Aimed at the problem of hardly extraction early crack in critical infrastructure components of major equipments, a sparse code shrinkage (SCS) denoising for weak signals based on independent component analysis (ICA) is proposed. Namely, the probability density function (PDF) of the signal independent coefficients is estimated by the generalized Gaussian model (GGM) in the ICA space. And the nonlinear denoising is finished by maximum a posteriori (MAP) estimate. By extracting different SNRs of weak crack signal, the results show that this method can extract the signal from the input SNR less than -27dB. And the waveform and spectrum of the extracted signal are substantially consistent with the original signal. At the same time, the results are much better than those from the wavelet denoising method. The method is very suitable for weak signal extraction.