Skip to Main Content
A performance analysis of an automatic signature-verification system based upon the use of a three-axis force-sensitive pen to transduce the dynamics of the handwritten signature into electrical signals suitable for real-time computer processing is presented. False-rejection and imposter-acceptance (i.e. type I/type II error) curves are presented for a variety of operating conditions. For typical `real world' conditions, the equal-error rate (where type I error rate equals type II error rate) is about one percent. The performance analysis was based upon a data base of 5220 true signatures obtained from 58 subjects over a four- month period, and 648 attempted forgeries obtained from 12 forgers. The forgers were given copies of the true signers' signatures, told how the verification system operates and what it measures, allowed to watch video tapes with close-up views of the signatures to be forged as they were being written, and allowed to practice for a three-week period. Particular advantages of the signature-verification system reported here are high performance, low storage requirements (typically 200 to 300 bits per user), and low-cost implementation in a stand-alone microprocessor unit.