Abstract: From the time and the frequency domains a study of de-noising problem for lower signal to noise radio is made: the adaptive noise reduction with LMS algorithm is used in the time domain analysis and the wavelet de-noising is used in the frequency domain study.The two methods are analyzed by contrast,and their advantage and weakness are discussed.A sample signal of speech with noise is simulated,and the de-noised sample signal is compared with the pure speech sample signal.The research result shows a good effect on noise reduction by means of the selected methods.