F.N. Buzylyov, A.V. Gerasimov, V.B. Vyrov, S.E. Zheleznova, A.N. Moroz

Algorithm of spectrum subtraction, used for suppression of additive stationary noise, is presented in brief. Recom-mendations are given on methods enabling to avoid short-time and frequency-limited energy bursts in pure signal that are heard as "bells" or "flowing water". It should be noted that suggested methods allow one to solve the problem of spectrum subtraction ? "musical noise". One of the most widespread ways of "musical noise" suppression consists in spectrum time smoothing. For this purpose the recursive time filtering is applied to STFT coefficients of pickup signal. It concluded that good suppression can be achieved by using adaptive two-dimensional filtering algorithms, such as non-local averaging or bilateral filters, applied for noise reduction when working with images.