The goal of this work was to perform statistical analysis of experimental trajectories of temperature-myographic BFB training. Approximation of signals was implemented using fractional-linear functions. It was found that the number of approximation functions decreased with increasing number of sessions. Moreover, if the number of approximation functions was large enough, it was possible to find the asymptotic behavior of monitored signal with the help of initial values of the signal. It means that it is possible to reveal the potentialities of patients during 3 or 4 sessions of training. Estimation of the time interval of active work showed that there were two various laws for values of temperature and values of myogram: for the first monitored variable, power law; for the second variable, exponential law. But for both monitored parameters the constants were positive, which meant that there was a minimum of time interval for the active "work" for every patient. Correlation analysis (i.e., analysis of autocorrelation and cross-correlation functions) of temperature and myogram signals showed that there was a correlation between monitored parameters. It means that for both signals we can consider both positive and negative time lag dependence between values of temperature and myogram. Statistical analysis of signal behavior made it possible to construct a phenomenological model of temperature-myogram BFB-training, which comprises the basic parameters of patients. With the help of the model it is possible, in particular, to calculate the sojourn time of system at any part of the phase plane. This characteristic can be used as an integral estimation of the efficiency of BFB-training.