Abstract: Underwater acoustic beacon signal detection is one of the important technologies to search for black boxes. Aiming at the detection difficulties caused by low sound source level and narrow pulse of underwater acoustic beacon signal, an improved matched filtering method based on multi-pulse integration is proposed by full use of the characteristics of short beacon signal period and strong correlation between adjacent pulse signals. This method overcomes the matched processing gain decrease due to signal distortion in standard signal matched processing through front and back pulse matching formula, and improves the integral gain through multi-pulse accumulation. The simulation results show that when the cumulative number of multiple pulses is 6, the detection performance of the improved method is about 4dB higher than that of the traditional single pulse matched filter in the ideal white noise environment; in the presence of obvious Doppler effect, the adaptability of the traditional method is reduced, while the performance of the improved method is improved by 9 dB. The validity of this method is also verified by processing the lake test data.