The work aims to propose an adaptive multi model speed sensorless control strategy to solve the problem that the models in the traditional extended Kalman filter algorithm may not match the actual working conditions so as to improve the control performance of packaging and printing machines, reduce the number of sensors used in packaging and printing machines and the cost of motor control systems, and reduce the failure rate of packaging and printing machines and the volume of motors. Based on the traditional extended Kalman filtering algorithm, multiple models were introduced to interact with each other in the input stage using a state transition probability matrix. At the same time, hidden Markov models were used to design state and observation sequences for multiple models. By iteratively updating the state transition probability matrix of the observation matrix for the interaction stage of multiple models, the matching degree of the model in the face of environmental disturbances was improved. The Matlab/Simulink simulation showed that the improved algorithm significantly improved the estimation accuracy of speed, and its anti-interference ability was significantly improved in the face of environmental disturbances. Compared with the traditional extended Kalman filter algorithm, the improved algorithm in this paper improves the accuracy of the control system, and also improves the dynamic performance and robustness, making the improved algorithm more suitable for packaging and printing machinery.