The above analysis of the spectral parameters (SP) and the processes 
of transformation that are used to predict the spectral envelope (SE) 
of speech signals in Recommendation G.729. 
The technique of comparing the characteristics of the processes in 
forecasting SE various alternative equivalent spaces SP. 
The results of experimental studies predicting SE in classical linear 
spaces SP (LSP) and LFA highest cleavage (LSP-HC). 
It is shown that the transition to space LSP-HC reduces prediction error SE.