A New Efficient Non-uniform Filtering Method Based On Level-crossing Sampling Scheme

In the theme of filter design,many people put forward various conceptions in order to improve filter performance and reduce computational complexity.In order to improve the computational complexity fundamentally,the sampling method of digital filtering system is more and more widely concerned.The most popular and basic is the uniform sampling method based on Shannon sampling theorem.But for a certain kind of research object,the uniform sampling method does not apply.Using uniform sampling to process signals that change only in certain periods of time(such as ECG,sound signals,etc.)produces a large number of redundant signal sampling points that make the multiplication of subsequent operations much larger.In order to solve this problem,the Nonuniform sampling filter is produced as a very effective method.My research direction is non-uniform sampling filter system.I begin to study and try my best to change the existing non-uniform sampling signal still exist in the field of complex algorithm and difficult to process.LCSS horizontal cross-sampling is one of the non-uniform sampling schemes.This method utilizes the characteristics of the signal itself and effectively reduces the number of sampling points in the signal without carrying information.Then the computational complexity of the subsequent processing signal is reduced.In the research army of LCSS horizontal cross-sampling,many people have put forward their own ideas and improved methods.Some people directly improve the sampling principle of LCSS horizontal cross-sampling,such as the CLS method.Others add windows to the algorithm to optimize the active region sampling method of active signal.But in the previous research methods,almost all the non-uniform sampling methods have to reconstruct the signal uniformly.Therefore,based on the previous idea that "the signal processing after non-uniform sampling does not need uniform reconstruction",I put forward my own improvement method,the main work is as follows:1.A sampling,filtering and recognition system based on offline over-sampling filter is proposed.After LCSS non-uniform sampling,the special method of over-sampling solves the problem that the signal sampling points and filter responses are difficult to align.It does not need to do uniform reconstruction,which can greatly improve the filtering efficiency and reduce the computational complexity.Moreover,the operation principle of filter response formula is improved,which greatly reduces the operation amount of the system and the number of multipliers,in order to reduce the energy consumption in the signal processing process of the system.2.Because the LCSS sampling method is based on the maximum frequency and amplitude of the signal,we propose a new idea of dynamically windowing the signal with ASA in time domain,and then selecting different LCSS parameters according to the maximum frequency of each window.(Total number of sampling lines,sampling line density,etc.)Finally,a new improved dynamic LCSS filtering system is obtained.The improved system is more flexible in sampling process and the number of sampling points is reduced to further reduce the computational complexity of the system in the signal processing process.3.In view of the above design,the ECG and speech signals with distinctive feature points are used to detect the effects.The reduction degree of the signal passing through the system and the distortion degree of the signal feature point are used to ensure that the system does not lose important data information on the basis of reducing the computational complexity.