Study On The Methods Of Detection And Feature Extraction Of Gastric Electromagnetic Signal

Stomach is an important apparatus in digestive system.According to the statistics,the number of people with stomach disease has surpassed 1 percent of the total worldwide population.Although experts have been studying it for many years,some common organic and functional stomach disease still remains unclear.Detection technique of electrogastrogram(EGG)takes on simple operation,no hurts and no pains,so it is of great significance for the diagnosis and treatment of stomach disease.EGG contains the slow wave signal which reflects gastric contraction rhythm,but it lost the fast wave signal which describes the contraction strength of stomach due to attenuation of human tissue.To solve this problem,this paper puts forward a new method of using the sensor based on the giant magneto impedance effect(GMI)to measure the gastric magnetic signal and extracted the gastric magnetic fast wave successfully.This paper mainly studies the methods of gastric magnetic signal denoising,schemes of feature extraction and classification methods.This paper derives the formulas of magnetic field intensity and electric potential inspired by electrical control activity(GEA)based on the description of GEA about its generation at first.In addition,this article also analyzes ellipsoidal gastric electrical dipole model and simulates electromagnetic field distribution of the stomach in the ANSYS environment.The magnetic field intensity data obtained from the simulation to can help determine the optimal placement of GMI sensor,it makes sense to improve the SNR and the selection of the follow-up gastric magnetic signal processing algorithms.Considering the features of EGG,which is ultra-weak amplitude,super-low frequency and extremely poor SNR.This paper realized the AD conversion of the signal with a high resolution 24 bit microcontroller in the hardware design part.A new method of morphology and RLS adaptive algorithm is proposed to separate high frequency component and biological signal baseline drift exist in the gastric magnetic signal in this paper.In addition,this paper also introduced the wavelet packet algorithm to filter out rest noise signals and finally get pure gastric magnetic data.Obtaining the characteristic information of signal is the task of feature extraction.This article introduced a method of matching pursuit combined with nonparametric basis function to get statistical information which describles the amplitude change of fast wave signal.In order to reduce the misdiagnosis rate,this thesis also extracts features of EGG through three-order B-spline curves and obtained the characteristic parameters of gastric electrical signals in the end.Cluster gastric magnetic signals by K-means algorithm and analyzed the signal with time and frequency domain methods.