PURPOSE: To provide the method and device for signal processing which can perform optional numerical calculation processing and signal processing with practical precision by using a limited scale of hardware.

CONSTITUTION: As for an addition result obtained by adding a constant to an input signal, a width multiplication arithmetic part A calculates width multiplication terms less than a multiplier required to obtain requested precision and the arithmetic result of an expansion expression arithmetic part C which adds all multiplication results, obtained by multiplying the respective width multiplication terms by a weight coefficient defined previously by a coefficient derivation part 2, by an adder 1 is connected as a fixed information quantity to a next stage; while constants and weight coefficients are determined by the coefficient derivation part 2 in order by an automatic generating method for one function which consists of plural connected processing units as one unit and a compound function automatic generating method constituted by connecting plural units, the addition result is repeated as an input signal for next time to structure a specific Taylor arithmetic system.