next up previous [pdf]

Next: Acknowledgments Up: Liu and Fomel: Local Previous: Application to multicomponent data


We have introduced a new time-frequency decomposition that uses regularized nonstationary regression with Fourier bases to represent the time-frequency variation of nonstationary signals. The decomposition is invertible and provides an explicit control on the time and frequency resolution of the time-frequency representation. Experiments with synthetic and field data show that the proposed local time-frequency decomposition can depict nonstationary variation and provide a useful domain for practical applications, such as ground-roll noise attenuation and multicomponent image registration.