Since telemetry vibration signal in frequency domain has characteristics of complex composition, nonlinear and non-stationary, as well as strong noise, a telemetry vibration signal detection method based on multi-scale spectral kurtosis map was proposed. Firstly, the collected telemetry vibration signal was zero drift amended and eliminated the trend term. Secondly, adaptive decomposition method was used to multi-scale decompose the signal, and the spectral kurtosis was used to remove false component: Thirdly, the time-frequency distribution was calculated by screening component, and divided multi-layer and multi-scale, then the spectral kurtosis of the corresponding scale band signal was calculated to draw spectral kurtosis map, and the filter band was selected based on the principle of maximum kurtosis; Finally, CZT transform was used to refine the filtered signal spectrum for abnormal frequency vibration signal. The results of simulation and practical application demonstrate effectiveness of this method.