Research And Implementation Of The Parametric Loudspeaker Inverse Control Based On Neural Network

The acoustic directional technology, which is based on the parametric array theory,is the hotspot and frontier technology between nonlinear acouistics and nonlinear signalprocessing field. Based on the fact that the audio signal of parametric loudspeaker hasthe outstanding features of energy concentration and high directivity, this kind oftechnology has a broad application prospect. The development of parametricloudspeaker was restricted by the complex nonlinear signal processing problems."Berktay" theory, which only takes into account the second-order nonlinear terms, isstill not well meet the requirement of the parametric loudspeaker’s signal processingand system control. The nonlinear control method will be a new way to break throughthe technical bottleneck.This article aims to apply the neural network that can be infinitely approximateoptional nonlinear system in order to make the parametric loudspeaker’s model betterapproximate the well-know acoustic wave KZK equation, which solves the problem thatthe KZK equation gets the higher-order analytical solution difficultly. Furthermore, itcan provide a more accurate theoretical basis for parametric loudspeaker’s signalprocessing and system control. Centering on this goal, the main research contents of thisthesis are as follows:According to the theory and method of neural network modeling, the mathematicmodel of parametric array acoustic system is established. These models are mainlybased on BP network, RBF network and CMAC network. Moreover, Further studyincludes support vector machine’s (SVM) regression application (SVR) and BPmodeling which is optimized by genetic algorithm (GA-BP) for parametric loudspeaker.The nonlinear simulation model of the parametric array is established by MATLABsoftware. Then, the models which are establised by GA-BP algorithm and SVMalgorithm have been tested. The results show that the parametric acoustic system modelwhich based on GA-BP has strong generalization capability and rapid globalconvergence features, so it’s more suited to establish the model of the parametricloudspeaker, thus this lay the foundations of identification and inverse identification for directional loudspeaker.The neural network inverse control method is used to establish inverse controlmodel of parametric loudspeaker by MATLAB/SIMULINK based on the nonlinearneural network model of the parametric array has been established. The main workincludes: The parametric loudspeaker’s model and inverse model are identified off-lineby GA-BP algorithm. The BP-PID inverse control system which combines PID controllink is proposed. The result show that PID control parameters has a greater impact onthe whole system performance, thus the BP-PID adaptive inverse control system whichPID control parameters can self-tuning is proposed. Furthermore, the CMAC-PIDadaptive inverse control system is proposed. Sinusoidal signal and square wave signalsare used to make simulations on three kinds of complex inverse control system, theexperiments results show that the neural network inverse control system has a goodperformance.In accordance with the theoretical analysis and simulation results, the performanceof BP-PID inverse control algorithm has been tested. The digital signal processinghardware platform which based on the BF533chip has been developed successfully.The hardware test platform has been built for the neural network inverse control system,and the BP-PID control algorithm program has been designed. Furthermore, thedifferent excitation source was used to test inverse control system, and the results ofsystem test show that the BP-PID inverse control method can effectively improve thenolinear distortion of the parametric loudspeaker.