Using edge information in time-frequency representations for chirp parameter estimation

Time-frequency representations of a signal can provide a useful means for obtaining parameter estimates for signals consisting of various chirps. We demonstrate the utility of including edge information extracted from these time-frequency representations when using a Hough transformation to perform this task. In particular, we show that using the edge information: (1) reduces the variance of the chirp parameter estimates in the case where the chirp signal has a single component; and (2) reduces the amount of spurious cross talk results when the signal has multiple chirp components. We further demonstrate a variation of our technique that detects the onset and duration of individual chirp components. We propose this technique as a fast preprocessing step for other algorithms such as maximum likelihood estimation which can provide very accurate parameter estimates. (c) 2004 Elsevier Inc. All rights reserved.