Voiceprint Classification for an Extended Population
L. G. Kersta · The Journal of the Acoustical Society of America · 1966
The binary-code automated classification system for Voiceprints previously described for a 12-member population has been extended to a 120 member population [Voiceprint classification, J. Acoust. Soc. Am. 37, 1217(A) (1965)]. The classification system is based upon the uniqueness of a selected quantized amplitude, frequency vs time matrix. Experimental data indicate unique speaker selection of better than 90 per cent for the tested population. When the classification is based serially on five selected cue words and subgrouping the selection accuracy is improved. The goal sought was for a method, preferably automated, by which the population in which a subjective search must be made for unique identification could be reduced. This has been accomplished. An automated identification method equating that of trained subjects has not yet been achieved. The major part of the work reported here was done at the Bell Telephone Laboratories.