DETECTION OF A POTENTIALLY SHIFTING SIGNAL PARAMETER.
COLUMBIA UNIV DOBBS FERRY N Y HUDSON LABS
Pagination or Media Count:
Parameter instability or non-stationarity in an input process is a potential problem for various signal processing schemes. In a processor, for example, which hard-clips the input, the clipping level must be at the true mean value of the signal in order to avoid severe distortion. A sudden shift in the mean can occur, however, if a large low-frequency transient is present in the signal or if a malfunction occurs in the electronics. Another non-stationarity problem in signal processing arises in matched filtering if there is a sudden change in the amplitude values of several Fourier components of a known signal waveform. If a change occurs, the old matching function is no longer optimal. The problems of a mean shift, a trend change, or an amplitude shift are special cases of the parameter shift problem in the general linear model. In this paper the detection of a shift of one parameter will be discussed. Author