Detection of weak signals using wavelet de-noising

The authors present the theory and the technique for using the wavelet transform to detect weak signals which are buried in noise. Based on the features of wavelet transform denoising and auto-correlation matched filtering, a weak signal time-frequency information extracting method is developed The frequency-domain information is extracted by an auto-correlation matched filter and. is used to deduce the wavelet decomposition scale levels. Using the different decomposition features of signal and noise, non-linear filtering is performed in each scale. A reconstructed signal is obtained with improved signal-to-noise ratio (SNR), which makes the detection of time-domain information easy. Computer simulation results show that signal time-frequency information can be extracted at SNRless than or equal to0dB by using the method.