The linearization-based approach of logical functions defined by the orthogonal terms and not only the truth tables or the decision diagrams was studied. The linearization procedure may consist of less than n iterations if the system is described using less than n basic vectors. The procedure can be extended successfully to incompletely defined functions at the expense, true enough, of an increased complexity. For linearization by the truth table, the algorithm optimizes the number of vertices at the levels beginning from the roots of the binary tree. There are at least n levels. At each level, the autocorrelation function is calculated, then the best is selected, and the function is convoluted. The algorithm for linearization on the orthogonal terms calculates the autocorrelation function with the greatest Hamming weight.