WO2018149147A1 - 一种呼吸率提取方法及装置 - Google Patents

一种呼吸率提取方法及装置 Download PDF

Info

Publication number
WO2018149147A1
WO2018149147A1 PCT/CN2017/104671 CN2017104671W WO2018149147A1 WO 2018149147 A1 WO2018149147 A1 WO 2018149147A1 CN 2017104671 W CN2017104671 W CN 2017104671W WO 2018149147 A1 WO2018149147 A1 WO 2018149147A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
respiratory
rate
weighting factor
current time
Prior art date
Application number
PCT/CN2017/104671
Other languages
English (en)
French (fr)
Inventor
胡静
Original Assignee
广州视源电子科技股份有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 广州视源电子科技股份有限公司 filed Critical 广州视源电子科技股份有限公司
Publication of WO2018149147A1 publication Critical patent/WO2018149147A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis

Definitions

  • the invention relates to the field of respiratory detection, and in particular to a respiratory rate extraction method and device.
  • Breathing is an important physiological process of the human body.
  • the monitoring of human breathing is also an important part of modern medical monitoring technology.
  • Patients with lesions of the respiratory system itself or other important organ lesions will affect the respiratory center to a certain extent.
  • the failure of multiple organ systems often involves the failure of respiratory function, and the failure of respiratory function leads to the failure of other organs, causing each other.
  • the prior art mainly uses the following methods to detect respiratory motion: impedance volume method: measuring the change of chest impedance with a high frequency constant current source to extract respiratory information; sensor method: using temperature, pressure, humidity and airflow sensor as a nostril sensor; capacitance method: when When breathing, the capacitance value changes accordingly; the breath sound method: the breath is recognized by picking up the breath sound; the ultrasonic method: the Doppler phenomenon is generated by the ultrasonic wave, and the respiratory frequency is detected.
  • impedance volume method measuring the change of chest impedance with a high frequency constant current source to extract respiratory information
  • sensor method using temperature, pressure, humidity and airflow sensor as a nostril sensor
  • capacitance method when When breathing, the capacitance value changes accordingly
  • the breath sound method the breath is recognized by picking up the breath sound
  • the ultrasonic method the Doppler phenomenon is generated by the ultrasonic wave, and the respiratory frequency is detected.
  • the use of these methods not only requires the addition of signal acquisition components, but also the effects of motion and the
  • the QRS wave refers to the largest amplitude group in the normal ECG, reflecting the whole process of ventricular depolarization.
  • the normal ventricular depolarization begins in the middle of the interventricular septum and depolarizes from the left to the right, so the QRS complex first presents a small downward q wave.
  • the normal chest lead QRS complex is more constant.
  • Extracting respiratory signals from ECG signals is a respiratory signal detection technology that does not require dedicated sensors and hardware modules to detect respiratory signals. It only needs to use ECG monitors to obtain ECG signals.
  • ECG-Derived Respiration, EDR is a respiratory signal detection technology that does not require dedicated sensors and hardware modules to detect respiratory signals. It only needs to use ECG monitors to obtain ECG signals.
  • the above two detection methods are bound to the human body, making dynamic breathing detection possible.
  • the existing technique for extracting respiratory signals from ECG signals mainly uses a waveform method in the calculation.
  • the method determines the current respiratory wave to rise or fall through the average value of the waveform (ie, the baseline value) over a period of time.
  • the method of extremum finds the peaks and troughs of the waveform.
  • the effective peak or trough is determined according to a certain threshold condition, and the waveform period is calculated according to the period of the effective peak or trough, thereby obtaining the respiration rate.
  • this algorithm has the advantages of relatively intuitive and small computational complexity, the respiratory waveform acquired in the actual process is more or less affected by ECG activity.
  • the waveform has a baseline drift, the calculated baseline value cannot be updated very quickly. , the waveform will be missed and the respiratory rate will be low, and the result will be greatly deviated.
  • an object of the present invention is to provide a respiratory rate extraction method and apparatus, which can accurately and reliably measure the respiratory rate, and can reduce measurement fluctuations or errors caused by external or environmental interference.
  • the invention provides a breathing rate extraction method, comprising:
  • the respiratory rate at the current time is calculated according to the first respiratory rate, the first weighting factor, the second respiratory rate, and the second weighting factor.
  • the method before performing the wavelet transform on the ECG signal to obtain the second respiratory signal, and calculating the second respiratory rate at the current time according to the second respiratory signal, the method further includes:
  • the ECG signal is downsampled.
  • the extracting the ECG signal by using an autoregressive model to obtain a first respiratory signal, and calculating the first respiratory rate according to the first respiratory signal according to the first respiratory signal is specifically:
  • an autocorrelation separation algorithm is used to extract the ECG signal, and the first respiratory signal is extracted;
  • the method further includes:
  • the representation of the observation value at each moment is optimized by using the moving average model whose model order is q, and the observed value at each moment is obtained, where q is the moving average number of items.
  • the wavelet transform is performed on the ECG signal to obtain a second respiratory signal
  • the second respiratory rate calculated at the current time according to the second respiratory signal is specifically:
  • the frequency band is layered, and the frequency range of each layer is calculated;
  • the calculating, according to the first respiratory rate, the first weighting factor, the second breathing rate, and the second weighting factor, a respiratory rate at the current time specifically:
  • the first respiration rate is set to the respiration rate at the current time
  • the first breathing rate and the second breathing rate are compared according to the first weighting factor and the second weighting factor A weighted summation is performed to calculate the respiratory rate at the current time.
  • the present invention also provides a respiratory rate extraction device, the device comprising:
  • An autoregressive extraction unit configured to extract an ECG signal by using an autoregressive model, obtain a first respiration signal, and calculate a first respiration rate at a current time according to the first respiration signal;
  • a wavelet transform extracting unit configured to perform wavelet transform on the electrocardiographic signal to obtain a second respiratory signal, and calculate a second respiratory rate at the current time according to the second respiratory signal
  • a weight calculation unit configured to perform signal quality index analysis on the first respiratory signal and the second respiratory signal, to obtain a first weighting factor corresponding to the first respiratory signal and corresponding to the second respiratory signal Second weighting factor;
  • the respiratory rate calculation unit is configured to calculate a respiratory rate at the current time according to the first respiratory rate, the first weighting factor, the second respiratory rate, and the second weighting factor.
  • the autoregressive extraction unit specifically includes:
  • An autoregressive building block is configured to obtain an observation value at each moment according to the observed values of the p historical moments before the moment and the random interference at each moment of the collected ECG signals;
  • a feature calculation module configured to generate a coefficient matrix according to p weighting parameters corresponding to the observation values of the p historical moments, to obtain a feature of the respiratory signal at each moment;
  • a signal extraction module configured to extract an ECG signal by using an autocorrelation separation algorithm according to the characteristics of the respiratory signal at each moment, and extract a first respiratory signal
  • the first respiratory rate calculation module is configured to calculate a first respiratory rate of the current time according to the first respiratory signal.
  • the wavelet transform extracting unit specifically includes:
  • Band stratification module for sampling according to Shannon-Nyquist sampling principle and said ECG signal The rate is stratified, and the frequency range of each layer is calculated;
  • a layer number determining module configured to determine a number of layers required for wavelet decomposition and reconstruction according to a frequency range and a passband frequency of each layer of the frequency band layer;
  • a signal decomposition module configured to perform signal decomposition according to a number of layers required for the wavelet decomposition and a pre-selected mother wavelet, to obtain a multi-layer waveform divided by frequency bands;
  • a signal reconstruction module configured to perform signal reconstruction according to a coefficient corresponding to the number of layers required for the wavelet reconstruction and the multi-layer waveform obtained by the decomposition, to obtain a second respiratory signal
  • the second respiratory rate calculation module is configured to calculate a second respiratory rate at the current time according to the second respiratory signal.
  • the weight calculation unit specifically includes:
  • a first determining module configured to: when determining that the first weighting factor is greater than a preset reference value and the second weighting factor is less than the reference value, setting the first breathing rate to a breathing rate of a current time;
  • a second determining module configured to: when determining that the first weighting factor is less than a preset reference value and the second weighting factor is greater than the reference value, set the second breathing rate to a breathing rate of a current time;
  • a weighting calculation module configured to: when determining that the second weighting factor is greater than a preset reference value by the first weighting factor, pairing the first breathing according to the first weighting factor and the second weighting factor The rate and the second respiration rate are weighted and summed to calculate the respiration rate at the current time.
  • the respiratory rate extraction method and device provided by the present invention the first respiratory rate and the second respiratory rate are obtained by processing the electrocardiographic signal by using an autoregressive time series technique and a wavelet transform technique, and according to the first respiratory rate
  • the corresponding first weighting factor and the second weighting factor corresponding to the second breathing rate obtain the breathing rate at the current time, and the calculation result is more accurate and reliable than the existing scheme of obtaining the breathing signal from the ECG signal by a single technique. It can reduce measurement fluctuations or errors caused by external or environmental interference, so that more accurate and stable measurement results can be obtained.
  • FIG. 1 is a schematic flow chart of a respiratory rate extraction method according to an embodiment of the present invention.
  • FIG 2 is an original waveform diagram of an electrocardiographic signal provided by an embodiment of the present invention.
  • FIG. 3 is a waveform diagram of an ECG signal after a power frequency notch according to an embodiment of the present invention.
  • FIG. 4 is a waveform diagram of a first respiratory signal extracted by an autoregressive model according to an embodiment of the present invention.
  • FIG. 5 is a waveform diagram of a second respiratory signal extracted by wavelet transform according to an embodiment of the present invention.
  • FIG. 6 is a schematic structural diagram of a respiratory rate extraction device according to an embodiment of the present invention.
  • an embodiment of the present invention provides a respiratory rate extraction method, including the following steps:
  • the original ECG signal can be detected by an electrocardiograph or a related electrocardiograph, wherein the original heart directly collected by the electrocardiograph or the related electrocardiograph is obtained.
  • the electrical signal contains a large number of power frequency interference, and a 50 Hz power frequency notch is needed to filter out the power frequency interference.
  • the ECG signal after filtering the power frequency interference is shown in Figure 3.
  • an autoregressive model is a statistical method for processing a time series, and predicts the performance of the variable in the current period by using the performance of each period before the same variable, and Suppose they are linear.
  • step S101 may include:
  • S1011 Obtain an observation value at each moment according to the observed values of the p historical moments before the moment and the random interference of each moment of the collected ECG signals.
  • the current time observation value is y t
  • the current time random interference is a t
  • the p-th historical time observation value is y tp
  • ⁇ (B) 1- ⁇ 1 B-...- ⁇ p B p
  • ⁇ p is the weighting parameter of the pth historical time
  • p The order of the model, representing the number of autoregressive terms.
  • S1012 Generate a coefficient matrix according to p weighting parameters corresponding to the observed values of the p historical moments, and obtain a characteristic of the respiratory signal at each moment.
  • FIG. 4 it is a waveform diagram of the extracted first respiratory signal.
  • the first respiratory rate R1 can be calculated, specifically:
  • the peak (or trough) of the first respiratory signal is found in the waveform diagram of the first respiratory signal by the extremum method, see the dot mark in FIG.
  • the period T1 of the current time is obtained by extracting the time interval between the two newly generated peaks.
  • the sampling rate conversion of the period can obtain the first respiratory rate R1 at the current time.
  • R1 60/T1.
  • S102 Perform wavelet transformation on the ECG signal to obtain a second respiratory signal, and calculate a second respiratory rate at the current time according to the second respiratory signal.
  • S1021 Perform frequency band stratification according to the Shannon-Nyquist sampling principle and the sampling frequency of the to-be-processed ECG signal, and calculate a frequency range of each layer.
  • the sampling frequency of the ECG signal is fs
  • the target frequency band is f1-f2 (Hz)
  • the number of layers decomposed by the wavelet transform is N, which is known by the Nyquist law.
  • the number of layers of wavelets to be reconstructed is N2 to N1 layers.
  • S1022. Determine a number of layers required for wavelet decomposition and reconstruction according to a frequency range of each layer in the frequency band layer and a preset passband frequency.
  • the sampling frequency of the directly collected ECG signal is high (typically 500 Hz)
  • the efficiency of the wavelet transform is affected. Therefore, the downsampling can be performed before the wavelet transform. Assuming that the ECG signal is downsampled to 100 Hz, fs is 100 Hz, and the highest frequency of the signal is 50 Hz. According to formulas (2), (3), (4), the corresponding frequency bands of each layer are as follows:
  • Frequency band Frequency range / Hz Frequency band Frequency range / Hz A1 0 to 25 D1 25 ⁇ 50 A2 1 ⁇ 12.5 D2 12.5 ⁇ 25 A3 0 ⁇ 6.25 D3 6.25 ⁇ 12.5 A4 0 ⁇ 3.125 D4 3.125 ⁇ 6.25 A5 0 ⁇ 1.625 D5 1.625 ⁇ 3.125 A6 0 ⁇ 0.8125 D6 0.8125 ⁇ 1.625 A7 0 ⁇ 0.40625 D7 0.40625 ⁇ 0.8125 A8 0 ⁇ 0.203125 D8 0.203125 ⁇ 0.40625 A9 0 ⁇ 0.10156 D9 0.10156 ⁇ 0.203125
  • the frequency range of the respiratory signal is usually 0.1 to 0.4 Hz
  • the frequency band is extended to 0.1 to 0.8 Hz in consideration of the shortness of breath
  • the frequency range of the electrocardiogram is 0.9 to 6 Hz, so that the respiratory signal and the heart can be well separated.
  • the electrical signal therefore, uses the approximation coefficients (D9/D8/D7) of layers 9, 8, and 7 to reconstruct the signal.
  • S1023 Perform signal decomposition according to the number of layers required for the wavelet decomposition and the pre-selected mother wavelet to obtain a multi-layer waveform divided by frequency bands.
  • the embodiment of the present invention uses the coif3 wavelet base as the mother wavelet for wavelet decomposition.
  • other mother wavelets such as db wavelets, may also be selected, which are not specifically limited in the present invention.
  • an N-layer waveform when decomposing, an N-layer waveform can be obtained. At this time, a waveform corresponding to the N2 to N1 layer can be extracted for reconstruction.
  • the number of layers required for wavelet reconstruction is the 9th, 8th, and 7th layers.
  • the wavelet coefficients corresponding to the number of layers required for the wavelet reconstruction can be calculated (by calculating the ECG signal and The product of the wavelet base is obtained and the multi-layer waveform obtained by the decomposition is subjected to signal reconstruction to obtain a second respiratory signal.
  • FIG. 5 a waveform diagram of a second respiratory signal extracted by wavelet transform according to an embodiment of the present invention.
  • the second respiratory rate R2 can be calculated, specifically:
  • the peak (or trough) of the second respiratory signal is found in the waveform diagram of the second respiratory signal by the extremum method, see the dot mark in FIG.
  • the period T2 is obtained by extracting the time interval between the two most recently generated peaks.
  • the real-time second respiratory rate R2 can be obtained.
  • power spectrum analysis may be performed on the first respiratory signal and the second respiratory signal, and a spectral distribution of the first respiratory signal and the second respiratory signal may be analyzed to obtain a first weighting factor corresponding to a breathing signal and a second weighting factor corresponding to the second breathing signal.
  • a first weighting factor corresponding to the first respiratory signal and a second corresponding to the second respiratory signal may also be obtained by performing peak spectrum analysis on the first respiratory signal and the second respiratory signal.
  • the weighting factor is not specifically limited in the present invention.
  • the respiratory rate R at the current time can be calculated by weighting the first respiratory rate R1 and the second respiratory rate R2.
  • ⁇ 1 is the first weighting factor and ⁇ 2 is the second weighting factor.
  • the respiratory rate extraction method processes the ECG signal by using a time series technique of an autoregressive model and a wavelet transform technique to obtain a first respiratory rate and a second respiratory rate, and according to the first respiratory rate
  • the corresponding first weighting factor and the second weighting factor corresponding to the second breathing rate obtain the breathing rate at the current time, and the calculation result is more accurate and reliable than the existing scheme of obtaining the breathing signal from the ECG signal by a single technique. It can reduce measurement fluctuations or errors caused by external or environmental interference, so that more accurate and stable measurement results can be obtained.
  • the autoregressive model may be optimized by using a moving average model when extracting respiratory signals using an autoregressive model. specifically:
  • the system matrix should include weighting parameters of random interference in addition to weighting parameters of observations at historical moments, and the present invention I will not repeat them here.
  • the autoregressive model is optimized based on a moving average model to minimize the residual of the model, thereby achieving noise reduction, especially the effect of reducing white noise.
  • the step S104 may also be:
  • the signal quality of the second breathing signal may be considered to be poor.
  • the first breathing rate R1 is directly set to the breathing rate R of the current time.
  • the signal quality of the first breathing signal may be considered to be poor.
  • the first breathing rate R1 is directly set to the breathing rate R of the current time.
  • the first breathing rate and the second weighting factor are compared according to the first weighting factor and the second weighting factor.
  • the respiration rate is weighted and summed to calculate the respiration rate at the current time.
  • the weighting factor is small, it indicates that the corresponding respiratory signal quality is poor, and the respiratory rate corresponding to the poor quality respiratory signal is directly removed to ensure the accuracy and stability of the calculation result.
  • an embodiment of the present invention further provides a respiratory rate extraction apparatus 100, including:
  • the autoregressive extraction unit 10 is configured to extract the collected electrocardiographic signal by using an autoregressive model to obtain a first respiration signal, and calculate a first respiration rate at the current moment according to the first respiration signal.
  • the autoregressive extraction unit 10 specifically includes:
  • the autoregressive construction module 11 is configured to obtain an observation value at each moment according to the observed values of the p historical moments before the moment and the random interference at each moment of the collected ECG signals;
  • the feature calculation module 12 is configured to generate a coefficient matrix according to p weighting parameters corresponding to the observed values of the p historical moments, to obtain a feature of the respiratory signal;
  • the signal extraction module 13 is configured to combine the obtained characteristics of the respiratory signal, and adopt an autocorrelation separation algorithm to extract the ECG signal and extract the first respiratory signal;
  • the first respiratory rate calculation module 14 is configured to calculate a first respiratory rate of the current time according to the first respiratory signal.
  • the wavelet transform extracting unit 20 is configured to perform wavelet transform on the electrocardiographic signal to obtain a second respiratory signal, and calculate a second respiratory rate at the current time according to the second respiratory signal.
  • the wavelet transform extracting unit 20 specifically includes:
  • the frequency band layering module 21 is configured to perform frequency band stratification according to the Shannon-Nyquist sampling principle and the sampling frequency of the ECG signal, and calculate a frequency range of each layer;
  • a layer number determining module 22 configured to determine a number of layers required for wavelet decomposition and reconstruction according to a frequency range and a passband frequency of each layer of the frequency band layer;
  • the signal decomposition module 23 is configured to perform signal decomposition according to the number of layers required for the wavelet decomposition and the pre-selected mother wavelet to obtain a multi-layer waveform divided by frequency bands;
  • the signal reconstruction module 24 is configured to perform signal reconstruction according to the coefficient corresponding to the number of layers required for the wavelet reconstruction and the multi-layer waveform obtained by the decomposition, to obtain a second respiratory signal;
  • the second respiratory rate calculation module 25 is configured to calculate a second respiratory rate at the current time according to the second respiratory signal.
  • the weight calculation unit 30 is configured to perform signal quality index analysis on the first respiratory signal and the second respiratory signal to obtain a first weighting factor corresponding to the first respiratory signal and corresponding to the second respiratory signal Second weighting factor;
  • the respiratory rate calculation unit 40 is configured to calculate a respiratory rate at the current time according to the first respiratory rate, the first weighting factor, the second respiratory rate, and the second weighting factor.
  • the auto-regressive extraction unit 10 further includes a moving average optimization module 15 for optimizing the random interference at each moment using a moving average model with a model order q, where q is the moving average number of items.
  • the weight calculation unit 40 specifically includes:
  • the first determining module 41 is configured to: when determining that the first weighting factor is greater than a preset reference value, When the second weighting factor is less than the reference value, setting the first breathing rate to a breathing rate at a current time;
  • the second determining module 42 is configured to: when determining that the first weighting factor is less than a preset reference value and the second weighting factor is greater than the reference value, set the second breathing rate to a current breathing rate ;
  • the weighting calculation module 43 is configured to: when it is determined that the second weighting factor of the first weighting factor is greater than a preset reference value, pair the first according to the first weighting factor and the second weighting factor The respiration rate and the second respiration rate are weighted and summed, and the respiration rate at the current time is calculated.
  • the respiratory rate extraction device 100 processes the electrocardiographic signal by using an autoregressive time series technique and a wavelet transform technique to obtain a first respiration rate and a second respiration rate, and according to the first respiration rate.
  • the first weighting factor and the second weighting factor corresponding to the second breathing rate obtain the breathing rate at the current time, and the calculation result is more accurate and reliable than the existing scheme of obtaining the breathing signal from the ECG signal by a single technique. It can reduce measurement fluctuations or errors caused by external or environmental interference, so that more accurate and stable measurement results can be obtained.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Pathology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Physiology (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种呼吸率提取方法及装置，方法包括：利用自回归模型对接收的心电信号进行提取，得到第一呼吸信号，并根据第一呼吸信号计算得到当前时刻的第一呼吸率（S101）；对心电信号进行小波变换得到第二呼吸信号，并根据第二呼吸信号计算得到当前时刻的第二呼吸率（S102）；对第一呼吸信号和所第二呼吸信号进行信号质量指数分析，得到与第一呼吸信号对应的第一权重因子和与第二呼吸信号对应的第二权重因子（S103）；根据第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率（S104）。能够便捷有效提取呼吸信号，从而计算得到准确稳定的呼吸率。

Description

一种呼吸率提取方法及装置 技术领域
本发明涉及呼吸检测领域，尤其涉及一种呼吸率提取方法及装置。
背景技术
呼吸是人体重要的生理过程，对人体呼吸的监护检测也是现代医学监护技术的一个重要组成部分。患者不论是呼吸系统本身的病变或是其他重要脏器的病变发展到一定程度都会影响呼吸中枢。多脏器系统功能衰竭往往累及呼吸功能的衰竭，呼吸功能的衰竭又导致其他脏器功能的衰竭，互为因果。
现有技术对呼吸运动主要使用下列方法检测：阻抗容积法：用高频恒流源测量胸部阻抗的变化来提取呼吸信息；传感器法：使用温度、压力、湿度和气流传感器作为鼻孔传感器；电容法：当呼吸时导致电容值产生相应的变化；呼吸音法：通过拾取呼吸音识别呼吸；超声法：利用超声波产生多谱勒现象，检测出呼吸频率。使用这些方法不但需要增加信号采集部件，而且受到运动和环境的影响，不适合用于日常监护。
大量临床资料显示，呼吸运动会引起心电图的变化。通过心电图，我们可以观察到在呼吸周期内由胸部运动和心脏位置变化所引起的心电波形峰峰值的改变。这是由于呼吸周期内，描述心脏电波主要传播方向的心脏电轴旋转造成QRS波群形态发生了变化。QRS波是指正常心电图中幅度最大的波群，反映心室除极的全过程。正常心室除极始于室间隔中部，自左向右方向除极，故QRS波群先呈现一个小向下的q波。正常胸导联QRS波群形态较恒定。从心电信号中提取呼吸信号(ECG-DerivedRespiration，EDR)是一种呼吸信号检测技术，这种技术不需要专用传感器和硬件模块检测呼吸信号，只需要用心电监护仪获取心电信号，避免了上述两种检测方法对人体的束缚，使动态呼吸检测成为可能。
但现有从心电信号中提取呼吸信号的技术，在计算时主要采用波形法，该方法通过一段时间内波形的平均值(即基线值)，来判定当前呼吸波处于上升或下降趋势，用极值的方法求得波形的波峰、波谷。根据一定的阈值条件来判定有效的波峰或波谷，再根据有效波峰或波谷的周期计算波形周期，从而得到呼吸率。这种算法虽然具有比较直观、运算量小的优点，但在实际过程中获取的呼吸波形或多或少会受到心电活动的影响，当波形出现基线漂移时，计算的基线值无法很快更新，会导致波形漏检致使呼吸率值偏低，其结果会有较大偏差。
发明内容
针对上述问题，本发明的目的在于提供一种呼吸率提取方法及装置，实现准确可靠的呼吸率的测量，并可减轻由于外界或环境的干扰而引起的测量波动或误差。
本发明提供了一种呼吸率提取方法，包括：
利用自回归模型对采集的心电信号进行提取，得到第一呼吸信号，并根据所述第一呼吸信号计算得到当前时刻的第一呼吸率；
对所述心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率；
对所述第一呼吸信号和所述第二呼吸信号进行信号质量指数分析，得到与所述第一呼吸信号对应的第一权重因子和与所述第二呼吸信号对应的第二权重因子；
根据所述第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率。
优选地，在对所述心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率之前，还包括：
对所述心电信号进行降采样。
优选地，所述利用自回归模型对采集的心电信号进行提取，得到第一呼吸信号，并根据所述第一呼吸信号计算得到当前时刻的第一呼吸率具体为：
根据采集的心电信号的位于每个时刻之前的p个历史时刻的观测值和每个时刻的随机干扰，得到每个时刻的观测值；
根据与p个历史时刻的观测值对应的p个加权参数，生成系数矩阵，得到每个时刻的呼吸信号的特征；
根据所述每个时刻的呼吸信号的特征，采用自相关分离算法对心电信号进行提取，提取得到第一呼吸信号；
根据所述第一呼吸信号计算得到当前时刻的第一呼吸率。
优选地，在依次将心电信号的每个时刻的观测值用位于该时刻之前的p个历史时刻的观测值和一个该时刻的随机干扰进行表示之后，在根据与p个历史时刻的观测值对应的p个加权参数，生成系数矩阵，得到每个时刻的呼吸信号的特征之前，还包括：
利用模型阶数为q的滑动平均模型对每个时刻的观测值的表示进行优化，得到优化后的每个时刻的观测值，其中，q为滑动平均项数。
优选地，所述对所述心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率具体为：
根据香农-奈奎斯特采样原理及所述心电信号的采样频率进行频段分层，计算得到每层的频率范围；
依据所述频段分层中每层的频率范围及预置的通带频率确定小波分解和重构所需的层数；
根据与所述小波分解所需的层数及预先选择的母小波进行信号分解，得到按频段划分的多层波形；
根据与所述小波重构所需的层数对应的小波系数及分解得到的所述多层波形进行信号重构，得到第二呼吸信号；
根据所述第二呼吸信号计算得到当前时刻的第二呼吸率。
优选地，所述根据所述第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率，具体为：
当判断所述第一权重因子大于预设的基准值且所述第二权重因子小于所述 基准值时，将所述第一呼吸率设置为当前时刻的呼吸率；
当判断所述第一权重因子小于预设的基准值且所述第二权重因子大于所述基准值时，将所述第二呼吸率设置为当前时刻的呼吸率；
当判断所述第一权重因子计所述第二权重因子均大于预设的基准值时，根据所述第一权重因子及所述第二权重因子对所述第一呼吸率和第二呼吸率进行加权求和，计算得到当前时刻的呼吸率。
本发明还提供了一种呼吸率提取装置，所述装置包括：
自回归提取单元，用于利用自回归模型对心电信号进行提取，得到第一呼吸信号，并根据所述第一呼吸信号计算得到当前时刻的第一呼吸率；
小波变换提取单元，用于对心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率；
权重计算单元，用于对所述第一呼吸信号和所述第二呼吸信号进行信号质量指数分析，得到与所述第一呼吸信号对应的第一权重因子和与所述第二呼吸信号对应的第二权重因子；
呼吸率计算单元，用于根据所述第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率。
优选地，所述自回归提取单元具体包括：
自回归构建模块，用于根据采集的心电信号的位于每个时刻之前的p个历史时刻的观测值和每个时刻的随机干扰，得到每个时刻的观测值；
特征计算模块，用于根据与p个历史时刻的观测值对应的p个加权参数，生成系数矩阵，得到每个时刻的呼吸信号的特征；
信号提取模块，用于根据所述每个时刻的呼吸信号的特征，采用自相关分离算法对心电信号进行提取，提取得到第一呼吸信号；
第一呼吸率计算模块，用于根据所述第一呼吸信号计算得到当前时刻的第一呼吸率。
优选地，所述小波变换提取单元具体包括：
频段分层模块，用于根据香农-奈奎斯特采样原理及所述心电信号的采样频 率进行频段分层，计算得到每层的频率范围；
层数确定模块，用于依据所述频段分层每层的频率范围及通带频率确定小波分解和重构所需的层数；
信号分解模块，用于根据与所述小波分解所需的层数及预先选择的母小波进行信号分解，得到按频段划分的多层波形；
信号重构模块，用于根据与所述小波重构所需的层数对应的系数及分解得到的所述多层波形进行信号重构，得到第二呼吸信号；
第二呼吸率计算模块，用于根据所述第二呼吸信号计算得到当前时刻的第二呼吸率。
优选地，所述权重计算单元具体包括：
第一判断模块，用于当判断所述第一权重因子大于预设的基准值且所述第二权重因子小于所述基准值时，将所述第一呼吸率设置为当前时刻的呼吸率；
第二判断模块，用于当判断所述第一权重因子小于预设的基准值且所述第二权重因子大于所述基准值时，将所述第二呼吸率设置为当前时刻的呼吸率；
加权计算模块，用于当判断所述第一权重因子计所述第二权重因子均大于预设的基准值时，根据所述第一权重因子及所述第二权重因子对所述第一呼吸率和第二呼吸率进行加权求和，计算得到当前时刻的呼吸率。
本发明提供的呼吸率提取方法及装置，通过利用自回归时间序列技术与小波变换技术相结合的方式处理心电信号得到第一呼吸率及第二呼吸率，并根据与所述第一呼吸率对应的第一权重因子和与所述第二呼吸率对应的第二权重因子得到当前时刻的呼吸率，相比于现有由单一技术从心电信号得到呼吸信号的方案，计算结果更准确可靠，并可减轻由于外界或环境的干扰而引起的测量波动或误差，从而能够得到更为准确稳定的测量结果。
附图说明
图1是本发明实施例提供的呼吸率提取方法的流程示意图。
图2是本发明实施例提供的心电信号的原始波形图。
图3是本发明实施例提供的经过工频陷波后的心电信号的波形图。
图4是本发明实施例提供的通过自回归模型提取得到的第一呼吸信号的波形图。
图5是本发明实施例提供的通过小波变换提取得到的第二呼吸信号的波形图。
图6是本发明实施例提供的呼吸率提取装置的结构示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
请参见图1，本发明实施例提供了一种呼吸率提取方法，包括如下步骤：
S101，利用自回归模型对采集的心电信号进行提取，得到第一呼吸信号，并根据所述第一呼吸信号计算得到当前时刻的第一呼吸率。
如图2所示，在本发明实施例中，原始的心电信号可通过心电测试仪或相关心电仪器检测得到，其中，由心电测试仪或相关心电仪器直接采集的原始的心电信号包含大量的工频干扰，需要进行50Hz工频陷波，以滤除工频干扰。其中，滤除工频干扰后的心电信号如图3所示。
在本发明实施例中，自回归模型(Autoregressive model，AR)是统计上的一种处理时间序列的方法，是用同一变量之前各期的表现情况，来预测该变量本期的表现情况，并假设它们为线性关系。
具体地，步骤S101可包括：
S1011，根据采集的心电信号的位于每个时刻之前的p个历史时刻的观测值和每个时刻的随机干扰，得到每个时刻的观测值。
在本发明实施例中，设当前时刻的观测值为yt，当前时刻的随机干扰为at， 第p个历史时刻的观测值为yt-p，则根据自回归模型可得到如下公式：
φ(B)yt＝at          (1)
其中，φ(B)＝1-φ1B-...-φpBp，B为延迟算子，满足Byt＝yt-1，φp为第p个历史时刻的加权参数，p为模型的阶数，表示自回归项数。
S1012，根据与p个历史时刻的观测值对应的p个加权参数，生成系数矩阵，得到每个时刻的呼吸信号的特征。
S1013，根据所述每个时刻的呼吸信号的特征，采用自相关分离算法对心电信号进行提取，提取得到第一呼吸信号。
如图4所示，为提取得到的第一呼吸信号的波形图。
S1014，根据所述第一呼吸信号计算得到当前时刻的第一呼吸率。
在本发明实施例中，在得到所述第一呼吸信号后，即可计算得到第一呼吸率R1，具体为：
通过求极值法在所述第一呼吸信号的波形图中寻找第一呼吸信号的波峰(或者波谷)，参见图4中的点标记。
通过提取最近生成的两个波峰之间的时间间隔，以得到当前时刻的周期T1。
对所述周期进行采样率换算即可得到当前时刻的第一呼吸率R1。
例如，R1＝60/T1。
S102，对所述心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率。
具体地：
S1021，根据香农-奈奎斯特采样原理及所述待处理心电信号的采样频率进行频段分层，计算得到每层的频率范围。
根据香农-奈奎斯特采样原理，设所述心电信号的采样频率为fs，目标频段为f1-f2(Hz)，运用小波变换分解的层数为N，由奈奎斯特定律可知，则有：
f1＝(fs/2)/2N1          (2)
f2＝(fs/2)/2N2          (3)
N＞N1(N1＞N2)          (4)
即需要重构的小波的层数为N2～N1层。
S1022，依据所述频段分层中每层的频率范围及预置的通带频率确定小波分解和重构所需的层数。
需要说明的是，由于直接采集的心电信号的采样频率较高(一般为500Hz)，会影响小波变换的效率，因此，在进行小波变换前，可先进行降采样。假设所述心电信号被降采样至100Hz，则fs为100Hz，信号最高频率为50Hz，根据公式(2)、(3)、(4)可知，每一层对应的频段如下：
频段 频率范围/Hz 频段 频率范围/Hz
A1 0～25 D1 25～50
A2 1～12.5 D2 12.5～25
A3 0～6.25 D3 6.25～12.5
A4 0～3.125 D4 3.125～6.25
A5 0～1.625 D5 1.625～3.125
A6 0～0.8125 D6 0.8125～1.625
A7 0～0.40625 D7 0.40625～0.8125
A8 0～0.203125 D8 0.203125～0.40625
A9 0～0.10156 D9 0.10156～0.203125
由于呼吸信号的频段范围通常为0.1～0.4Hz，考虑到呼吸急促的情况，将频段扩展为0.1～0.8Hz，心电的频率范围为0.9～6Hz，因此，能够很好地分离呼吸信号和心电信号，所以，选用第9、8、7层的近似系数(D9/D8/D7)来重构信号。
S1023，根据与所述小波分解所需的层数及预先选择的母小波进行信号分解，得到按频段划分的多层波形。
在本发明实施例中，经验证，coifN小波和dmey小波的提取效果较佳，且优选地，以coif3小波基作为母小波时，具有最佳的提取效果。因而本发明实施例采用coif3小波基作为母小波进行小波分解。当然，可以理解的是，在本发明 的其他实施例中，也可选取其他的母小波，如db小波等，本发明不做具体限定。
在本发明实施例中，在分解时，可得到N层波形，此时，可提取N2～N1层对应的波形来进行重构。
S1024，根据与所述小波重构所需的层数对应的系数及分解得到的所述多层波形进行信号重构，得到第二呼吸信号。
由步骤S1023可知，小波重构所需的层数为第9、8、7层，此时，即可根据与所述小波重构所需的层数对应的小波系数(通过计算心电信号与小波基的积得到)及分解得到的所述多层波形进行信号重构，得到第二呼吸信号。
参见图5，为根据本发明实施例的小波变换提取得到的第二呼吸信号的波形图。
S1025，根据所述第二呼吸信号计算得到当前时刻的第二呼吸率。
在本发明实施例中，在获得所述第二呼吸信号后，即可计算第二呼吸率R2，具体为：
通过求极值法在所述第二呼吸信号的波形图中寻找第二呼吸信号的波峰(或者波谷)，参见图5中的点标记。
通过提取最近生成的两个波峰之间的时间间隔，以得到周期T2。
根据采样率换算即可得到实时的第二呼吸率R2。
S103，对所述第一呼吸信号和所述第二呼吸信号进行信号质量指数分析，得到与所述第一呼吸信号对应的第一权重因子和与所述第二呼吸信号对应的第二权重因子。
在本发明实施例中，可对所述第一呼吸信号和所述第二呼吸信号进行功率谱分析，分析所述第一呼吸信号和所述第二呼吸信号的谱分布，得到与所述第一呼吸信号对应的第一权重因子和与所述第二呼吸信号对应的第二权重因子。
当前，也可通过对所述第一呼吸信号和所述第二呼吸信号进行峰值谱分析来得到与所述第一呼吸信号对应的第一权重因子和与所述第二呼吸信号对应的第二权重因子，本发明不做具体限定。
S104，根据所述第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率。
在本发明实施例中，可通过对第一呼吸率R1及第二呼吸率R2进行加权平均来计算得到当前时刻的呼吸率R。
即：
R＝μ1*R1+μ2*R2          (5)
其中，μ1为第一权重因子，μ2为第二权重因子。
需要说明的是，在进行加权平均之前，需先对μ1和μ2进行归一化处理，具体地，假设μ1+μ2＝a，则需要分别对μ1和μ2乘以归一化系数1/a进行归一化，保证归一化后的μ1+μ2＝1。
本发明提供的呼吸率提取方法，通过利用自回归模型的时间序列技术与小波变换技术相结合的方式处理心电信号得到第一呼吸率及第二呼吸率，并根据与所述第一呼吸率对应的第一权重因子和与所述第二呼吸率对应的第二权重因子得到当前时刻的呼吸率，相比于现有由单一技术从心电信号得到呼吸信号的方案，计算结果更准确可靠，并可减轻由于外界或环境的干扰而引起的测量波动或误差，从而能够得到更为准确稳定的测量结果。
需要说明的是，在本发明的优选实施例中，为了进一步消除心电信号中的白噪声，在利用自回归模型进行呼吸信号提取时，还可利用滑动平均模型对自回归模型进行优化。具体地：
对于滑动平均模型，当前时刻的观测值可表示为当前时刻的随机干扰与过去q个历史时刻的随机干扰，即yt＝θ(B)·at，其中，θ(B)＝1-θ1-...-θqBq，B为延迟算子，θq为过去第q个时刻的随机干扰的加权系数，q为滑动平均项数。利用所述滑动平均模型对所述自回归模型进行优化后即可得到：φ(B)yt＝θ(B)·at。
在本发明优选实施例中，在利用滑动平均模型对所述自回归模型进行优化的同时，则系统矩阵除了包括历史时刻的观测值的加权参数外，还应包括随机干扰的加权参数，本发明在此不做赘述。
在本发明优选实施例中，基于滑动平均模型对所述自回归模型进行优化，使模型的残差最小，从而可达到降噪，尤其是降低白噪声的效果。
需要说明的是，为了防止第一呼吸信号或第二呼吸信号质量较差而导致的计算结果不准确，在本发明的优选实施例中，所述步骤S104还可为：
S1041，当判断所述第一权重因子大于预设的基准值且所述第二权重因子小于所述基准值时，将所述第一呼吸率设置为当前时刻的呼吸率。
当所述第二权重因子小于所述基准值时，可以认为第二呼吸信号的信号质量较差，此时，直接将所述第一呼吸率R1设置为当前时刻的呼吸率R。
S1042，当判断所述第一权重因子小于预设的基准值且所述第二权重因子大于所述基准值时，将所述第二呼吸率设置为当前时刻的呼吸率。
当所述第一权重因子小于所述基准值时，可以认为第一呼吸信号的信号质量较差，此时，直接将所述第一呼吸率R1设置为当前时刻的呼吸率R。
S1043，当判断所述第一权重因子及所述第二权重因子均大于预设的基准值时，根据所述第一权重因子及所述第二权重因子对所述第一呼吸率和第二呼吸率进行加权求和，计算得到当前时刻的呼吸率。
即：R＝μ1*R1+μ2*R2。
本优选实施例中，如果权重因子较小，则说明对应的呼吸信号质量较差，则直接去掉与质量较差呼吸信号对应的呼吸率，保证计算结果的准确和稳定。
参阅图6，本发明实施例还提供一种呼吸率提取装置100，包括：
自回归提取单元10，用于利用自回归模型对采集的心电信号进行提取，得到第一呼吸信号，并根据所述第一呼吸信号计算得到当前时刻的第一呼吸率。
其中，所述自回归提取单元10具体包括：
自回归构建模块11，用于根据采集的心电信号的位于每个时刻之前的p个历史时刻的观测值和每个时刻的随机干扰，得到每个时刻的观测值；
特征计算模块12，用于根据与p个历史时刻的观测值对应的p个加权参数，生成系数矩阵，得到呼吸信号的特征；
信号提取模块13，用于结合得到的呼吸信号的特征，采用自相关分离算法，对心电信号进行提取，提取得到第一呼吸信号；
第一呼吸率计算模块14，用于根据所述第一呼吸信号计算得到当前时刻的第一呼吸率。
小波变换提取单元20，用于对心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率。
其中，所述小波变换提取单元20具体包括：
频段分层模块21，用于根据香农-奈奎斯特采样原理及所述心电信号的采样频率进行频段分层，计算得到每层的频率范围；
层数确定模块22，用于依据所述频段分层每层的频率范围及通带频率确定小波分解和重构所需的层数；
信号分解模块23，用于根据与所述小波分解所需的层数及预先选择的母小波进行信号分解，得到按频段划分的多层波形；
信号重构模块24，用于根据与所述小波重构所需的层数对应的系数及分解得到的所述多层波形进行信号重构，得到第二呼吸信号；
第二呼吸率计算模块25，用于根据所述第二呼吸信号计算得到当前时刻的第二呼吸率。
权重计算单元30，用于对所述第一呼吸信号和所述第二呼吸信号进行信号质量指数分析，得到与所述第一呼吸信号对应的第一权重因子和与所述第二呼吸信号对应的第二权重因子；
呼吸率计算单元40，用于根据所述第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率。
优选地，所述自回归提取单元10还包括滑动平均优化模块15，用于利用模型阶数为q的滑动平均模型对每个时刻的随机干扰进行优化，其中，q为滑动平均项数。
优选地，所述权重计算单元40具体包括：
第一判断模块41，用于当判断所述第一权重因子大于预设的基准值且所述 第二权重因子小于所述基准值时，将所述第一呼吸率设置为当前时刻的呼吸率；
第二判断模块42，用于当判断所述第一权重因子小于预设的基准值且所述第二权重因子大于所述基准值时，将所述第二呼吸率设置为当前时刻的呼吸率；
加权计算模块43，用于当判断所述第一权重因子计所述第二权重因子均大于预设的基准值时，根据所述第一权重因子及所述第二权重因子对所述第一呼吸率和第二呼吸率进行加权求和，计算得到当前时刻的呼吸率。
本发明提供的呼吸率提取装置100，通过利用自回归时间序列技术与小波变换技术相结合的方式处理心电信号得到第一呼吸率及第二呼吸率，并根据与所述第一呼吸率对应的第一权重因子和与所述第二呼吸率对应的第二权重因子得到当前时刻的呼吸率，相比于现有由单一技术从心电信号得到呼吸信号的方案，计算结果更准确可靠，并可减轻由于外界或环境的干扰而引起的测量波动或误差，从而能够得到更为准确稳定的测量结果。
以上所揭露的仅为本发明一种较佳实施例而已，当然不能以此来限定本发明之权利范围，本领域普通技术人员可以理解实现上述实施例的全部或部分流程，并依本发明权利要求所作的等同变化，仍属于发明所涵盖的范围。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的程序可存储于一计算机可读取存储介质中，该程序在执行时，可包括如上述各方法的实施例的流程。其中，所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory，ROM)或随机存储记忆体(Random Access Memory，RAM)等。

Claims (10)

  1. 一种呼吸率提取方法，其特征在于，包括：
    利用自回归模型对采集的心电信号进行提取，得到第一呼吸信号，并根据所述第一呼吸信号计算得到当前时刻的第一呼吸率；
    对所述心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率；
    对所述第一呼吸信号和所述第二呼吸信号进行信号质量指数分析，得到与所述第一呼吸信号对应的第一权重因子和与所述第二呼吸信号对应的第二权重因子；
    根据所述第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率。
  2. 根据权利要求1所述的呼吸率提取方法，其特征在于，在对所述心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率之前，还包括：
    对所述心电信号进行降采样。
  3. 根据权利要求1所述的呼吸率提取方法，其特征在于，所述利用自回归模型对采集的心电信号进行提取，得到第一呼吸信号，并根据所述第一呼吸信号计算得到当前时刻的第一呼吸率具体为：
    根据采集的心电信号的位于每个时刻之前的p个历史时刻的观测值和每个时刻的随机干扰，得到每个时刻的观测值；
    根据与p个历史时刻的观测值对应的p个加权参数，生成系数矩阵，得到每个时刻的呼吸信号的特征；
    根据所述每个时刻的呼吸信号的特征，采用自相关分离算法对心电信号进行提取，提取得到第一呼吸信号；
    根据所述第一呼吸信号计算得到当前时刻的第一呼吸率。
  4. 根据权利要求3所述的呼吸率提取方法，其特征在于，在依次将心电信号的每个时刻的观测值用位于该时刻之前的p个历史时刻的观测值和一个该时刻的随机干扰进行表示之后，在根据与p个历史时刻的观测值对应的p个加权参数，生成系数矩阵，得到每个时刻的呼吸信号的特征之前，还包括：
    利用模型阶数为q的滑动平均模型对每个时刻的观测值的表示进行优化，得到优化后的每个时刻的观测值，其中，q为滑动平均项数。
  5. 根据权利要求1所述的呼吸率提取方法，其特征在于，所述对所述心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率具体为：
    根据香农-奈奎斯特采样原理及所述心电信号的采样频率进行频段分层，计算得到每层的频率范围；
    依据所述频段分层中每层的频率范围及预置的通带频率确定小波分解和重构所需的层数；
    根据与所述小波分解所需的层数及预先选择的母小波进行信号分解，得到按频段划分的多层波形；
    根据与所述小波重构所需的层数对应的小波系数及分解得到的所述多层波形进行信号重构，得到第二呼吸信号；
    根据所述第二呼吸信号计算得到当前时刻的第二呼吸率。
  6. 根据权利要求1所述的呼吸率提取方法，其特征在于，所述根据所述第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率，具体为：
    当判断所述第一权重因子大于预设的基准值且所述第二权重因子小于所述基准值时，将所述第一呼吸率设置为当前时刻的呼吸率；
    当判断所述第一权重因子小于预设的基准值且所述第二权重因子大于所述 基准值时，将所述第二呼吸率设置为当前时刻的呼吸率；
    当判断所述第一权重因子计所述第二权重因子均大于预设的基准值时，根据所述第一权重因子及所述第二权重因子对所述第一呼吸率和第二呼吸率进行加权求和，计算得到当前时刻的呼吸率。
  7. 一种呼吸率提取装置，其特征在于，所述装置包括：
    自回归提取单元，用于利用自回归模型对采集的心电信号进行提取，得到第一呼吸信号，并根据所述第一呼吸信号计算得到当前时刻的第一呼吸率；
    小波变换提取单元，用于对心电信号进行小波变换得到第二呼吸信号，并根据所述第二呼吸信号计算得到当前时刻的第二呼吸率；
    权重计算单元，用于对所述第一呼吸信号和所述第二呼吸信号进行信号质量指数分析，得到与所述第一呼吸信号对应的第一权重因子和与所述第二呼吸信号对应的第二权重因子；
    呼吸率计算单元，用于根据所述第一呼吸率、第一权重因子、第二呼吸率及第二权重因子，计算得到当前时刻的呼吸率。
  8. 根据权利要求7所述的呼吸率提取装置，其特征在于，所述自回归提取单元具体包括：
    自回归构建模块，用于根据采集的心电信号的位于每个时刻之前的p个历史时刻的观测值和每个时刻的随机干扰，得到每个时刻的观测值；
    特征计算模块，用于根据与p个历史时刻的观测值对应的p个加权参数，生成系数矩阵，得到每个时刻的呼吸信号的特征；
    信号提取模块，用于根据所述每个时刻的呼吸信号的特征，采用自相关分离算法对心电信号进行提取，提取得到第一呼吸信号；
    第一呼吸率计算模块，用根据所述第一呼吸信号计算得到当前时刻的第一呼吸率。
  9. 根据权利要求7所述的呼吸率提取装置，其特征在于，所述小波变换提取单元具体包括：
    频段分层模块，用于根据香农-奈奎斯特采样原理及所述心电信号的采样频率进行频段分层，计算得到每层的频率范围；
    层数确定模块，用于依据所述频段分层每层的频率范围及通带频率确定小波分解和重构所需的层数；
    信号分解模块，用于根据与所述小波分解所需的层数及预先选择的母小波进行信号分解，得到按频段划分的多层波形；
    信号重构模块，用于根据与所述小波重构所需的层数对应的系数及分解得到的所述多层波形进行信号重构，得到第二呼吸信号；
    第二呼吸率计算模块，用于根据所述第二呼吸信号计算得到当前时刻的第二呼吸率。
  10. 根据权利要求8所述的呼吸率提取装置，其特征在于，所述呼吸率计算单元具体包括：
    第一判断模块，用于当判断所述第一权重因子大于预设的基准值且所述第二权重因子小于所述基准值时，将所述第一呼吸率设置为当前时刻的呼吸率；
    第二判断模块，用于当判断所述第一权重因子小于预设的基准值且所述第二权重因子大于所述基准值时，将所述第二呼吸率设置为当前时刻的呼吸率；
    加权计算模块，用于当判断所述第一权重因子计所述第二权重因子均大于预设的基准值时，根据所述第一权重因子及所述第二权重因子对所述第一呼吸率和第二呼吸率进行加权求和，计算得到当前时刻的呼吸率。
PCT/CN2017/104671 2017-02-20 2017-09-29 一种呼吸率提取方法及装置 WO2018149147A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710090553.4A CN106901694A (zh) 2017-02-20 2017-02-20 一种呼吸率提取方法及装置
CN201710090553.4 2017-02-20

Publications (1)

Publication Number Publication Date
WO2018149147A1 true WO2018149147A1 (zh) 2018-08-23

Family

ID=59209241

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/104671 WO2018149147A1 (zh) 2017-02-20 2017-09-29 一种呼吸率提取方法及装置

Country Status (2)

Country Link
CN (1) CN106901694A (zh)
WO (1) WO2018149147A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113288111A (zh) * 2021-05-10 2021-08-24 厦门理工学院 一种基于呼吸率的疲劳预测方法、终端设备及存储介质

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106901694A (zh) * 2017-02-20 2017-06-30 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN108309259A (zh) * 2018-02-02 2018-07-24 中国建设银行股份有限公司山西省分行 服务推送方法、服务器及可测生理参数的银行终端系统
US20220280063A1 (en) * 2019-08-23 2022-09-08 Data Solutions, Inc. Respiration detection system and respiration detection method
CN113143228A (zh) * 2021-04-30 2021-07-23 中科院计算所泛在智能研究院 一种应用于压电传感器信号的心率呼吸率提取方法

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101528126A (zh) * 2006-12-21 2009-09-09 弗雷森纽斯医疗护理德国有限责任公司 用于确定呼吸频率的方法和装置
WO2011106249A1 (en) * 2010-02-26 2011-09-01 Nellcor Puritan Bennett Llc Spontaneous breathing trial manager
WO2016139652A1 (en) * 2015-03-05 2016-09-09 Oridion Medical 1987 Ltd. Identification of respiration waveforms during cpr
CN106073784A (zh) * 2016-08-17 2016-11-09 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN106344022A (zh) * 2016-09-18 2017-01-25 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN106388824A (zh) * 2016-10-14 2017-02-15 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN106388825A (zh) * 2016-10-14 2017-02-15 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN106901694A (zh) * 2017-02-20 2017-06-30 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107647868B (zh) * 2012-09-19 2021-05-11 瑞思迈传感器技术有限公司 用于确定睡眠阶段的系统和方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101528126A (zh) * 2006-12-21 2009-09-09 弗雷森纽斯医疗护理德国有限责任公司 用于确定呼吸频率的方法和装置
WO2011106249A1 (en) * 2010-02-26 2011-09-01 Nellcor Puritan Bennett Llc Spontaneous breathing trial manager
WO2016139652A1 (en) * 2015-03-05 2016-09-09 Oridion Medical 1987 Ltd. Identification of respiration waveforms during cpr
CN106073784A (zh) * 2016-08-17 2016-11-09 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN106344022A (zh) * 2016-09-18 2017-01-25 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN106388824A (zh) * 2016-10-14 2017-02-15 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN106388825A (zh) * 2016-10-14 2017-02-15 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置
CN106901694A (zh) * 2017-02-20 2017-06-30 广州视源电子科技股份有限公司 一种呼吸率提取方法及装置

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113288111A (zh) * 2021-05-10 2021-08-24 厦门理工学院 一种基于呼吸率的疲劳预测方法、终端设备及存储介质

Also Published As

Publication number Publication date
CN106901694A (zh) 2017-06-30

Similar Documents

Publication Publication Date Title
WO2018149147A1 (zh) 一种呼吸率提取方法及装置
CN107714023B (zh) 基于人工智能自学习的静态心电图分析方法和装置
CN108478209B (zh) 心电信息动态监护方法和动态监护系统
Korürek et al. Clustering MIT–BIH arrhythmias with Ant Colony Optimization using time domain and PCA compressed wavelet coefficients
Sutha et al. Fetal electrocardiogram extraction and analysis using adaptive noise cancellation and wavelet transformation techniques
JP6457117B2 (ja) 妊娠中の被験者の連続非侵襲モニタリング
CN101065058A (zh) 使用部分状态空间重构监视生理活动
JP2018512243A5 (zh)
CN106073784B (zh) 一种呼吸率提取方法及装置
Hadjem et al. ST-segment and T-wave anomalies prediction in an ECG data using RUSBoost
CN109259756B (zh) 基于非平衡训练的二级神经网络的ecg信号处理方法
CN109219391A (zh) 用于确定对象血压的设备和方法
CN112998690A (zh) 一种基于脉搏波多特征融合的呼吸率提取方法
Jenkal et al. Enhanced algorithm for QRS detection using discrete wavelet transform (DWT)
EP3764896B1 (en) Method and apparatus for monitoring a human or animal subject
Baldin et al. ECG signal spectral analysis approaches for high-resolution electrocardiography
Rashkovska et al. Clustering of heartbeats from ECG recordings obtained with wireless body sensors
CN112274159B (zh) 一种基于改进带通滤波器的压缩域心电信号质量评估方法
Kaya et al. Abnormality detection in noisy biosignals
Dembrani et al. Accurate detection of ECG signals in ECG monitoring systems by eliminating the motion artifacts and improving the signal quality using SSG filter with DBE
WO2018023698A1 (zh) 一种胎儿心电分离方法及装置
CN112754495A (zh) 一种基于云计算技术的便携式穿戴无感心电监测系统
Liao et al. AAFA-Net: Adaptive Amplitude-Frequency Attention Network for Fetal Monitoring from Noninvasive Abdominal Recordings
Raut et al. Development of Algorithm for Extraction of Fetal from Maternal ECG on Benchmark Database and Prototype Development for Acquisition
Bikulčienė et al. The measure of ECG complexity by matrix analysis

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17897163

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 13.01.2020)

122 Ep: pct application non-entry in european phase

Ref document number: 17897163

Country of ref document: EP

Kind code of ref document: A1