Modern bioengineering is aimed at using advanced technologies to develop new methods for studying and regulating biosystems. It covers a wide range of social problems related to medicine, agriculture, ecology and industrial industry. At the same time, to obtain the necessary information about the state of these systems and the processes occurring in them, the entire spectrum of natural manifestations of their vital activity is used. One of the aspects of obtaining such information is the study of biosignals in the acoustic range of the spectrum. However, acoustic signals in biosystems have characteristic features due to the stochastic non-stationary form of representation, which significantly complicates the selection of a useful signal. The authors studied the spectral non-stationarity of sound biosignals, and as an example considered the biosystem of a beehive, where these features are manifested to a significant extent. This is due to the fact that the sound signals of the honeybee biosystem manifest themselves as dynamic models of random processes carrying significant information not only in the form of parameters, but also in the form of local non-stationarities. The information component of such a model is quite multifaceted and is of interest both for theoretical analysis and for practical application.