For the sake of resolving the problem of large torque ripple problem of switched reluctance motor, especially under hysteresis control strategy, back propagation (BP) neural network control strategy is introduced on the account of traditional direct instantaneous torque control, and the direct instantaneous torque control is optimized and adjusted with the square of the torque error as the performance index function, which restrains the torque ripple of the motor during operation. In addition, based on the overall double closed-loop control, sliding mode control is introduced for the improvement of the speed link, which improves the response speed and robustness of the system. Finally, Matlab/Simulink is used for simulating the traditional direct instantaneous torque control and optimized direct instantaneous torque control, the effectiveness and feasibility of the proposed strategy are proved.