A novel method for extracting weak signal based on the theory of stochastic resonance was proposed. It was implemented by using the cooperative effect of noise instead of adding noise to a weak signal. The previous studies show that several disadvantages exist in the adding noise method. Due to the fact that adding noise to the weak signal will decline the signal-noise-radio(SNR) of the input of the nonlinear system, it will lower the detection effect. Additionally, the resonance point is determined manually, and also obvious offset of the peak position may occur in the result analyzed. Here the novel method was applied to analyze the weak laser-Raman spectrum of a CCl_4 sample (liquid in capillary) which was measured with SPEX-1403 laser-Raman-spectrometer from 250 cm-1 to 418 cm-1 on 5 mW (output power of laser). The spectrum was analyzed with the novel method by varying the data to [-4^13, 3^50] and declining μ from 2.0 and compared with the result acquired by wavelet analysis. The results show that the novel method can greatly increase the SNR of input signal easily to detect the weak signal. With the deep research of the theory of stochastic resonance and improvement of the new method, it will probably become a power tool to analyze spectrum. A novel method for extracting weak signal based on the theory of stochastic resonance was proposed. It was implemented by using the cooperative effect of noise instead of adding noise to a weak signal. Due to the fact that adding noise to the weak signal will decline the signal-noise-radio (SNR) of the input of the nonlinear system, it will lower the detection effect. The new method was applied to analyze the weak laser-Raman spectrum of a CCl_4 sample (liquid in capillary) which was measured with a SPEX-1403 laser-Raman-spectrometer from 250 cm- 1 to 418 cm-1 on 5 mW (output power of laser). The spectrum was analyzed with the novel method by varying the data to [-4 ^ 13, 3 ^ 50] and declining μ from 2.0 and compared with the result acquired by wavelet analysis The results show that the novel method can greatly increase the SNR of input signal easily to detect the weak signal. With the deep research of the theory of stochastic resonance and improvement of the new method, it will probably became a power tool to analyze spectrum.