Channel estimation in noisy conditions using time-frequency domain filtering

We propose a time-frequency domain filtering technique that improves the performance of channel estimation in noisy environments. A chirp signal is used as a training sequence to estimate the impulse response of a linear time-invariant (LTI) system. The additive white noise in the channel decreases the performance of the estimation. Since the chirp signal is localized but the noise is spread in the joint time-frequency plane, the output of the LTI system can be denoised by using time-frequency techniques. The chirplet decomposition is used for noise filtering. It is shown that, chirplets increase the accuracy of the estimation results compared to the classical cross-spectral method.