Fast Kurtogram can equally divide the spectrum into several groups of different numbers of frequency bands by constructing a finite impulse response filter. The calculation speed of this method is very fast, but the way of dividing the frequency spectrum is not connected with the characteristics of the signal. The incomplete information extracted during actual application may lead to the failure to detect obvious fault information in the results. This paper proposes a variable spectrum segmentation method. First, initialize the window of the power spectral density and search for the minimum sequence in it. The position of the minimum point in the spectrum is used as boundaries. Secondly, more group boundaries can be obtained by increasing the size of the window. Since the fluctuation of the power spectral density is related to the actual signal, there are different ways of dividing for different signals. Meyer wavelet is used to construct the filter and reconstruct the components of the signal. Finally, calculate the spectral negentropy of each component, and multiple sets of boundaries can construct a new variable spectral segmentation negentropy diagram (VSEntrogram). The rolling bearing outer ring fault data prove the effectiveness of the method.