基于小波分解的颈动脉波特征提取算法

Feature Extraction Algorithm of Carotid Artery Pulse Wave Based on Wavelet Transform

目的:利用小波变换的时频局域化性质,检测出存在于颈动脉波信号(CAP)中的奇异点和奇异角,并且精确检测奇异角出现的位置。方法:小波变换具有多分辨率等特点,能够通过放大信号的任意细节部分进行时域分析。采用离散小波变换法结合db1小波能够检出脉搏信号中的奇异U角。利用计算CAP时域特征点的小波变换极大值坐标来精确定位脉搏时域特征点,通过检测脉搏的特征参数以及脉搏的突变特征参数,可以客观判定人体脉搏变化规律。结果:CAP信号WT分解很好地抑制了各种病理性、基线漂移等干扰,为进一步进行特征提取创造了条件,基于第一细节信号d1的特征点定位几乎不受各种病理性、基线漂移等干扰的影响,定位比其他传统处理技术更为准确。结论:本文提出了基于小波分解的颈动脉波特征点提取算法,取得高达100%的检测率。在含有大量噪声和伪差的脉搏信号中,仍具有较高正确检出率和良好的抗噪性。根据计算得到CAP信号时域特征点的小波变换极大值的坐标,再利用极大值表征准确测定脉象时域特征点的坐标,能够克服脉搏时域特征点定位不准的问题。

Objective： Wavelet transform was used to check the existence of singular point and strange a-ngle which exist in carotid artery wave signal （CAP） because of its time-frequency local change nature. At the same time, strange angle＇s position could be located. Methods： Wavelet transform has the multi-resolution characteristics, which can analyze signals in time do- main by amplifying signals＇ any detailed part. Combine discrete wavelet transform method with db 1 wavelet could judge exis- tence of strange angle of U. Human pulse changing rule could objective determined by calculating wavelet transform＇s maxi- mum position of CAP time domain feature points could located the time domain feature points of pulse accurately, and detect- ing characteristic parameters of pulse and mutation characteristic parameters of pulse. Results： WT resolving of CAP signals could restrain variety of pathologies and baseline wander＇s interference, that create the conditions for feature extraction. Feature points location based on first details of the signal d 1 be unacted on variety of pathologies and baseline wander＇s interference. Its position precision is higher than other traditional processing technique. Conclusions： Wavelet transform algorithm was pro- posed for extracting carotid artery pulse＇s feature points which obtained better detection rate. Under the interference of noises and pseudo differential pulse signal, it still has higher correct detection rate and better noise resistance. Based on the wavelet transform maximum coordinates of CAP signals＇ time domain feature points and coordinates of the time domain feature points of pulse calculated by maxima representation, which could resolve the problem of position invalid of pulse feature points in time domain