In order to realize the detection and recognition of underwater acoustic signals, the characteristic expression of coherent detection signal under spectral aliasing was given theoretically, and a kind of signal demodulation method based on Hilbert transform was proposed by studying the characteristics of underwater multi-source coherence detection signals. Acoustic frequency demodulation of underwater multi-sources was realized in case of spectral aliasing. In this method, the Hilbert transform is carried out after filtering and smoothing processes to obtain the analytic form of the signal. Then, the second filtering and smoothing processes are carried out for the squares of the analytic signal modulo to separate the aliasing bands. After that, the spectrum of the obtained signal is analyzed. Frequency of each underwater acoustic signal can be calculated according to the frequency shift value finally. A laser coherence detection system was built in the optical darkroom, and the underwater acoustic signals of 2~6 kHz were experimented. The results show that the proposed method can effectively separate the signal bands that are mixed together in the detection signal, the accurately extract the frequency of each underwater acoustic signal, and the frequency extraction repeatability is not more than 2.5 Hz.