Power Spectral Density Estimation of Noisy Signal Based on Wavelet

The power spectral density estimation of signals plays an important role in the understanding and analysis of the spectral distribution of signals. In this paper, the authors adopt dyadic wavelet, multi-band wavelet and complex wavelet to estimate the power spectral density of noisy signals, especially to speech signals and complex modulation signals. They also analyze the properties of different wavelet methods for decomposition and denoising. The analysis shows that their methods are efficient in decreasing the estimation runtime and increasing the estimation accuracy, and can be used in real-time engineering applications.