New method of linear time-frequency analysis for signal detection

Previous work on standard signal detection via linear time-frequency (TF) transforms has focused on detectors based on the standard linear TF transforms: the short-time Fourier transform (STFT), the Gabor transform, and the wavelet transform (WT). This paper examines two methods of improving linear-TF-based detection. The first method uses the Coifman-Meyer-Wickerhauser (see IEEE Transactions on Information Theory, vol.38, no.2, p.713-18, 1992) "best basis" concept for detection based on adaptive-window-length local cosine and wavelet packet transforms. The second method uses the Mallat-Zhang (see IEEE Transactions on Signal Processing, vol.41, no.12, p.3397-3415, 1993) matching pursuit algorithm for detection based on a combination of STFT and (standard) wavelet functions.