Abstract: A detection method for underwater acoustic pulse under non-cooperative condition based on short time-fractional Fourier transform (ST-FRFT) is proposed in this paper. There are two kinds of underwater acoustic pulse signals, one is the Linear frequency modulation(LFM) signal and the other is the continuous wave (CW) signal. The matched filtering method cannot be used directly under non-cooperative situation due to lack of signal parameters information. The Fourier transform (FT) has good gain for the CW signal detection, but the gain decreases for the LFM signal detection, especially for the LFM signal with large modulation. The FRFT has not only good energy-concentration for the LFM signal but also contains the FT for the CW signal, so that it can be used to detect the CW and the LFM signal. In order to obtain time domain and frequency domain features of the received signal simultaneously, the detector was designed by the short-time fractional Fourier transform (ST-FRFT) in this paper, that it can be used to detect intercepted the LFM and CW pulse signal without any information about the parameters of signals. The detection performance for the CW and the LFM signal was analyzed by using the ST-FRFT and the STFT respectively, the results show that the processing gain of the LFM based on the FT was affected by modulation rate and non-linear change with window length, so it is difficult to estimate optimal window length under non-cooperative situation. The processing gain of the matched filter can be obtained approximately by the FRFT, therefore, the detection performance of the LFM signal was improved. The simulation results and experimental results in lake demonstrate the correctness of the theoretical analysis and the effectiveness of the proposed method.