Enhanced estimation of power spectral density of noise using the wavelet transform

In practice, different methods for enhancing speech hidden in noise are used but none of the available methods is universal; it is always designed for only a certain type of interference that is to be suppressed. Since enhancing speech masked in noise is of fundamental significance for further speech signal processing (subsequent recognition of speaker or type of language, compression, processing for transmission or storing, etc.), it is necessary to find a reliable method that would work even under considerable interference and will be modifiable for different types of interference and noise. The methods known to date can basically be divided into two large groups: single-channel methods and multi-channel methods. The basic problem of these methods lies in a rapid and precise method for estimating noise, on which the quality of enhancement method depends. If the noise is of stationary or quasi-stationary nature, its determination brings further difficulties. A method is proposed in the article for enhancing the estimation of power spectral density of noise using the wavelet transform.