This paper presents the use of vibration analysis based on autoregressive (AR) trispectrum in milling process to indicate the relationship between machined surface quality of workpiece and spectrum character under a variety of different cutting parameter settings, such as the depth of cut, the feed rate and the spindle speed. The metal cutting experiments were performed on steel A3 material without using any cutting fluid. The experimental results for slices analysis of trispectra indicated that the spectral peaks were abundant as the milling process went smoothly. The roughness value measured reflects the machined surface quality of workpice. According to the changes for spectral peak and roughness value, a conclusion was made that the proposed trispectrum slices method for choosing cutting parameters and monitoring the machined surface quality in milling process was practicable.