Abstract: To improve the BER performance of the Extended Binary Phase Shift Keying (EBPSK) demodulation under low SNR，the LDPC code is introduced. But the special impacting filter used in EBPSK demodulator brings more difficulty in LDPC decoding after obtaining posterior probability information. The support vector machine(SVM) method in this paper selects only several samples from the filter output used for LDPC decoder. The simulation results show that higher SNR gain can be obtained with this method. The effect of the posterior probability information got by different sampling rates and different ways on decoding performance is analyzed. Which indicate that more accurate posterior probability can be obtained with SVM method under low sampling rate and low SNR. So SVM is an effective method in EBPSK demodulator to get posterior probability information for LDPC decoding.