A novel maximum torque per ampere (MTPA) method based on power perturbation for a field-oriented control interior permanent magnet synchronous motor (IPMSM) drive system is proposed in this study. The proposed MTPA method, which is parameter independent and can improve the motor operation at both start-up and low speed, is designed based on the power perturbation by using the signal injection in the current angle. Moreover, the influence of current and voltage harmonics to the MTPA control can be eliminated effectively. Furthermore, to enhance the robustness of the control system, an online tuning scheme for an integral-proportional controller using a new wavelet fuzzy neural network with disturbance torque feedforward control is developed where the disturbance torque is obtained from an improved disturbance torque observer. Finally, some experimental results using an IPMSM drive system based on a low-price digital signal processor are presented. From the experimental results, the proposed control approach can guarantee the control performance of the speed loop, even under a cyclic fluctuating load.