Application Research Of Wavelet Transform And Neural Network In Spatial Position Perception

At present,most methods of ultrasonic denoising are based on Fourier transform and short-time Fourier transform.Because the ultrasonic signal is a kind of non-stationary signal,and often appears as local singular,and the basis of the Fourier transform is a global basis,there is no local ability,so the Fourier transform in processing such signals,its shortcomings and shortcomings become increasingly prominent.Although the short-time Fourier transform makes up for the shortcomings of the Fourier transform to some extent,the window of the shorttime Fourier transform is fixed and its width does not change during a transformation,so the short-time Fourier transform still cannot solve the problem of processing unsteady signals.In addition,it is found in the research that the equipment debugging before measurement,calculation in measurement and statistics after measurement will introduce a large uncertainty error,affecting the positioning accuracy.Therefore,these uncertainty errors have become practical problems that must be solved in this paper.To solve those problems,wavelet transform and neural network are introduced in this paper.For the denoising of ultrasonic noise signal,the wavelet threshold denoising method based on wavelet transform is introduced to realize the denoising of noise signal.PSO-RBF neural network model is introduced to reduce the error of three-sided positioning algorithm.In the process of their research,this paper also builds a set of spatial location sensing system.Specific research contents are as follows:Aiming at the above problems,this paper introduces wavelet transform theory to reduce the noise of ultrasonic signal.Because wavelet transform can be used for time-frequency analysis of signals and has local processing ability,it is an ideal tool for processing non-stationary signals and also a hotspot of current research on processing non-stationary signals.Aiming at the uncertainty errors introduced in equipment debugging,calculation and statistics,this paper designs the pso-rbf neural network model based on the ability of neural network to approach the truth value to the greatest extent.This paper mainly completed the following research work:1.Through the investigation and review of the literature,the main problems in the research of improving the ultrasonic positioning accuracy are found out,and the environmental factors affecting the ultrasonic signal are analyzed,and the characteristics of ultrasonic signal affected by noise are investigated.2.Wavelet threshold denoising method is applied to denoising processing of ultrasonic signals,and the following researches are mainly carried out:(1)The selection of wavelet basis in wavelet threshold denoising method is studied.As the wavelet bases of different signals are different,the selection of wavelet bases is mainly based on experiments.In this paper,three kinds of nonstationary signals are selected as test signals.(2)The method of choosing the number of layers in wavelet denoising is studied.In the wavelet threshold denoising method,the selection of the number of layers determines the integrity of signal decomposition,and too low number of decomposition layers will lead to the inconspicuous matching characteristics of signal and noise.The number of decomposition layers is too large,and the signal error after reconstruction is large.In this paper,the following methods are adopted for the selection of the number of layers: firstly,the approximate range of the number of layers is selected by referring to the empirical method,and then the wavelet basis selected by the above experiments is tested by using three kinds of non-stationary signals,and the number of layers is determined from the test results.(3)The fusion index method of wavelet noise reduction parameter selection is studied.Above studies are based on root mean square error,signal-to-noise ratio,smoothness and correlation coefficient to evaluate the efficacy of the wavelet and the number of layers to dry,this section mainly studies how to integrate different evaluation index into a composite value,simplify the parameter selection process,this paper proposes a composite evaluation based on wavelet denoising parameter selection method.In this method,the wavelet base and the layer number are determined by constructing the composite evaluation value.3.The wavelet denoising method and pso-rbf neural network model studied above are applied to the ultrasonic localization algorithm to improve the traditional three-sided localization algorithm.4.A set of ultrasonic positioning system is designed.Mainly completed some design work:Completed the overall design of the system;The thesis completes the hardware design of the system.The hardware circuit mainly includes: node control circuit,node power circuit,ultrasonic transmitting circuit,ultrasonic receiving circuit,radio frequency circuit and interface conversion circuit.The completion of the system software design.Software design is divided into upper computer program and each node program.The upper computer software mainly includes various algorithms,control programs and communication programs.The designed node program mainly includes: ultrasonic transmitting and receiving program,radio frequency transmitting and receiving program,ranging algorithm program and so on.5.The performance of the designed system is analyzed.Experimental results show that the wavelet denoising method and pso-rbf neural network model proposed in this paper are correct and feasible.The experimental results also prove that the system designed in this paper improves the positioning accuracy to some extent and has certain application value.