Aiming at the problem of long diagnosis and treatment recovery period and low recovery efficiency of traditional stroke patients, a finger rehabilitation training control system based on speech recognition is designed to help patients carry out finger exercise training and obtain the perception ability of finger angle, speed and position, and provide rehabilitation physicians with finger rehabilitation evaluation and training data reference. In order to realize the finger rehabilitation training control system, based on the finger movement perception system, the system is divided into hardware circuits, lower computer control systems, and voice recognition human computer interaction systems. Combining the unique advantages of the HMM algorithm, the HMM algorithm is applied to the voice interaction system for pattern matching, and the simulation test of the finger rehabilitation system is performed. The application results show that the rehabilitation training system based on speech recognition proposed in this study meets the design requirements, has good safety and reliability, and has high application value for future finger rehabilitation training.