NCJ Number
57708
Date Published
1976
Length
8 pages
Annotation
TWO KINDS OF ERROR RATES WERE TESTED, THE NONADMITTANCE OF AUTHORIZED PERSONNEL (TYPE 1) AND THE ADMITTANCE OF NONAUTHORIZED PERSONNEL (TYPE 2), FOR BOTH THE AUTOMATIC SPEAKER VERIFICATION (ASV) AND THE AUTOMATIC HANDWRITING VERIFICATION (AHV) SYSTEM.
Abstract
THESE ADVANCED DEVELOPMENT MODES WERE PROCURED BY THE BASE AND INSTALLATION SECURITY SYSTEMS PROGRAM OFFICE OF THE U.S. AIR FORCE SYSTEMS COMMAND'S ELECTRONIC SYSTEMS DIVISION. SPEAKER VERIFICATION IS PERFORMED USING A DIRECTIONAL DYNAMIC MICROPHONE AS THE SENSOR, A PREPROCESSOR THAT DIGITIZES AND FILTERS THE SIGNAL, AN ENTRANT TERMINAL, A PROMPTER THAT SELECTS A PSEUDO-RANDOM 4-WORD PHRASE TO BE SPOKEN BY THE ENTRANT, AND A PROCESSOR THAT COMPARES THE CRITICAL FEATURES WITHIN THE SPEECH PATTERN WITH A REFERENCE FILE. SIGNATURE VERIFICATION UTILIZES A PEN TRANSDUCER AS THE SENSOR, COMBINING A BALLPOINT PEN WITH A PRESSURE-SENSING INSTRUMENT. THE RESULTANT SIGNAL IS DIGITIZED, FILTERED, AND COMPARED TO THE REFERENCE FILE OF PERTINENT SIGNATURE CHARACTERISTICS FOR THAT INDIVIDUAL. SUFFICIENT TRIALS WERE MADE TO DERIVE STATISTICALLY DATA ON THE PERFORMANCE OF BOTH SYSTEMS USING A SAMPLE OF OVER 200 PERSONS. BOTH THE EFFECT OF A VARIETY OF PARAMETERS ON THE ERROR RATES AND THE INDEPENDENCE OF ERRORS FOR REPEATED USE BY ENTRANTS WERE DETERMINED. RESULTS INDICATE AN OVERALL ERROR RATE OF 2.5 PERCENT FOR ASV TYPE I ERRORS, AND AN OVERALL ERROR RATE OF 0.64 PERCENT FOR ASV TYPE II ERRORS. TYPE I ERRORS FOR AHV PERFORMANCE, ON THE OTHER HAND, OCCURRED AT A RATE OF 6.81 PERCENT FOR ALL USERS, WHILE TYPE II ERRORS WERE AT A RATE OF 3.19 PERCENT. IT IS CONCLUDED THAT THE POSSIBILITY FOR THESE TYPES OF ERRORS, PARTICULARLY WITH ASV SYSTEMS, ARE ACCEPTABLY SMALL. TABLES AND FIGURES ARE INCLUDED IN THE REPORT. SEE ALSO NCJ 38864, 52569, 57707, AND 57709-57726. (AUTHOR MODIFIED--PRG)