J.R. Boston, L. Baloa, and G. Akyol (USA)
fuzzy logic, signal detection, uncertainty, QRS detection.
Fuzzy combination of evidence in signal detection incorporates a measure of uncertainty and can conclude that a signal is present, that a signal is not present, or that no decision can be made. That is, the decision is uncertain. An uncertain decision can avoid an outright error, and identification of uncertain evidence can be of value in applications in which recognition of the inability to make a decision has lower cost than an error. This paper describes an approach to fuzzy combination of evidence that allows uncertain classifications. The amount of uncertainty introduced depends on the membership functions used, and an extension of Bayesian cost analysis is developed to evaluate the performance of the detector as a function of the amount of uncertainty. The analysis is illustrated using the detection of QRS complexesinanoisyelectrocardiogram(ECG) for three different applications: critical care monitoring ofECG; monitoring long-term trends in heart rate; identification of the precise time of occurrence of a QRS complex to synchronize an experimental procedure with the heartbeat. Results using ECG waveforms with simulated noise are presented. The lowest cost may be achieved by maximizing the uncertainty, allowing an optimum amount of uncertainty, or allowing no uncertainty, depending on the relative costs associated with uncertain decisions and errors.
Important Links:
Go Back
