In order to reduce the impact of electromagnetic interference on signal transmission, the balise uplink signal transmission process and its susceptibility to interference signals were analyzed, and an adaptive demodulation method based on the symbolic least mean square (LMS) algorithm was designed. In order to achieve the demodulation method in the hardware platform, the range of the intermediate variable of the adaptive algorithm of LMS algorithm was determined through simulation calculation. And the weight coefficient was updated by truncation operation. Setting the equalizer length, step factor, and median filter coefficient to 1,1/64, and 16, respectively, can obtain excellent demodulation performance without occupying too much hardware resources. The demodulation algorithm was verified on field programmable gata array (FPGA). The experiment shows that the bit error rate of adaptive demodulation in FPGA is 0.000 001 when the signal-to-noise ratio is 6 dB. And the measured bit error rate is basically consistent with the simulation bit error rate when the signal-to-noise ratio is greater than or equal to 6 dB. The bit error rate of FPGA adaptive demodulation method is less than 10^-6 at different train speed levels.