Abstract: The existing feature parameters are unable to effectively identify the modulation mode of bionic underwater acoustic communication signals mimicking dolphin whistles. This paper proposes an automatic identification method of the frequency-modulated bionic underwater acoustic communication signals mimicking dolphin whistles that combines multiple features. Initially, the whistle contour is estimated through the application of preprocessing and least squares polynomial fitting. Subsequently, the feature parameters are extracted based on the estimated whistle contour. The simulation results demonstrate that the extracted features exhibit robust and effective identification capabilities. Ultimately, the support vector machine classifier is employed to facilitate the automated identification. The recognition effect of this technology is validated through a lake experiment, and the influence of modulation parameters (symbol width and frequency offset) on recognition rate is analyzed. The results demonstrate that the modulation frequency offset exerts a more pronounced influence on the average recognition rate. When the frequency offset is 50 Hz and the signal-to-noise ratio exceeds 5 dB, the average recognition rate is approximately 90%. When the frequency offset is not less than 100 Hz and the signal-to-noise ratio is greater than 0 dB, the average recognition rate is approximately 95%.