Due to the complexity of ruin scenes in the post-disaster search and rescue operations, the UWB echo signal reflected from the human target is often contaminated by the clutters and noises from the surrounding environment, which makes it difficult to extract the chest-wall micro-motion information that indicating the presence of human target, causing misjudgment and leakage. Thus, it remains as a major challenge in disaster relief work to reliably and effectively extract the vital signs modulated in the echo signal. The paper proposes a novel feature extraction methods based on wavelet transform to conquer the challenge. And a comparison with the traditional power spectral density based method was carried out. Experimental results show a better detection performance with the improved detection accuracy in ruin scenes using the new method. And it has the potential to be applied in more complicated ruin scenes.