The acoustic emission signals of rolling bearing with different type of defects are de-noised and illustrated by the continuous wavelet transform and scalogram. Morlet wavelet function is selected and the wavelet parameters are optimized based on the principle of minimal wavelet entropy. The soft-threshold de-noising is used to filter the wavelet transform coefficients. The de-noised signals obtained by reconstructing the wavelet coefficients show the obvious impulsive features. Based on the optimized waveform parameters, the wavelet scalogram is used to analyze the real AE signal from the defective rolling bearing in experimental test rig. The results indicate that the proposed method is useful and efficient for signal purification and features extraction.