The invention relates to a recursion demodulation method for an electrical tomography system, which is characterized by comprising the following steps: (1) establishing an initial equation and a recursion equation according to an excitation frequency f and a sampling frequency f s of a system; (2) taking a zero phase moment of an excitation signal as an initial moment, and calculating initial values of a data matrix B and a state matrix P based on an initial equation in step (1) according to two sampling point data of a measurement signal immediately after the initial moment; (3) substituting the initial values of the data matrix B and the state matrix P as well as newly added sampling point data into the recursion equation in step (1) so as to update B and P; and (4) judging whether a demodulation result meets the accuracy requirement of the system or not, if not, returning to step (3), and if so, finishing demodulation and outputting the amplitude and phase data of the measurement signal. The recursion demodulation method only needs at least two sampling point data, and can improve demodulation accuracy and anti-noise ability by adding sampling points.