Impulsive modulated signals are widely existed in mechanical systems especially when faults occurred Traditional envelope analysis has provided an important and effective approach to extract the impulsive features but failed to get an ideal result under strong background noise. A new approach to demodulation and de-noising the impulsive modulated signal with strong background noise using multi-scale morphological filters is put forward and examined in this study. Adaptive method for selecting the length and amplitude of structure elements using local-peak-value is introduced. Four different type of structure elements, i.e. flat, line, triangle and cosine SE, are constructed and applied in demodulate and de-noising the simulated impulse modulated signal with strong noise The effects of the four type SE and traditional envelope demodulation method are compared. Results show that the line, triangle and cosine SE presented an excellent ability to extract the impulsive features from strong noised signal. The effect of the flat SE is not very ideal but still better than the envelope demodulation method
