Research On Pulse Wave Signal Acquisition System Design And Time-frequency Feature Analysis Method

Pulse diagnosis is a very important diagnostic method in traditional Chinese medicine theory,and it still has a high status in the development of modern Chinese medicine theory and practice.The diagnosis results of pulse diagnosis depend on the personal experience and skills accumulated by doctors for a long time.This complex,subjective and fuzzy theory not only requires doctors to spend a lot of time to master,but also does not meet the requirements of the development of modern medicine.In order to promote the modernization of pulse diagnosis,many scholars have done a lot of work in collecting and analyzing digital pulse wave signals in recent decades.According to the logical order,the main research contents of modern acquisition and analysis of pulse wave include the process of objective acquisition,preprocessing and data analysis.However,there are some deficiencies in the current research of each link,such as the single function of the sensor,the low level of practicability,the preprocessing algorithm can not deal with complex clinical data well,and the performance of feature extraction method is insufficient.In order to solve these problems,a series of researches are carried out in the aspects of design and optimization of acquisition system,construction of signal quality evaluation system,improvement of preprocessing algorithm and process,parametric representation of signal and feature extraction:(1)A pulse wave sensing sensor array based on multi-sensor cooperation is designed,and a relatively complete pulse wave acquisition system is constructed.Two kinds of strain gauges with different properties,foil strain gauge and semiconductor strain gauge,are used in the pulse wave signal sensor,which overcomes the shortcoming of insufficient performance caused by single type sensor used in most equipment.It can record accurate static contact pressure signal and dynamic pulse wave signal at the same time.Based on clinical practice,the ergonomic design of the acquisition system,automatic adjustment and correction of sensor performance,circuit modularization and other aspects are optimized.A closed-loop control system is constructed,which can automatically adjust the pulse pressure or remind the operator to adjust the sensor according to the acquired signal.(2)A pulse wave preprocessing framework based on signal quality detection is proposed to solve the problem of period segmentation and signal quality evaluation in complex clinical situations.By using the segmentation point screening method of local point detection,the framework can obtain better periodic segmentation performance in clinical databases with diverse signal morphology,interference and quality.By analyzing the types and causes of abnormal signals in the database,a real-time signal quality evaluation system is proposed,which can be used for auxiliary operation of pulse wave signal acquisition process.Through the combination of signal quality evaluation system and preprocessing process,the detection ability of abnormal signal and abnormal period in off-line data processing is enhanced,and the reliability of subsequent data analysis is effectively improved.(3)In order to represent the contour information and detail information more effectively,a parameterization method of pulse wave average period signal based on discrete Fourier series is proposed.Considering that pulse wave signal is produced and acted by physiological processes of different frequencies,the method extracts contour information and detail information of different frequency components in the signal respectively.It avoids the error of existing methods in key point detection and over fitting of local information,and improves the expression ability of the model for signal.Experimental results show that this method can not only obtain smaller representation error,but also has stronger ability of disease identification in the same kind of methods.(4)Aiming at the accurate representation of physiological information such as main wave,tidal wave and dicrotic wave in single period pulse wave signal,a feature extraction method based on Gabor time-frequency atom and sparse representation is proposed.This method not only takes into account the different frequency of different physiological activities,but also represents the time difference of different physiological activities.It also uses the good morphological changes and time-frequency characteristics of Gabor atoms to sparse represent the average periodic signal of pulse wave.Experiments show that the proposed Gabor feature vector,which represents the main fluctuation information of pulse wave,can achieve better recognition performance in disease classification based on pulse wave signal.