In the light of coexistence of random noise and power interference noise in rotor vibration signals, the noise reduction method based on the combination of singular value and singular vector is proposed. First, the signals processed with signal value decomposition (SVD) method, the effective order of singular value can be obtained from singular value spectrum, in this way the random noise can be reduced. From the relationship between singular value and singular vector we can get the corresponding singular vectors. Then, these singular vectors are calculated with fast Fourier transformation (FFT). The singular vectors corresponding to the power interference are filtered based on the FFT amplitude spectrum feature. Finally, the time-domain signal with all the other singular values and singular vector are reconstructed to reduce the random noise and power interference noise in the target signals. Simulation experimental results demonstrate that the proposed method can effectively reduce the random noise and power interference noise, which is obviously superior to the notch filter method.