The rescue robot can detect the human bodys infrared thermal signal effectively, and realize the remote recognition and rescue of distress personnel. The traditional method used spectrogram detection method, but the detection effect was not good in low SNR environment. A new method of infrared thermal signal detection is proposed based on time frequency analysis. A parametric model of the human body infrared heat release signal is constructed, and the dynamic smoothing process of the human body infrared heat release signal is processed by the Doppler frequency fuzzy number search method. The timefrequency analysis of nonstationary time varying signal is carried out, and the negative frequency part is eliminated. Antiinterference filter is realized. The Hough transform is implemented by the postpositive high order accumulated amount of the signal, the focus of signal in the time frequency plane is increased. The signal is suppressed and the noise is suppressed. The simulation results show that this algorithm has higher accuracy and better antiinterference ability, and the search and recognition ability of the rescue robot is improved.