Abstract-In the article, the method of early diagnostics of nascent defects of mechanical assemblies (friction nodes, bearing pairs, gears) is considered by on-line determination of high-order statistical moments of the vibro-acoustic signal. To determine such statistical moments, a parabolic spline approximation of the discrete signal values is used. The spline approximation coefficients are determined at each discrete sampling interval by digital filter expressions. Expressions for the second, third and fourth statistical moments of the vibroacoustic signal are determined using its parabolic and cubic spline approximation. The signal processing is simple enough and can be implemented in a microprocessor controller. The characteristics of the proposed method are determined by an example of processing a vibro-acoustic signal of a harmonic shape. It is shown that with a relatively rare signal sampling, the error in determining the dispersion and the fourth statistical moment (Excess) of the signal is rather small. The use of a parabolic spline approximation of a vibro-acoustic signal makes it possible to significantly reduce the frequency of the signal sampling and to reduce the amount of data needed to determine the statistical characteristics of the signal.