The method of music digital recording analysis is described in this paper. The aim of the method is to detect sound frequencies and their durations for further musical notation creation. The method is based on the step by step averaging of the sound signal with parallel amplification to save the oscillation amplitude that is the sound volume. Averaging is carried out with the discrete analog of the Steklov functions. The decrease of the amplitude during the averaging process is in a right proportion with the frequency of the signal. So it provides the possibility to suppress the high-frequency harmonics and noise using step by step smoothing. This allows to detect the fundamental frequency as a pure single harmonic. After subtracting this harmonic from the original signal this algorithm can be used again to detecting the other harmonics of the higher frequencies. Algorithm is implemented in a computer program with the Java programming language.