
"A voice input is received from the caller, and the characteristics are applied to a plurality of acoustic models, which includes a generic acoustic model and acoustic models of any previously identified callers, to obtain a plurality of respective acoustic scores," the patent reads.
"The caller is identified as one of the previously identified callers or as a new caller based on the plurality of acoustic scores.
"If the caller is identified as a new caller, a new acoustic model is generated... which is specific to the new caller."
The technique breaks the speaker's voice into segments and then compares them segment by segment to pick the most probable subject based on previous experience.
Pascovici envisages the system being built into PCs, servers and handheld computers, and installed in call centres to speed up the transfer of callers to the correct department.