This article develops a data-adaptive time-frequency representation that overcomes some limitations of the short-time Fourier transform, while avoiding the cross-terms that make the Wigner distribution and other bilinear representations difficult to interpret. The adaptive time-frequency representation uses Gaussian basis functions but varies their time width and chirp rate with time and frequency to achieve high signal concentration everywhere. A measure of local signal concentration allows fully automated determination of the optimal basis parameters. The adaptive method is expensive computationally, but may provide much better performance than any currently known technique.
|Original language||English (US)|
|Number of pages||9|
|Journal||IEEE Transactions on Acoustics, Speech, and Signal Processing|
|State||Published - Dec 1990|
ASJC Scopus subject areas
- Signal Processing