Current harmonics generated by nonlinear loads and the use of semiconductor switching drives caused widespread concern and attracted attention in power systems at all times. This paper applied an adaptive detection approach based on Hopfield Neural Network optimum theory to the harmonic detection. It presents the principle of estimation first, and then the neural network architecture will be built and simulated. The adaptive neural network-based signal processing technique is used to know the harmonic parameters. This knowledge would make it possible to compensate the harmonic components. By emulating this harmonic detection system in MATLAB, the result verifies the validity and the rapidity of the approach.