With the development of high speed and high precision NC machining technology, high cutting stability of NC machine tool is required. The uncertainty processing is insufficient in the traditional cutting state monitoring. In this paper, an uncertainty processing method for cutting state monitoring is proposed based on modal interval theory. The uncertainty in traditional monitoring methods is described by using the width of modal interval to solve the monitoring uncertainty problem. In order to verify effectiveness of the proposed method, a cutting experimental platform is built. The cutting information of the NC machining is obtained by acceleration sensor. The cutting states are divided into three processing stages: stable, transition and chatter state by using timefrequency analysis. The interval feature of the different stages is extracted by using the wavelet packet energy percentage based on modal interval. The interval feature is encoded by Lloyd algorithm, and regarded as the input vectors of generalized hidden Markov model. Finally, the cutting stats of NC machine are identified by generalized hidden Markov model state recognition method. The experimental results show that the proposed generalized hidden Markov model recognition method based on modal interval is superior to the traditional hidden Markov model recognition method.