Bootstrap-based signal denoising

We present frequency domain bootstrap-based signal denoising schemes applicable to real-valued non-Gaussian signals embedded in additive white Gaussian noise. The two proposed schemes separate the noisy signal into frequency bands using Fourier or wavelet transforms. Each frequency band is tested for Gaussianity by evaluating its kurtosis. The bootstrap method is applied to increase the reliability of the kurtosis estimate. Noise effects are minimized using thresholding schemes on the frequency bands that are estimated to be Gaussian. The estimate of the signal is obtained by applying the appropriate inverse transform to the modified data. The denoising schemes are tested using a stationary and non-stationary signal type. Results show that FFT-based denoising schemes perform better than WT-based denoising schemes on the stationary sinusoidal signal, whereas WT-based schemes outperform FFT-based schemes on chirp outperforms soft thresholding.