Research On Multi-sensor Fusion Technique Base On Wavelet Neural Network

Multi-sensor information fusion is the essence of dealing with uncertain information; uncertain information needs to be able to handle the mathematical method. Wavelet neural network is combined a wavelet transform theory with artificial neural networks. A new neural network model, which with a strong approximation ability combines the wavelet transform time-frequency localization good nature and self-learning ability of neural networks, and fault tolerance in dealing with complex non-linear, uncertain system is superior to other issues to show the convergence speed of traditional neural networks, fault tolerance, with a wide range of applications. In this paper, information fusion will be a variety of coal mine detection signal by wavelet neural network and genetic algorithm fusion.This paper analyzes the multi-sensor information fusion, wavelet transform, artificial neural networks, genetic algorithms the basic theory, operation mechanism, and their respective advantages and disadvantages. Characteristics of BP network the Sigmoid function itself, in fact, only a sub-optimal neural network need to be improved; BP algorithm uses the error function gradient descent iterative way, the convergence speed is very slow, and easily falls into local optimum, and the specific The extreme position and the initial value of the relevant weights. Wavelet network using genetic algorithms to optimize the parameters in the study, to achieve the global optimization.In this paper, in coal mine the monitoring of gas emission conductes a study to design a multi-sensor information fusion based on the model, the fusion algorithm is compared wavelet neural network with genetic algorithm with the BP standard method. Simulation results show that the algorithm has greatly improved monitoring of the accuracy error of the network significantly faster convergence, and the monitoring results are satisfactory.