Research On Fetal ECG Signal Detection Method Based On Compressed Sensing

In recent years,the development of signal processing technology is in full swing,and telemedicine has become possible under this background,gradually appearing in people’s life.For medical signals,the requirements of real-time transmission,accuracy of signals and portability of equipment are strict.The signal processing technology based on compressed sensing has gradually become the most widespread and widely used research method in this link.Therefore,based on fetal ECG signals,this paper conducts an in-depth study on compressed sensing technology.Aiming at the shortcomings of existing methods,a new measurement matrix with simpler hardware implementation and better compressed sampling performance is proposed,and a reconstruction algorithm with higher accuracy and better real-time performance is proposed.Specific work is as follows:First of all,in view of the deficiencies of existing sparse representation studies on fetal ECG signals,this paper analyzes the sparsity of fetal ECG signals,compares the sparse representation effects of discrete cosine transform,discrete orthogonal transform and redundant dictionary method through experiments,and carefully considers the changes of internal parameters of wavelet transform.A large number of experimental results show that,Discrete orthogonal transformation is more suitable for sparse representation of fetal ECG signals,in which the sparse representation matrix composed of Rbio5.5 wavelet is the best.Secondly,aiming at the problem that traditional random measurement matrix has good compression performance but difficult hardware realization,while traditional fixed measurement matrix has simple hardware realization but poor compression performance,the construction method of measurement matrix was designed in this paper according to the characteristics of Logistics-Tent sine sequence.The LTS Bernoulli measurement matrix based on Logistics-Tent sine chaotic sequence was proposed,and the theoretical correlation of the LTS Bernoulli measurement matrix was deduced by principle,and it was proved that it was lower than the traditional Gaussian measurement matrix and could meet the RIP characteristics.Thirdly,in view of the shortcomings of the existing greedy algorithms,which meet the running time requirements but have poor accuracy,and the existing optimization algorithms,which have high accuracy but large computing resources and long running time,considering the joint characteristics and sparse block characteristics of fetal ECG signals,OLS algorithm is used to develop the characteristics of high atomic accuracy,and a joint block multiple orthogonal least squares algorithm is proposed.It has excellent performance in reconstruction accuracy and running time.The experimental simulation results on OSET original fetal ECG signal data set show that,combined with Rbio5.5 wavelet sparse representation matrix and LTS Bernoulli measurement matrix mentioned above,compared with the latest traditional greedy algorithm,the compression rate of accurate reconstruction can be increased to about80%,and the running time can be saved by about 50%.