Abstract: Inspired by the signal processing model of the human auditory system， a method of the degree of association between signal components was proposed. Firstly， the signal was successively processed with Gammatone bandpass filter， inversed order filter and halfwave rectification. Then， the once， twice and thrice extremums were picked out in the waveform after halfwave rectification to generate three kinds of units sequences. By using these units sequences， two synchronisms between adjacent filter channels signals and the once and twice extremums points in different filter channels were calculated respectively. The former synchronism may present the correlation， and the other reflects the degree of fluctuation association. A vibration signal measured from a gear system with tooth backlash is analyzed by the method. The results show the proposed method can discover those signal components with high degree of association， and it is favorable for reducing data size and revealing the dynamic characteristics of the measured system.

Key words: bandpass filter, timefrequency decomposition, auditory model, faults diagnosis, envelope