Robust transient signal detection using the oversampled Gaborrepresentation

The Gabor representation has been investigated previously as an effective and efficient means of detecting transient signals. The major problem associated with the most commonly used form of the Gabor representation is that it assumes that signals have integer arrival times and modulation frequencies-an assumption that, if violated, causes signal miss-match and deterioration in detector performance. The present correspondence presents the concept of oversampling the Gabor representation, which makes the representation more robust to certain types of signal mismatch. To show this, the performance of an oversampled Gabor detector is compared to that of critically sampled Gabor detectors. Gabor coefficients are calculated in the oversampled representation via the least squares method, which has been shown to be superior to the more commonly used biorthogonal function method in terms of corresponding detector performance