PERSONAL IDENTIFICATION SYSTEM USING MULTIPLE PARAMETERS HAVING LOW CROSS-CORRELATION

A personal identification system for use in security and various other types of pattern matching applications provides accurate recognition of a known person. This system uses a large quantity of course but inexpensive sensors to create a profile of a person seeking to enter the controlled space or to take other action. These sensors measure different statistically independent characteristics of the person, such as height, weight, stride, voice timbre, and time to perform various functions. The measurements so obtained may individually be of very low resolution. They are used to search a data base of patterns related to admissible individuals. Active agent methods are used to support learning and adaptation of the sensor set to address sensor drift and individual characteristic changes over time. A combination of many parameters will identify a person with very low probability of both false negative and false positive errors. The resulting low sensor resolution required allows such a system may be implemented at very low cost in comparison to other biometric systems.