personal informtion to identify individuals that call their bank or credit card companies is totally outmoded. Too often this information can be compromised. Using the unique "voiceprint" rather than mother's maiden name seems far more useful and SECURE!
Telephone-based authenticationPosted on: 03-15-10 | By: Michael SOne question I have is, how sensitive does the mic have to be in order to pick up the voiceprint? Can this be done over the phone, or do you need a more high-fidelity sample of the voice to be authenticated?
A user comment on this articlePosted on: 03-15-10 | By: JonasI believe I can often identify a person's ethnicity just by listening to the person speaking English on the radio or TV with some degree of reliability. Sometimes the speech characteristics are obvious and sometimes very subtle. I am not saying why this is the case but rather that it is a fact. Perhaps someday someone will create such an analysis program on a computer. It would offer an opportunity for ethnic profiling. It is not clear if this work is a step in that direction or is just a way to match (or not) a sample with pre-existing samples.
Can envision lots of applicationsPosted on: 03-15-10 | By: Salvatore SalamoneThis is very useful technology. I can see it being incorporated into many applications and having many uses.
It would certainly make two-factor authentication easier to use, manage and deploy compared to the many hardware token methods used today.
