Abstract: The core of underwater acoustic reconnaissance is to capture the pulse signals transmitted by other platforms without prior knowledge. Continuous wave (CW) signal and frequency modulation (FM) signal are the most widely used detection pulses. CW signal can be effectively detected by power spectrum entropy algorithm at low signal to noise ratio (SNR), but the performance of this algorithm for FM signal detection is degraded. Fractional Fourier transform (FRFT) can gather energy of FM signal. By using the properties of FRFT and power spectrum entropy algorithm, a fractional power spectrum entropy detector is designed, which can detect unknown signal in the fractional domain. And then, the energy accumulation of FRFT for FM signal is analyzed, and the method of FRFT order search is optimized. The simulation results and experimental results at sea demonstrate that the new detector has good detection performance on FM signal and has no effect on CM signal. The detector can complete unified automatic detection for unknown underwater acoustic pulse signal.