The invention provides a multi-gear 
code rate adaptive 
demodulation system and method based on a neural network so as to solve the technical problems of 
high complexity of realization of an existing multi-gear 
code rate adaptive 
demodulation system and large computation amount of a 
demodulation method. In the demodulation 
system, an ADC sampling module samples an analog modulation 
signal; a code element feature point extraction module carries out detection on 
phase jump points of the sampled 
signal by utilizing a one-dimensional 
convolution neural network trained by a neural 
network construction module; a 
code rate estimation module estimates code rate of the sampled 
signal according to the detection result; a signal-to-
noise ratio 
estimation module estimates signal-to-
noise ratio of the sampled signal; a controller module selects low-pass filter coefficients and an interpolation structure of a demodulation module according to the code rate 
estimation result and the signal-to-
noise ratio 
estimation result, and calculates sampling rate conversion ratio of the demodulation module; and finally, the demodulation module carries out demodulation on the sampled signal according to the selected low-pass filter coefficients and the interpolation structure as well as the calculated sampling rate conversion ratio.