Home - News ( United States | United Kingdom | Italy | Germany ) - Football scores

Showing content from https://patents.google.com/patent/CN104778342B/en below:

CN104778342B - A kind of heart sound feature extracting method based on wavelet singular entropy

Publication number

CN104778342B

CN104778342B CN201410015681.9A CN201410015681A CN104778342B CN 104778342 B CN104778342 B CN 104778342B CN 201410015681 A CN201410015681 A CN 201410015681A CN 104778342 B CN104778342 B CN 104778342B

Authority

CN

China

Prior art keywords

heart sound

wavelet

singular

cardiechema signals

signal

Prior art date

2014-01-14

Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)

Expired - Fee Related

• Espacenet
• Global Dossier
• Discuss

• 238000000034 method Methods 0.000 title claims abstract description 15
• 238000000354 decomposition reaction Methods 0.000 claims abstract description 20
• 238000000605 extraction Methods 0.000 claims abstract description 14
• 230000009466 transformation Effects 0.000 claims abstract description 7
• 230000011218 segmentation Effects 0.000 claims abstract description 6
• 239000011159 matrix material Substances 0.000 claims abstract description 4
• 238000006243 chemical reaction Methods 0.000 claims description 2
• 238000003672 processing method Methods 0.000 abstract 1
• 230000005236 sound signal Effects 0.000 description 26
• 238000005070 sampling Methods 0.000 description 7
• 238000004364 calculation method Methods 0.000 description 4
• 238000004458 analytical method Methods 0.000 description 3
• 238000001228 spectrum Methods 0.000 description 2
• 208000001778 Coronary Occlusion Diseases 0.000 description 1
• 230000002526 effect on cardiovascular system Effects 0.000 description 1
• 238000005516 engineering process Methods 0.000 description 1
• 230000008092 positive effect Effects 0.000 description 1
• 230000003595 spectral effect Effects 0.000 description 1
• 238000010183 spectrum analysis Methods 0.000 description 1

• Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
• Ultra Sonic Daignosis Equipment (AREA)

The invention discloses a kind of heart sound feature extracting method based on wavelet singular entropy, and it is related to a kind of processing method of cardiechema signals.Cardiechema signals sample frequency of the present invention is 8000Hz；Utilize Hilbert-Huang（HHT）Heart sound envelope is extracted, is then segmented again based on heart sound envelope so as to obtain one section of complete cardiechema signals, both including s1, s2 and s3, s4；6 rank wavelet decompositions are carried out to segmentation cardiechema signals using DB6 as morther wavelet；Singular value decomposition is all carried out to the contour signal and detail signal obtained after wavelet transformation, respectively obtains respective singular value matrix；Calculate respective singular entropy；Then weighting obtains the characteristic value of cardiechema signals.Wavelet singular entropy is introduced into cardiechema signals feature extraction by the present invention, is not only able to highlight the feature of cardiechema signals HFS complexity, and can determine the respective frequencies of low frequency signal, and cardiechema signals day part is quantitatively described；Reduce the data operation quantity of cardiechema signals processing, improve the arithmetic speed of heart sound characteristics extraction.

技术领域technical field

本发明属于信号处理领域，具体涉及一种基于小波奇异熵的心音特征提取方法。The invention belongs to the field of signal processing, in particular to a heart sound feature extraction method based on wavelet singular entropy.

背景技术Background technique

心音信号是利用信号采集设备采集到的发自心脏的声音，其中包含重要的心血管生物信息。心音信号在不同的人身上有着完全不同的特征并且具有极高的稳定性。Heart sound signal is the sound from the heart collected by signal acquisition equipment, which contains important cardiovascular biological information. Heart sound signals have completely different characteristics in different people and have extremely high stability.

目前国内外很多科研人员尝试了多种的方法来分析心音信号。Gauthier采用快速傅里叶变换（FFT）分析心音信号(Gauthier D, Akay Y M, Paden R G, et al. SpectralAnalysis of Heart Sounds Associated with Coronary Occlusions [C].6thInternational Special Topic Conference on Information Technology Applicationsin Biomedicine,2007:49-52)。傅里叶变换虽然具有良好的性质，能够实现时域到频域相互转换。但从傅里叶变换公式可以看出，它是以正弦波及其高次谐波为标准基的，因此它是对信号的一种总体上的分析，具有单一的局部定位能力，也就是在时域的良好定位是以频域的全部信号分析为代价的，对频域的良好定位是以时域的全部信号分析为代价的，傅立叶变换的频率谱中要么频率是准确的而时间是模糊的，要么时间是准确的而频率是模糊的，它不可能同时在时域和频域都具有良好的定位的能力。而心音信号主要集中在s1和s2上，相对整个心音周期s1和s2是两个相对短暂的时间段，因此使用傅里叶变换对心音信号进行分析并不是一个很好的选择。At present, many researchers at home and abroad have tried various methods to analyze heart sound signals. Gauthier uses Fast Fourier Transform (FFT) to analyze heart sound signals (Gauthier D, Akay YM, Paden RG, et al. Spectral Analysis of Heart Sounds Associated with Coronary Occlusions [C]. 6thInternational Special Topic Conference on Information Technology Applications in Biomedicine, 2007: 49-52). Although Fourier transform has good properties, it can realize mutual conversion from time domain to frequency domain. However, it can be seen from the Fourier transform formula that it is based on the sine wave and its higher harmonics, so it is an overall analysis of the signal and has a single local positioning capability, that is, in time Good positioning in the frequency domain is at the expense of all signal analysis in the frequency domain, and good positioning in the frequency domain is at the cost of all signal analysis in the time domain. In the frequency spectrum of the Fourier transform, either the frequency is accurate and the time is ambiguous. , or the time is accurate but the frequency is ambiguous, it is impossible to have a good positioning ability in both the time domain and the frequency domain. The heart sound signals are mainly concentrated on s1 and s2, which are two relatively short time periods compared to the entire heart sound cycle s1 and s2, so it is not a good choice to use Fourier transform to analyze the heart sound signals.

此外还有其他功率谱密度，样本熵方法分析心音信号。但这些方法都存在各自的缺点：功率谱是以信号服从高斯分布假设为前提，同时丢失了信号的相位信息；近似熵计算过程中存在自身数据段的比较,导致结果存在偏差。In addition there are other power spectral density, sample entropy methods for analyzing heart sound signals. However, these methods have their own shortcomings: the power spectrum is based on the assumption that the signal obeys the Gaussian distribution, and the phase information of the signal is lost at the same time; there is a comparison of its own data segments in the approximate entropy calculation process, which leads to deviations in the results.

发明内容Contents of the invention

本发明的目的是解决目前在心音信号特征提取中存在的问题，提供一种基于小波奇异熵的心音特征提取方法。The purpose of the present invention is to solve the existing problems in the heart sound signal feature extraction at present, and provide a heart sound feature extraction method based on wavelet singular entropy.

实现本发明的技术方案是：其特征在于该方法包括下面6个过程。The technical solution for realizing the present invention is: it is characterized in that the method includes the following 6 processes.

a)心音采集。心音信号采样频率为4000Hz。a) Heart sound collection. The heart sound signal sampling frequency is 4000Hz.

b）心音包络提取和分段。在实际采集中得到的心音过长因此需要对心音进行包络提取，在此采用的包络提取方法是利用希尔伯特—黄变换（HHT）提取心音包络，然后再基于心音包络进行分段从而得到一段完整的心音信号，既包括s1,s2，和s3,s4。b) Heart sound envelope extraction and segmentation. The heart sound obtained in the actual collection is too long, so it is necessary to extract the envelope of the heart sound. The envelope extraction method used here is to use the Hilbert-Huang Transform (HHT) to extract the heart sound envelope, and then based on the heart sound envelope. Segmentation to obtain a complete heart sound signal, including s1, s2, and s3, s4.

c）小波变换。采用DB6作为母小波对分段心音信号进行5阶小波分解。c) Wavelet transform. Using DB6 as the mother wavelet, the segmental heart sound signal was decomposed by 5th order wavelet.

d）奇异值分解。因为心音信号的频率成分主要集中在300 Hz以下，所以在此需要对小波变换后得到的轮廓信号CA（低频信息）和细节信号CD（高频信息）都进行奇异值分解。分别得到各自的奇异值矩阵Sa和Sd。奇异值分解格式如下。d) Singular value decomposition. Because the frequency components of the heart sound signal are mainly concentrated below 300 Hz, it is necessary to perform singular value decomposition on both the contour signal CA (low frequency information) and the detail signal CD (high frequency information) obtained after wavelet transformation. The respective singular value matrices Sa and Sd are obtained respectively. The singular value decomposition format is as follows.

其中，而为矩阵A的全部非零奇异值。in ,and are all non-zero singular values of matrix A.

e）计算奇异熵。把Sa和Sd分别代入下式得到轮廓信号CA的奇异熵Ha和细节信号CD的奇异熵Hd。e) Calculate the singular entropy. Substitute Sa and Sd into the following formula to obtain the singular entropy Ha of the contour signal CA and the singular entropy Hd of the detail signal CD.

其中，。in , .

f）计算特征值。按加权把轮廓信号CA的奇异熵Ha和细节信号CD的奇异熵Hd进行合并，得到最终特征值。f) Calculate the eigenvalues. The singular entropy Ha of the contour signal CA and the singular entropy Hd of the detail signal CD are combined by weighting to obtain the final feature value.

所述的心音特征值采用的是心音小波分解后的奇异熵，并将两者按照加权的形式结合。The heart sound eigenvalue adopts the singular entropy after the heart sound wavelet decomposition, and combines the two in a weighted form.

本发明具有积极地效果：本发明的一种基于小波奇异熵的心音特征提取方法，不但能够可以凸显心音信号高频部分复杂度的特征，而且可以确定低频信号的对应频率，并对心音信号各时段进行定量描述。本发明将小波奇异熵引入心音信号特征提取中，减少了心音信号处理的数据运算量，提高了心音特征值提取的运算速度。The present invention has positive effects: a heart sound feature extraction method based on wavelet singular entropy of the present invention can not only highlight the characteristics of the complexity of the high-frequency part of the heart sound signal, but also determine the corresponding frequency of the low-frequency signal, and analyze each heart sound signal. Quantitative description of time period. The invention introduces the wavelet singular entropy into the feature extraction of the heart sound signal, reduces the data calculation amount of the heart sound signal processing, and improves the calculation speed of the heart sound feature value extraction.

附图说明Description of drawings

图1为采样的心音信号；Fig. 1 is the heart sound signal of sampling;

图2 为分段的心音信号；Figure 2 is a segmented heart sound signal;

图3为 小波变换后心音轮廓信号；Fig. 3 is the heart sound contour signal after wavelet transform;

图4为小波变换后心音细节信号；Fig. 4 is the heart sound detail signal after wavelet transform;

图5为本发明流程示意图；Fig. 5 is a schematic flow chart of the present invention;

具体实施方式detailed description

（实施例1）(Example 1)

本具体实施方式的一种基于小波奇异熵的心音特征提取方法，具体过程如图1所示，说明如下：A kind of heart sound feature extraction method based on wavelet singular entropy of this specific embodiment, concrete process as shown in Figure 1, explains as follows:

一、心音采集。心音信号采样频率为8000Hz。图2为采集到的正常心音信号。1. Heart sound collection. The heart sound signal sampling frequency is 8000Hz. Figure 2 is the collected normal heart sound signal.

二、心音分段。采集到的心音过长因此需要对心音进行包络提取，通过这个过程从而得到一段完整的心音信号，既包括s1,s2和s3,s4。在此采用的包络提取方法是利用希尔伯特—黄变换（HHT）提取心音包络，然后再基于心音包络进行分段。图3为经过HHT变换，提取到的一个独立完整的正常心音信号。2. Segmentation of heart sounds. The collected heart sound is too long, so it is necessary to extract the envelope of the heart sound. Through this process, a complete heart sound signal is obtained, including s1, s2 and s3, s4. The envelope extraction method used here is to use the Hilbert-Huang transform (HHT) to extract the heart sound envelope, and then segment based on the heart sound envelope. Fig. 3 is an independent and complete normal heart sound signal extracted after HHT transformation.

三、小波变换。本研究采用DB6作为母小波对心音信号进行分解。小波分解尺度与采样频率和小波基有关。根据小波的采样定理，对于一个长度为M的原始采样信号，离散小波分解最多可以把信号分解成log₂M个频率级，因此小波分解的尺度是有限的并且分解的级数越多,则其所需的计算量越大所以应根据实际应用和需要来选择小波的分解尺度。在本文中采用的心音信号采样频率为4000Hz，根据小波子带含义，不同的小波系数代表不同频段信息，比如1阶分解的细节信号cd1表示的频率为：2000 Hz ~4000 Hz；2阶分解的细节信号cd2表示的频率为：1000 Hz ~2000 Hz……。而通过短时傅里叶变换分析心音信号，发现第一心音s1的频率成分主要集中在50 Hz ~150 Hz范围内，而第二心音s2的频率成分主要集中在50 Hz ~200 Hz范围内, 250 Hz ~300 Hz范围内出现第二个小峰值所以在此选择进行6层变换。这样得到6阶分解的轮廓信号ca6表示的频率为：0 Hz ~125 Hz, 6阶分解的细节信号cd6表示的频率为：125 Hz ~250Hz。Three, wavelet transform. In this study, DB6 was used as the mother wavelet to decompose the heart sound signal. The scale of wavelet decomposition is related to sampling frequency and wavelet basis. According to the wavelet sampling theorem, for an original sampling signal of length M, discrete wavelet decomposition can decompose the signal into log ₂ M frequency levels at most, so the scale of wavelet decomposition is limited and the more stages of decomposition, the more The greater the amount of calculation required, the wavelet decomposition scale should be selected according to the actual application and needs. The heart sound signal sampling frequency used in this paper is 4000Hz. According to the meaning of wavelet subbands, different wavelet coefficients represent different frequency band information. The frequency represented by the detail signal cd2 is: 1000 Hz ~ 2000 Hz... . By analyzing the heart sound signal through short-time Fourier transform, it is found that the frequency components of the first heart sound s1 are mainly concentrated in the range of 50 Hz to 150 Hz, while the frequency components of the second heart sound s2 are mainly concentrated in the range of 50 Hz to 200 Hz. , a second small peak appears in the range of 250 Hz ~ 300 Hz, so here we choose to perform 6-layer transformation. In this way, the frequency represented by the contour signal ca6 of the 6th order decomposition is: 0 Hz ~ 125 Hz, and the frequency represented by the detail signal cd6 of the 6th order decomposition is: 125 Hz ~ 250 Hz.

图4为小波变换后心音轮廓信号；图5为小波变换后心音细节信号。Fig. 4 is the heart sound contour signal after wavelet transformation; Fig. 5 is the heart sound detail signal after wavelet transformation.

四、奇异值分解。因为心音信号的频率成分主要集中在300 Hz以下，所以在此需要对cd6，ca6进行奇异值分解，代入下式。分别得到各自的奇异值矩阵Sd和Sa。4. Singular value decomposition. Because the frequency components of the heart sound signal are mainly concentrated below 300 Hz, it is necessary to perform singular value decomposition on cd6 and ca6, and substitute them into the following formula. The respective singular value matrices Sd and Sa are obtained respectively.

五、计算奇异熵。把Sd和Sa分别代入下式得到6阶分解的轮廓信号ca6的奇异熵Ha和6阶分解的细节信号cd6的奇异熵Hd。5. Calculate the singular entropy. Substitute Sd and Sa into the following formula respectively to obtain the singular entropy Ha of the contour signal ca6 decomposed by the 6th order and the singular entropy Hd of the detail signal cd6 decomposed by the 6th order.

其中，。in , .

六、提取特征值。不按加权把轮廓信号的奇异熵Ha和细节信号的奇异熵Hd进行合并，得到最终特征值。6. Extract eigenvalues. The singular entropy Ha of the contour signal and the singular entropy Hd of the detail signal are combined without weighting to obtain the final eigenvalue.

1. a kind of heart sound feature extracting method based on wavelet singular entropy, it is characterised in that this method includes following 6 processes：

A) heart sound gathers, and cardiechema signals sample frequency is 4000Hz；

b）Heart sound envelope extraction and segmentation, the heart sound obtained in actual acquisition is long therefore needs to carry heart sound progress envelope Take, the envelope extraction method used herein is to utilize Hilbert-Huang（HHT）Heart sound envelope is extracted, is then based on the heart again Sound envelope is segmented so as to obtain one section of complete cardiechema signals, both including s1, s2 and s3, s4；

c）Wavelet transformation, 5 rank wavelet decompositions are carried out to segmentation cardiechema signals as morther wavelet using DB6；

d）Singular value decomposition, because the frequency content of cardiechema signals is concentrated mainly on 300 below Hz, need herein to small The contour signal CA obtained after wave conversion（Low-frequency information）With detail signal CD（High-frequency information）Singular value decomposition is all carried out, respectively Respective singular value matrix Sa and Sd is obtained, singular value decomposition form is shown in formula 1；

（1）

Wherein, andFor matrix A Whole non-zero singular values；

e）Singular entropy is calculated, Sd and Sa is substituted into formula 2 respectively obtains contour signal CA singular entropy Ha and detail signal CD Singular entropy Hd；

（2）

Wherein,；

f）Characteristic value is calculated, contour signal CA singular entropy Ha and detail signal CD singular entropy Hd are merged by weighting, Obtain final characteristic value.

A kind of 2. described heart sound feature extracting method based on wavelet singular entropy before being required according to right 1, it is characterised in that：The heart Sound characteristic value is using heart sound wavelet decomposition rear profile signal and the singular entropy of detail signal, and the shape by both according to weighting Formula combines.

• 2014
  • 2014-01-14 CN CN201410015681.9A patent/CN104778342B/en not_active Expired - Fee Related

Granted publication date: 20171215

Termination date: 20200114

RetroSearch is an open source project built by @garambo | Open a GitHub Issue

Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo

HTML: 3.2 | Encoding: UTF-8 | Version: 0.7.4