In order to achieve the fault diagnosis of the rotating machine equipment，a method based on improved time synchronous averaging（TSA）denoising using correlation detection and fault features extracting using squared envelope spectrum is proposed.A reference sub-signal is arbitrarily taken from the sampled vibration signal. The reference sub-signal starts from the initial time of the vibration signal and moves point by point along the time axis to the end time. Meanwhile，the correlation coefficient between the reference sub-signal and its masking sub-signal in the vibration signal is calculated. The correlation coefficient optimal threshold are obtained using the step iterative algorithm. All the masking sub-signals whose correlation coefficients are greater than the optimal threshold are selected as overall synchronous signals whose arithmetic average is calculated for the denoised target signal. The target signal analysis is achieved using squared envelope spectrum. The analysis results of the simulated signal，gear and rolling bearing fault vibration signals have shown that the proposed method is an efficient implementation for rotating machine vibration signal denoising and fault feature extraction.