To determine whether a living object is human being or animal by detecting the respiration signal using the life-detector based on radar to penetrate the building or debris. To fully utilize the nonstationary character of the respiration signal, short time Fourier transform (STFT) is employed to get the signal's time-frequency representation. Singular value decomposition (SVD) is then used in the spectrogram to extract feature vector for pattern identification. Human beings and animals can be identified from the respiration sensed by the life-detector based on radar. Experimental results show that the method based on STFT and SVD is stable and efficient to differ from the human being and animal.