Research On The Application Of Single Component Bayesian Method In Sound Signal Analysis

In this paper,the empirical mode is mainly focused on the process of model establishment in speech signal analysis,using empirical mode decomposition method to generate a set of adaptive basis and decompose the data into a single component based on the characteristics of the acoustic grain data itself.In combination with Bayesian method,the acoustic grain signals are analyzed and predicted.By using Bayesian method and single component analysis method,the order number of prediction model is reduced effectively,the prediction precision and the computational resources are improved.By using empirical mode decomposition,the acoustic grain signal is decomposed and the robustness of this method is improved.Firstly,in the introduction,the background of speech signal testing and the application situation are introduced.Meanwhile,the single component signal obtained by empirical mode decomposition is obtained by combining bayesian method with a better vibration speech analysis.Then the origin of bayesian method and its application in recent years are introduced,and the linear recursive prediction model is selected.Then,introduces the method of separation of the major component of several comments,emphatically introduces the short-time Fourier transform,wavelet transform and empirical mode decomposition,after group,combined with bayesian method and empirical mode decomposition,the use of simulation analysis,for white noise sequence,analyzed the correlation between the eigenfunction,obtain the correlation coefficient of different order number only for10^-4.The feasibility of a single component separation method is verified.The non-stationary time domain signal obtained by the reaction spectrum inversion is used to compare the different generation parameters and select the exponential curve as the contour function.The feasibility of bayesian dynamic prediction is verified.Has carried on the simulation analysis for speech signal,compares the forecasting results of complex signal and a single component signal,finally applied to the actual speech signal analysis,concluded that each single component prediction error?0.0013 m/S2?and less than the overall prediction error?0.009 m/S2?10%.