Implementation Of ECG Signal Processing Algorithm Based On FPGA

In the era of rapid development of information technology,the incidence of cardiovascular and cerebrovascular diseases has increased year by year due to high-intensity work pressure.The prediction of this type of disease has always been a problem for doctors and our concern.Electrocardiogram(ECG)is an important basis for detecting and judging whether or not there is cardiovascular disease.The treatment level of ECG signals directly determines the accuracy of diagnosis of this disease.With the increasing requirements for the accuracy and timeliness of ECG signal processing in human society,the current conventional processing methods no longer meet the requirements of precision,so more accurate algorithm processing and faster algorithm implementation are needed.Based on the complexity and flexibility of neural networks,it is an effective way to predict the ECG signal class through NN training.In addition,from the perspective of the circuit,it is also helpful to improve the processing efficiency of the algorithm by increasing the clock operating frequency.In this paper,the ECG algorithm is modeled and tested by software such as MATLAB,and the simulation of the algorithm hardware is realized by VIVADO software.Firstly,this paper analyzes the development status of ECG signal processing and neural network,and clarifies the necessity of improving ECG processing algorithm and optimizing ECG algorithm hardware implementation module.The introduction of ECG generation and extraction principles provides a theoretical basis for ECG acquisition and feature extraction.In addition,based on the Huffman algorithm,the symbolic and translational segmentation processing is performed on the data to be compressed,so that the compression ratio of the module reaches 0.38.Secondly,this paper analyzes the ECG signal,establishes the entry point of ECG signal processing,and then innovates on the basis of the original QRS feature extraction algorithm,and reduces the hardware implementation difficulty under the premise of ensuring accuracy.Then,three kinds of methods commonly used in the field of classification and recognition are used to simulate the classification and recognition of ECG signals.The results show that under the premise of limited samples,the improved method based on convolutional network proposed in this paper is classified and identified with an accuracy rate of 0.97.In addition,based on the Huffman algorithm,the symbolic and translational segmentation processing is performed on the data to be compressed,and the compression ratio of 0.3 is obtained while preserving the details of the more compressed data.Finally,this thesis carries out partial hardware simulation of the implementation of ECG neural network algorithm and compression algorithm,and proposes a hardware compression scheme for ECG signals.The results show that a single sample of 1280 data is selected as the original storage file,and its storage size is 12.5K,and it is 4.3K after compression.In the aspect of neural network,the design process of some sub-modules is given,and the sub-module is optimized by parallel processing,pipeline,etc.The highest clock operating frequency of the sub-module can reach 0.5GHz.