WO2024098819A1 - 一种运动脉搏波去噪方法及装置 - Google Patents

一种运动脉搏波去噪方法及装置 Download PDF

Info

Publication number
WO2024098819A1
WO2024098819A1 PCT/CN2023/105832 CN2023105832W WO2024098819A1 WO 2024098819 A1 WO2024098819 A1 WO 2024098819A1 CN 2023105832 W CN2023105832 W CN 2023105832W WO 2024098819 A1 WO2024098819 A1 WO 2024098819A1
Authority
WO
WIPO (PCT)
Prior art keywords
correlation coefficient
signal
factor
sensor
axis
Prior art date
Application number
PCT/CN2023/105832
Other languages
English (en)
French (fr)
Inventor
俞晓峰
张通
张海威
杨小牛
Original Assignee
广东粤港澳大湾区黄埔材料研究院
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 广东粤港澳大湾区黄埔材料研究院 filed Critical 广东粤港澳大湾区黄埔材料研究院
Publication of WO2024098819A1 publication Critical patent/WO2024098819A1/zh

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/024—Detecting, measuring or recording pulse rate or heart rate

Definitions

  • the present invention relates to the technical field of digital signal processing, and in particular to a method and device for denoising a motion pulse wave.
  • Adaptive filtering is a commonly used filtering method for pulse wave denoising to obtain heart rate. It includes a parameter-adjustable digital filter (Adaptive filter) and an adaptive filtering algorithm (Adaptive Algorithm).
  • Adaptive filter a parameter-adjustable digital filter
  • Adaptive Algorithm an adaptive filtering algorithm
  • the input signal is passed through the parameter-adjustable digital filter to generate an output signal, and the expected signal is compared with the expected signal to obtain an error signal.
  • the filtering process needs to be adjusted several times.
  • Using different adaptive filtering algorithms will produce differences in filtering results.
  • Common adaptive filtering algorithms include: Least Mean Square (LMS), Recursive Least Squares Filtering Algorithm
  • x(n) is the input signal
  • d(n) and y(n) are the desired signal and output signal respectively
  • e(n) is the error signal
  • w(n) and w(n+1) are the weight vectors of the adaptive linear combiner at this moment and the next moment respectively
  • ⁇ is the step size factor
  • the convergence condition of the LMS algorithm is to obtain the minimum mean square error (that is, the error between the desired signal and the output signal of the filter is infinitely reduced).
  • Initial convergence speed, time-varying system tracking capability and steady-state offset are the three most important technical indicators for measuring the quality of adaptive filtering algorithms. Since there is inevitably interference noise at the signal input, the adaptive filtering algorithm will generate parameter offset noise. The greater the interference noise, the greater the offset noise caused. Small step size factor can reduce the parameter offset noise of adaptive filtering algorithm and improve the convergence accuracy of the algorithm. However, the reduction of step size factor will reduce the convergence speed and tracking speed of the algorithm. Therefore, the requirements of adaptive filtering algorithm with fixed step size factor on algorithm adjustment step size factor in terms of convergence speed, time-varying system tracking speed and convergence accuracy are contradictory.
  • the prior art usually uses a fixed step size value for adaptive filtering to remove motion artifacts.
  • the proportion of pulse wave signal components and motion noise components contained in the sensor signals obtained under different motion states (walking, running, ball games, etc.) and different motion intensities is different. Therefore, when using adaptive filtering, the selected step size factor also needs to be adjusted according to the actual situation.
  • the existing patent CN 108652609 A (a heart rate acquisition method, system and wearable device) establishes the relationship between the magnitude of each motion signal, the motion state and the magnitude of the step size factor, and gives the value range of the step size factor under running, cycling and walking states.
  • the existing technology considers limited motion states and cannot adapt to different motion scenes; each time the motion scene or motion intensity changes, the value of the step factor needs to be manually set; and the range of the set step factor is large, and an accurate reference cannot be given in practice; in addition, the signal strength obtained by different types of accelerometer chips in the same environment may also lead to numerical differences in the step factor.
  • the embodiment of the present invention provides a motion pulse wave denoising method and device, which uses a relationship for establishing an optimal step length factor to pre-process the step length factor, can adapt to different motion scenes, and adjust the step length factor in real time.
  • an embodiment of the present invention provides a method for denoising a motion pulse wave, the method comprising:
  • the optimal step factor is obtained as the first optimal step factor
  • the sensor signal is filtered according to the first optimal step size factor and a preset adaptive filtering algorithm to obtain a denoised pulse wave signal.
  • the present invention pre-processes the step length factor by fitting a relationship formula of the optimal step length factor, and can calculate the correlation coefficient of the sensor signal and the N-axis accelerometer signal obtained in real time. According to the fitted relationship formula, a more accurate optimal step length factor corresponding to the motion scene can be obtained without manually modifying the step length factor, which can make the filtered spectrum data more accurate and reliable, thereby improving the accuracy and credibility of obtaining the exercise heart rate value.
  • obtaining the optimal step length factor as the first optimal step length factor according to the fitting curve of each correlation coefficient and the step length factor includes:
  • M M groups of sensor signals and N-axis accelerometer signals under different motion scenarios; where M is a positive integer;
  • the step factors of the M groups are adjusted respectively, and the step factor with the highest waveform amplitude and the most obvious period after filtering is selected as the second optimal step factor in the current motion scene, and the corresponding step factors of the M groups are obtained;
  • a first optimal step size factor is obtained according to the first correlation coefficient and the relationship.
  • the present invention describes the proportion of motion intensity and pulse wave/motion artifacts in sensor signals by calculating the correlation coefficient between sensor signals and N-axis accelerometer signals in different scenarios, and reflects the information of motion artifacts through the signals of the accelerometer.
  • the larger the correlation coefficient the greater the motion intensity.
  • the greater the motion interference in the sensor signal the larger the step length factor is required for denoising, and the value of the step length factor can be reflected more simply and conveniently.
  • a relational expression is obtained through curve fitting to accurately describe the relationship between the correlation coefficient and the step length factor. Under different correlation coefficients, that is, in different motion scenarios, the optimal step length factor can be accurately found through the relational expression, thereby improving the accuracy of denoising.
  • the calculating of the correlation coefficient between each group of sensor signals and the N-axis accelerometer signal as a second correlation coefficient includes:
  • the sensor signal data and N-axis accelerometer data of the same time period and the same length are extracted, and the correlation coefficients between the sensor signal data and each axis data of the N-axis accelerometer are respectively calculated to obtain N groups of correlation coefficients as the second correlation coefficients.
  • the relationship is specifically:
  • ⁇ * is the optimal step size factor in the current motion scene
  • abs( ⁇ ) and exp( ⁇ ) are exponential function operations with absolute value and natural constant e as base, respectively.
  • the specific calculation formula of the correlation is the Pearson correlation coefficient calculation formula:
  • pwi and acci (j) are the i-th signal data of the sensor and the i-th signal data on the j-th axis of the N-axis accelerometer, respectively, 1 ⁇ j ⁇ N; and are the mean values of the signal data of the sensor and the signal data on the j-th axis of the N-axis accelerometer, respectively;
  • L is the length of extracting the sensor signal data and the N-axis accelerometer signal data;
  • a(j) is the correlation coefficient between the sensor signal and the signal data on the j-th axis of the N-axis accelerometer; or,
  • the present invention is calculated by a Pearson correlation coefficient calculation method or a Spearman correlation coefficient calculation method, and the absolute value of the average value of the correlation coefficient is used to reflect the strength of the correlation between the sensor signal and the accelerometer signal.
  • the correlation between the sensor signal and the accelerometer signal can be described more accurately, and a reliable correlation coefficient can be obtained.
  • the curve fitting in different motion scenes can be more scientific and reliable, thereby improving the denoising ability of the adaptive filtering.
  • an embodiment of the present invention provides a motion pulse wave denoising device, the device comprising:
  • a first correlation coefficient calculation module used to obtain the currently collected sensor signal and N-axis accelerometer signal, and respectively calculate the correlation coefficient between the sensor signal and the accelerometer signal of each axis as a first correlation coefficient; wherein N is a positive integer;
  • a first optimal step factor calculation module used to obtain the optimal step factor as the first optimal step factor according to the relationship between the curve fitting of each correlation coefficient and each step factor;
  • the pulse wave signal denoising module is used to filter the sensor signal according to the first optimal step size factor and a preset adaptive filtering algorithm to obtain a denoised pulse wave signal.
  • the first optimal step size factor calculation module is specifically:
  • M M groups of sensor signals and N-axis accelerometer signals under different motion scenarios; where M is a positive integer;
  • the step factors of the M groups are adjusted respectively, and the step factor with the highest waveform amplitude and the most obvious period after filtering is selected as the second optimal step factor in the current motion scene, and the corresponding step factors of the M groups are obtained;
  • a first optimal step size factor is obtained according to the first correlation coefficient and the relationship.
  • the calculating of the correlation coefficient between each group of sensor signals and the N-axis accelerometer signal as a second correlation coefficient includes:
  • the sensor signal data and N-axis accelerometer data of the same time period and the same length are extracted, and the correlation coefficients between the sensor signal data and each axis data of the N-axis accelerometer are respectively calculated to obtain N groups of correlation coefficients as the second correlation coefficients.
  • the relationship is specifically:
  • ⁇ * is the optimal step size factor in the current motion scene
  • abs( ⁇ ) and exp( ⁇ ) are exponential function operations with absolute value and natural constant e as base, respectively.
  • the specific calculation formula of the correlation coefficient is the Pearson correlation coefficient calculation formula:
  • pwi and acci (j) are the i-th signal data of the sensor and the i-th signal data on the j-th axis of the N-axis accelerometer, respectively, 1 ⁇ j ⁇ N; and are the mean values of the signal data of the sensor and the signal data on the j-th axis of the N-axis accelerometer, respectively;
  • L is the length of extracting the sensor signal data and the N-axis accelerometer signal data;
  • a(j) is the correlation coefficient between the sensor signal and the signal data on the j-th axis of the N-axis accelerometer; or,
  • the present invention calculates the correlation coefficient in the current motion scene to obtain the optimal step factor through a first correlation coefficient calculation module, a first optimal step factor calculation module and a pulse wave signal denoising module, and then denoises the sensor signal according to a preset adaptive filtering algorithm to obtain a pulse wave signal with less noise influence.
  • the relationship between the correlation coefficient and the step factor is obtained by fitting a good curve, so that a more accurate step factor can be obtained, which can adapt to different types of motion and motion intensities, thereby improving the denoising ability of the adaptive filtering algorithm.
  • FIG1 is a schematic flow chart of a method for denoising a motion pulse wave provided by an embodiment of the present invention
  • FIG2 is a flow chart of a curve fitting process of a motion pulse wave denoising method provided by an embodiment of the present invention
  • FIG3 is a schematic diagram of a curve fitting of a motion pulse wave denoising method provided by an embodiment of the present invention.
  • FIG4 is a flow chart of a method for denoising a motion pulse wave provided by an embodiment of the present invention.
  • FIG5 is a schematic diagram showing a comparison of a motion pulse wave before and after denoising provided by an embodiment of the present invention
  • FIG6 is a schematic diagram of heart rate values in different sports scenarios after sports pulse wave denoising provided by an embodiment of the present invention.
  • FIG. 7 is a schematic structural diagram of a motion pulse wave denoising device provided in an embodiment of the present invention.
  • a method for denoising a motion pulse wave provided by an embodiment of the present invention includes steps S101 to S103, specifically:
  • Step S101 obtaining the currently collected sensor signal and N-axis accelerometer signal, and respectively calculating the correlation coefficient between the sensor signal and the accelerometer signal of each axis as a first correlation coefficient; wherein N is a positive integer.
  • the calculated first correlation coefficient can be the average of the correlation coefficients between the sensor signal and the accelerometer signals of each axis, or the correlation coefficient between the average of the accelerometer signals of each axis and the sensor signal, which is not limited here; in addition, the N-axis accelerometer is not limited to a 3-axis accelerometer or a 6-axis accelerometer.
  • Step S102 According to the relationship between the correlation coefficients and the curve fitting of the step factors, an optimal step factor is obtained as a first optimal step factor.
  • FIG. 2 it is a flow chart of a curve fitting process of a motion pulse wave denoising method provided by an embodiment of the present invention, including steps S21 to S24, which are specifically as follows:
  • Step S21 collecting M groups of sensor signals and N-axis accelerometer signals in different motion scenes; wherein M is a positive integer.
  • Step S22 Calculate the correlation coefficient between each group of sensor signals and the N-axis accelerometer signal as a second correlation coefficient.
  • the sensor signal data and N-axis accelerometer data of the same time period and length are extracted, and the correlation coefficients between the sensor signal data and each axis data of the N-axis accelerometer are respectively calculated to obtain N groups of correlation coefficients as the second correlation coefficients.
  • the calculated second correlation coefficients of each group may be the mean of the correlation coefficients between the sensor signal and the accelerometer signals of each axis, or the correlation coefficients between the mean of the accelerometer signals of each axis and the sensor signal, which is not limited here.
  • Step S23 According to the preset adaptive filtering algorithm, the step factors of the M groups are adjusted respectively, and the step factor with the highest waveform amplitude and the most obvious period after filtering is selected as the second optimal step factor in the current motion scene to obtain the corresponding M group step factors.
  • the preset adaptive filtering algorithm may be an LMS algorithm, an RLS algorithm or an NLMS algorithm, which is not limited here.
  • the LMS algorithm is selected as the preset adaptive filtering algorithm.
  • the step factors of the LMS algorithm under M groups of correlation coefficients are adjusted respectively, and the step factors are debugged several times.
  • the step factor with the highest waveform amplitude and the most obvious period after filtering is selected as the optimal step factor in the current motion scene.
  • the correlation coefficient can reflect the intensity of exercise and the proportion of pulse wave/motion artifacts in the collected signal data, and the signal data of the accelerometer reflects the information of motion artifacts
  • the signal data of the sensor is the superposition of motion artifacts and pulse wave signals.
  • Step S24 performing curve fitting on the M groups of second correlation coefficients and the second optimal step size factor to obtain a relationship.
  • the first optimal step factor is obtained according to the relationship obtained from the fitting curves of each correlation coefficient and each step factor and the first correlation coefficient.
  • the relationship is:
  • FIG3 it is a schematic diagram of a curve fitting of a motion pulse wave denoising method provided by an embodiment of the present invention.
  • the specific calculation formula of the correlation coefficient is the Pearson correlation coefficient calculation formula:
  • pwi and acci (j) are the i-th signal data of the sensor and the i-th signal data on the j-th axis of the N-axis accelerometer, respectively, 1 ⁇ j ⁇ N; and are the mean values of the signal data of the sensor and the signal data on the j-th axis of the N-axis accelerometer, respectively;
  • L is the length of extracting the sensor signal data and the N-axis accelerometer signal data;
  • a(j) is the correlation coefficient between the sensor signal and the signal data on the j-th axis of the N-axis accelerometer; or,
  • the Kendall correlation coefficient can also be used for calculation. Among them, if the correlation coefficient is greater than 0, it indicates that the correlation between the sensor signal and the accelerometer signal is positively correlated, if the correlation coefficient is less than 0, it indicates that the correlation between the sensor signal and the accelerometer signal is negatively correlated, and if the weak correlation coefficient is 1 or -1, it indicates that the sensor signal and the accelerometer signal can be described by a straight line equation.
  • Step S103 filtering the sensor signal according to the first optimal step size factor and a preset adaptive filtering algorithm to obtain a denoised pulse wave signal.
  • the filtered pulse wave signal data is converted into a frequency spectrum, and the position of the highest frequency peak is extracted as the exercise heart rate frequency, so as to obtain the exercise heart rate value.
  • the present invention describes the proportion of motion intensity and pulse wave/motion artifacts in the sensor signal by calculating the correlation coefficient between the sensor signal and the N-axis accelerometer signal in different scenarios.
  • the larger the correlation coefficient the greater the motion intensity and the greater the motion interference in the sensor signal, and a larger step factor is required for denoising.
  • the value of the step factor can be reflected more simply and conveniently, and the relationship is obtained by curve fitting.
  • the formula accurately describes the relationship between the correlation coefficient and the step-size factor. Under different correlation coefficients, that is, in different types of exercise and exercise intensities, the optimal step-size factor can be accurately found through the pre-fitted relationship, thereby improving the accuracy of denoising.
  • the present invention also provides an embodiment of a complete operation process.
  • FIG. 4 it is a flow chart of a method for denoising a motion pulse wave provided by an embodiment of the present invention, including steps S301 to S306, specifically:
  • Step S301 obtain the currently collected sensor signal as the input signal of the adaptive filter, transmit the input signal to the adaptive filter, illustratively, select the LMS algorithm as the preset adaptive filter, and enter step S302.
  • Step S302 Adaptive filtering processes the input signal and obtains an output signal.
  • w(n) is the weight vector of the adaptive linear combiner at this moment
  • x(n) is the input signal
  • y(n) is the output signal
  • Step S303 Get the expected signal and proceed to step S304.
  • e(n) is the error signal
  • d(n) is the desired signal
  • w(n+1) is the weight vector of the adaptive linear combiner at the next moment
  • ⁇ * is the optimal step size factor in the current motion scene.
  • step S302 repeat steps S302 to S305 to reduce the mean square error, and finally proceed to step S306.
  • Step S306 output the denoised pulse wave signal.
  • the present invention adopts the LMS algorithm as an adaptive filter, and can obtain the optimal step size factor under the current motion scene according to the relationship between the curve fitting and the optimal step size factor under the current motion scene, so as to adjust the adaptive linear combiner at the next moment.
  • the weight vector is updated, and when the convergence condition is reached, the denoised pulse wave signal data is more accurate, thereby improving the accuracy and credibility of the obtained heart rate value.
  • FIG5 is a comparative schematic diagram of a motion pulse wave before and after denoising provided by an embodiment of the present invention.
  • the original pulse waveforms and three-axis accelerometer signals of the head/wrist/ankle and other parts under different exercise intensities are collected, and then several groups of sensor data of different exercise heart rates are screened from the collected data.
  • the selected heart rate values include: 80bpm, 95bpm, 120bpm, 140bpm, 165bpm, and the population for data collection is males aged 20 to 30.
  • the sampling time for each group is 40 seconds, and the sampling frequency of the sensor and the accelerometer is 50Hz.
  • FIG5a is the signal data of the original sensor
  • FIG5b is the pulse wave signal data after adaptive filtering and denoising with the optimal step size factor.
  • the present invention also provides a schematic diagram of denoised heart rate values for different types of sports and at different exercise intensities, see Figure 6, which is a schematic diagram of heart rate values in different sports scenarios after sports pulse wave denoising provided by an embodiment of the present invention, Figures 6a, 6b and 6c are the heart rate values for running, cycling and basketball, respectively.
  • Figure 6a, 6b and 6c are the heart rate values for running, cycling and basketball, respectively.
  • FIG. 7 is a schematic diagram of the structure of a motion pulse wave denoising device provided by an embodiment of the present invention, including a first correlation coefficient calculation module 201 , a first optimal step factor calculation module 202 and a pulse wave signal denoising module 203 .
  • the first correlation coefficient calculation module 201 is used to obtain the currently collected sensor signal and N-axis accelerometer signal, and respectively calculate the correlation coefficient between the sensor signal and each axis accelerometer signal as a first correlation coefficient; wherein N is a positive integer.
  • the first optimal step factor calculation module 202 is used to obtain the optimal step factor as the first optimal step factor according to the relationship between the curve fitting of each correlation coefficient and each step factor.
  • M groups of sensor signals and N-axis accelerometer signals in different motion scenes are collected; wherein M is a positive integer.
  • the step factors of the M groups are adjusted respectively, and the step factor with the highest waveform amplitude and the most obvious period after filtering is selected as the second optimal step factor in the current motion scene to obtain the corresponding step factors of the M groups.
  • the sensor signal data and N-axis accelerometer data of the same time period and the same length are extracted, and the correlation coefficients between the sensor signal data and each axis data of the N-axis accelerometer are respectively calculated to obtain N groups of correlation coefficients as the second correlation coefficients.
  • Curve fitting is performed on the M groups of second correlation coefficients and the second optimal step size factor to obtain a relationship.
  • a first optimal step size factor is obtained according to the first correlation coefficient and the relationship.
  • the pulse wave signal denoising module 203 is used to filter the sensor signal according to the first optimal step size factor and a preset adaptive filtering algorithm to obtain a denoised pulse wave signal.
  • the relationship is specifically:
  • ⁇ * is the optimal step size factor in the current motion scene
  • abs( ⁇ ) and exp( ⁇ ) are exponential function operations with absolute value and natural constant e as base, respectively.
  • the specific calculation formula of the correlation is the Pearson correlation coefficient calculation formula:
  • pwi and acci (j) are the i-th signal data of the sensor and the i-th signal data on the j-th axis of the N-axis accelerometer, respectively, 1 ⁇ j ⁇ N; and are the mean values of the signal data of the sensor and the signal data on the j-th axis of the N-axis accelerometer, respectively;
  • L is the length of extracting the sensor signal data and the N-axis accelerometer signal data;
  • a(j) is the correlation coefficient between the sensor signal and the signal data on the j-th axis of the N-axis accelerometer; or,
  • the present invention adopts the LMS algorithm as an adaptive filter, and the first optimal step factor calculation module 202 obtains the relationship between the curve fitting of each correlation coefficient and each correlation factor, so as to obtain the relationship between the correlation coefficient and the correlation factor.
  • the optimal correlation factor in the pre-motion scenario can make the denoising ability of the adaptive filter stronger, thereby obtaining more accurate pulse wave signal data.
  • the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc.
  • special hardware including special integrated circuits, special CPUs, special memories, special components, etc.
  • all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structures used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits.
  • software program implementation is a better implementation mode in most cases.
  • the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.
  • a computer device which can be a personal computer, a server, or a network device, etc.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Physics & Mathematics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Cardiology (AREA)
  • Physiology (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种运动脉搏波去噪方法及装置，包括获取当前采集的传感器信号、N轴加速度计信号，并分别计算传感器信号与各轴加速度计信号之间的相关系数为第一相关系数（S101）；其中，N为正整数；根据各相关系数和各步长因子的曲线拟合的关系式，获得最优步长因子为第一最优步长因子（S102）；根据第一最优步长因子和预设自适应滤波算法，对传感器信号进行滤波处理，获得去噪后的脉搏波信号（S103）。通过曲线拟合得到的关系式，描述相关系数和步长因子之间的关系，在不同的运动场景中，能够通过关系式找到最优步长因子，从而提高去噪的精度。

Description

一种运动脉搏波去噪方法及装置 技术领域
本发明涉及数字信号处理技术领域，尤其涉及一种运动脉搏波去噪方法及装置。
背景技术
自适应滤波是进行脉搏波去噪获取心率时常用的滤波方法，包含一个参数可调数字滤波器(Adaptive filter)、自适应滤波算法(Adaptive Algorithm)，输入信号通过参数可调数字滤波器产生输出信号，将期望信号与进行比较，得到误差信号，为了使误差信号尽可能小，需要经过若干次循环调整滤波过程。采用不同的自适应滤波算法对滤波结果会产生差异，常见的自适应滤波算法包括：最小均方滤波算法(Least Mean Square,LMS)、递推最小二乘法滤波算法
(Recursive Least Square,RLS)、归一化最小均方自适应滤波算法(Normalized Least Mean Square,NLMS)，常用的LMS算法迭代过程包括：
计算由输入信号通过参数可调数字滤波器产生的输出信号，y(n)＝wT(n)x(n)；
计算误差信号,e(n)＝d(n)-y(n)；
更新下一时刻的自适应线性组合器的权矢量，w(n+1)＝w(n)+2μe(n)x(n)；
其中，x(n)是输入信号，d(n)和y(n)分别是期望信号和输出信号，e(n)是误差信号，w(n)和w(n+1)分别是本时刻和下一时刻的自适应线性组合器的权矢量，μ是步长因子，LMS算法的收敛条件是获得均方误差最小(即期望信号和滤波器的输出信号的误差无限缩小)。
初始收敛速度、时变***跟踪能力及稳态失调是衡量自适应滤波算法优劣的三个最重要的技术指标。由于信号输入端不可避免地存在干扰噪声，自适应滤波算法将产生参数失调噪声，干扰噪声越大，则引起的失调噪声就越大。减 小步长因子可减小自适应滤波算法的参数失调噪声，提高算法的收敛精度。然而，步长因子的减小将降低算法的收敛速度和跟踪速度。因此，固定步长因子的自适应滤波算法在收敛速度、时变***跟踪速度与收敛精度方面对算法调整步长因子的要求是相互矛盾的。
因此，可通过设置合适的步长因子来提高滤波器的稳定性。现有技术通常会采用固定步长的值进行自适应滤波去除运动伪迹，然而，在运动测试的场景中，不同运动状态下(行走，跑步，球类等)以及不同运动强度下获得的传感器信号中所包含的脉搏波信号成分和运动噪声成分比例有所差异，因此，采用自适应滤波时，所选的步长因子也需要根据实际情况调节。现有专利CN 108652609 A(一种心率获取方法、***及可穿戴式设备)，建立了各运动信号的量级、运动状态与步长因子的量级之间的关系，给定了包括：跑步、骑行、行走状态下步长因子取值区间。
但是，现有技术所考虑的运动状态有限，不能自适应不同的运动场景；每更换一次运动场景或者运动强度，都需手动设置步长因子的取值；并且，所设置的步长因子的范围较大，在实际中无法给出准确的参考；此外，由不同类型的加速度计芯片在相同环境下获得的信号强度也可能带来步长因子的数值差异。
发明内容
本发明实施例的提供了一种运动脉搏波去噪方法及装置，采用建立最优步长因子的关系式对步长因子进行预处理，能够自适应不同的运动场景，实时对步长因子进行调整。
第一方面，本发明实施例提供了一种运动脉搏波去噪方法，所述方法包括：
获取当前采集的传感器信号、N轴加速度计信号，并分别计算所述传感器信号与各轴加速度计信号之间的相关系数为第一相关系数；其中，N为正整数；
根据各相关系数和各步长因子的曲线拟合的关系式，获得最优步长因子为第一最优步长因子；
根据所述第一最优步长因子和预设自适应滤波算法，对所述传感器信号进行滤波处理，获得去噪后的脉搏波信号。
本发明通过拟合最优步长因子的关系式，对步长因子进行预处理，可对实时获得的传感器信号和N轴加速度计信号计算相关系数，根据拟合好的关系式，可得到对应运动场景下的更加精准最优步长因子，而无需手动修改步长因子，能够使滤波后的频谱数据更加准确和可靠，从而提高获取运动心率值的准确度和可信度。
进一步，所述根据各相关系数和步长因子的拟合曲线，获得最优步长因子为第一最优步长因子，包括：
采集M组不同运动场景下的传感器信号和N轴加速度计信号；其中，M为正整数；
计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数；
根据所述预设自适应滤波算法，分别调节M组的步长因子，选取滤波后波形振幅最高、周期最明显的步长因子为当前运动场景下的第二最优步长因子，得到对应M组步长因子；
对M组第二相关系数和所述第二最优步长因子进行曲线拟合，得到关系式；
根据所述第一相关系数和所述关系式，获得第一最优步长因子。
本发明通过计算不同场景下的传感器信号和N轴加速度计信号的相关系数，来描述传感器信号中运动强度及脉搏波/运动伪迹的占比，通过加速度计的信号来反映运动伪迹的信息，相关系数越大，则表明运动强度越大，传感器信号中的运动干扰越大，则需要更大的步长因子来进行去噪，能够更加简单方便的反映出步长因子的取值，通过曲线拟合得到关系式，准确描述相关系数和步长因子之间的关系，在不同的相关系数下，即在不同的运动场景中，都能够通过关系式准确找到最优步长因子，从而提高去噪的精度。
进一步，所述计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数，包括：
提取相同时间段、相同长度的传感器信号的数据和N轴加速度计数据，分别计算传感器信号的数据与N轴加速度计的每一轴数据的相关系数，得到N组相关系数为所述第二相关系数。
优选地，所述关系式具体为：
或者其中，是当前运动场景下的相关系数，μ*是当前运动场景下的最优步长因子，abs(·)和exp(·)分别为取绝对值和取自然常数e为底的指数函数操作。
优选地，所述相关关系的具体计算公式为皮尔森相关系数计算公式：
其中，pwi和acci(j)分别是传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据，1≤j≤N；和分别是传感器的信号数据的均值和N轴加速度计的第j轴上的信号数据的均值；L是提取传感器信号数据和N轴加速度计信号数据的长度；a(j)是传感器信号和N轴加速度计的第j轴上的信号数据的相关系数；或者，
使用斯皮尔曼相关公式计算：
其中，为传感器的第i条信号数据和N轴传感器的第j(1≤j≤N)轴上的第i条信号数据的等级差。
本发明通过皮尔森相关系数计算方法或者斯皮尔曼相关系数计算方法来计算，并用相关系数的平均值的绝对值来反映传感器信号和加速度计信号的相关性强弱，相关系数的平均值的绝对值越大，则相关性越强，能够更加准确的描述传感器信号和加速度计信号的相关关系，得到可信的相关系数，能够对不同运动场景下的曲线拟合更具科学性和可靠性，从而提高自适应滤波的去噪能力。
第二方面，本发明实施例提供了一种运动脉搏波去噪装置，所述装置包括：
第一相关系数计算模块，用于获取当前采集的传感器信号、N轴加速度计信号，并分别计算所述传感器信号与各轴加速度计信号之间的相关系数为第一相关系数；其中，N为正整数；
第一最优步长因子计算模块，用于根据各相关系数和各步长因子的曲线拟合的关系式，获得最优步长因子为第一最优步长因子；
脉搏波信号去噪模块，用于根据所述第一最优步长因子和预设自适应滤波算法，对所述传感器信号进行滤波处理，获得去噪后的脉搏波信号。
进一步，所述第一最优步长因子计算模块具体为：
采集M组不同运动场景下的传感器信号和N轴加速度计信号；其中，M为正整数；
计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数；
根据所述预设自适应滤波算法，分别调节M组的步长因子，选取滤波后波形振幅最高、周期最明显的步长因子为当前运动场景下的第二最优步长因子，得到对应M组步长因子；
对M组第二相关系数和所述第二最优步长因子进行曲线拟合，得到关系式；
根据所述第一相关系数和所述关系式，获得第一最优步长因子。
进一步，所述计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数，包括：
提取相同时间段、相同长度的传感器信号的数据和N轴加速度计数据，分别计算传感器信号的数据与N轴加速度计的每一轴数据的相关系数，得到N组相关系数为所述第二相关系数。
优选地，所述关系式具体为：
或者其中，是当前运动场景下的相关系数，μ*是当前运动场景下的最优步长因子，abs(·)和exp(·)分别为取绝对值和取自然常数e为底的指数函数操作。
优选地，所述相关系数的具体计算公式为皮尔森相关系数计算公式：
其中，pwi和acci(j)分别是传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据，1≤j≤N；和分别是传感器的信号数据的均值和N轴加速度计的第j轴上的信号数据的均值；L是提取传感器信号数据和N轴加速度计信号数据的长度；a(j)是传感器信号和N轴加速度计的第j轴上的信号数据的相关系数；或者，
使用斯皮尔曼相关公式计算：
其中，为传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据的等级差。
本发明通过第一相关系数计算模块、第一最优步长因子计算模块和脉搏波信号去噪模块，计算当前运动场景下的相关系数获得最优步长因子，进而根据预设的自适应滤波算法对传感器信号进行去噪，得到噪声影响更小的脉搏波信号，无需根据不同的运动场景进行手动修改步长因子，通过拟合好的曲线得到相关系数与步长因子的关系式，能够获得更加精确的步长因子，能够适应不同的运动种类、运动强度，从而能够提高自适应滤波算法去噪的能力。
附图说明
图1是本发明实施例提供的一种运动脉搏波去噪方法的流程示意图；
图2是本发明实施例提供的一种运动脉搏波去噪方法的曲线拟合的流程示意图；
图3是本发明实施例提供的一种运动脉搏波去噪方法的曲线拟合的示意图；
图4是本发明实施例提供的一种运动脉搏波去噪方法的流程示意图；
图5是本发明实施例提供的一种运动脉搏波去噪前及去噪后的对比示意图；
图6是本发明实施例提供的一种运动脉搏波去噪后不同运动场景下的心率值示意图；
图7是本发明实施例提供的一种运动脉搏波去噪装置的结构示意图。
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。
参见图1，是本发明实施例提供的一种运动脉搏波去噪方法，包括步骤S101～S103，具体为：
步骤S101、获取当前采集的传感器信号、N轴加速度计信号，并分别计算所述传感器信号与各轴加速度计信号之间的相关系数为第一相关系数；其中，N为正整数。
在一个可选的实施方式中，获取到当前采集的传感器信号、N轴加速度计信号后，计算的所述第一相关系数可以为所述传感器信号与各轴加速度计信号之间的相关系数的均值，也可以为各轴加速度计信号的均值与所述传感器信号的相关系数，在此不做限定；此外，所述N轴加速度计不限于3轴加速度计、6轴加速度计。
步骤S102、根据各相关系数和各步长因子的曲线拟合的关系式，获得最优步长因子为第一最优步长因子。
具体地参见图2，是本发明实施例提供的一种运动脉搏波去噪方法的曲线拟合的流程示意图，包括步骤S21～步骤S24，具体如下：
步骤S21、采集M组不同运动场景下的传感器信号和N轴加速度计信号；其中，M为正整数。
步骤S22、计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数。
具体为，提取相同时间段、相同长度的传感器信号的数据和N轴加速度计数据，分别计算传感器信号的数据与N轴加速度计的每一轴数据的相关系数，得到N组相关系数为所述第二相关系数。
值得说明的是，计算的每组第二相关系数可以为所述传感器信号与各轴加速度计信号之间的相关系数的均值，也可以为各轴加速度计信号的均值与所述传感器信号的相关系数，在此不做限定。
步骤S23、根据所述预设自适应滤波算法，分别调节M组的步长因子，选取滤波后波形振幅最高、周期最明显的步长因子为当前运动场景下的第二最优步长因子，得到对应M组步长因子。
其中，所述预设自适应滤波算法可以为LMS算法、RLS算法或者NLMS算法，在此不做限定，作为示例，选取LMS算法为预设自适应滤波算法。分别调节在M组相关系数下LMS算法的步长因子，对步长因子进行若干次调试，选择滤波后波形振幅最高、周期最明显的步长因子为当前运动场景下的最优步长因子。
值得说明的是，由于相关系数能够反映运动强度及脉搏波/运动伪迹在采集的信号数据中的占比，而加速度计的信号数据反映了运动伪迹的信息，传感器的信号数据是运动伪迹和脉搏波信号的叠加，相关系数越大，表明当前运动强度越大，传感器的信号数据中的运动干扰越大，在自适应滤波器中需要设置更大的步长因子进行去噪；反之，相关系数较小，则传感器的信号数据中的运动干扰越小运动强度较低，采用小的步长因子可以获得更准确的脉搏波信号。
步骤S24、对M组第二相关系数和所述第二最优步长因子进行曲线拟合，得到关系式。
其中，根据各相关系数和各步长因子的拟合曲线得到的关系式、所述第一相关系数，获得第一最优步长因子。
优选地，根据拟合的曲线，所述关系式为：
或者通过线性拟合：其中，是当前运动场景下的相关系数，μ*是当前运动场景下的最优步长因子，abs(·)和exp(·)分别为取绝对值和取自然常数e为底的指数函数操作。参见图3，是本发明实施例提供的一种运动脉搏波去噪方法的曲线拟合的示意图。
优选地，所述相关系数的具体计算公式为皮尔森相关系数计算公式：
其中，pwi和acci(j)分别是传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据，1≤j≤N；和分别是传感器的信号数据的均值和N轴加速度计的第j轴上的信号数据的均值；L是提取传感器信号数据和N轴加速度计信号数据的长度；a(j)是传感器信号和N轴加速度计的第j轴上的信号数据的相关系数；或者，
使用斯皮尔曼相关公式计算：
其中，为传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据的等级差。
值得说明的是，除了使用上述皮尔森(Pearson)相关系数和斯皮尔曼(Spearman)相关系数公式进行计算以外，还可以采用肯德尔(Kendall)相关系数进行计算。其中，若相关系数大于0，则表明传感器信号和加速度计信号的相关性为正相关，若相关系数小于0，则表明传感器信号和加速度计信号的相关性为负相关，弱相关系数为1或者为-1，则表明传感器信号和加速度计信号可以由直线方程进行描述。
步骤S103、根据所述第一最优步长因子和预设自适应滤波算法，对所述传感器信号进行滤波处理，获得去噪后的脉搏波信号。
值得说明的是，将滤波后的脉搏波信号数据转化成频谱，提取频率峰最高的位置作为运动心率频率，即可获得运动心率值。
本发明通过计算不同场景下的传感器信号和N轴加速度计信号的相关系数，来描述传感器信号中运动强度及脉搏波/运动伪迹的占比，相关系数越大，则表明运动强度越大，传感器信号中的运动干扰越大，则需要更大的步长因子来进行去噪，能够更加简单方便的反映出步长因子的取值，通过曲线拟合得到关系 式，准确描述相关系数和步长因子之间的关系，在不同的相关系数下，即在不同的运动种类、运动强度中，都能够通过预先拟合好的关系式准确找到最优步长因子，从而提高去噪的精度。
本发明还提供了操作过程完整的实施例，参见图4，是本发明实施例提供的一种运动脉搏波去噪方法的流程示意图，包括步骤S301～S306，具体为：
步骤S301、获取当前采集的传感器信号为自适应滤波器的输入信号，将输入信号传输给自适应滤波，示例性的，选择LMS算法作为预设的自适应滤波器，进入步骤S302。
步骤S302、自适应滤波将输入信号进行处理，并得到输出信号，输出信号可以表示为：
y(n)＝wT(n)x(n)，
其中，w(n)是本时刻的自适应线性组合器的权矢量，x(n)为输入信号，y(n)为输出信号，进入步骤S303。
步骤S303、获取期望信号，进入步骤S304。
步骤S304、将获得的期望信号和计算得到的输出信号做差，得到误差信号，计算如下：
e(n)＝d(n)-y(n)，
其中，e(n)为误差信号，d(n)为期望信号，进入步骤S305。
步骤S305、根据曲线拟合好的关系式、得到的当前运动场景下的最优步长因子，更新下一时刻的自适应线性组合器的权矢量，计算如下：
w(n+1)＝w(n)+2μ*e(n)x(n)，
其中，w(n+1)是下一时刻的自适应线性组合器的权矢量，μ*是当前运动场景下的最优步长因子。
再进入步骤S302，重复步骤S302～S305，减小均方误差，最后进入步骤S306。
步骤S306、输出去噪后的脉搏波信号。
本发明通过采用LMS算法作为自适应滤波器，根据曲线拟合好的关系式、得到的当前运动场景下的最优步长因子，能够对下一时刻的自适应线性组合器 的权矢量进行更新，当达到收敛条件时的去噪的脉搏波信号的数据更加精准，从而提高获得的心率值的准确度和可信度。
示例性地，参见图5，是本发明实施例提供的一种运动脉搏波去噪前及去噪后的对比示意图。采集不同运动强度下头部/腕部/脚踝等部位的原始脉搏波形以及三轴加速度计信号，再从采集到的数据中筛选若干组的不同的运动心率的传感器数据，选取的心率值包括：80bpm，95bpm，120bpm，140bpm，165bpm，数据采集的人群为20～30岁的男性，每组采样时间为40秒，所述传感器和所述加速度计的采样频率皆为50Hz。图5a为原始传感器的信号数据，图5b为经过最佳步长因子的自适应滤波去噪后的脉搏波信号数据。
本发明还提供不同运动种类、在不同运动强度下的去噪后的心率值示意图，参见图6，图6是本发明实施例提供的一种运动脉搏波去噪后不同运动场景下的心率值示意图，图6a、图6b和图6c分别为跑步、骑行和篮球运动时的心率值，通过检测头部和胸部的心率，通过最佳步长因子的自适应滤波去噪后的心率值，可以自适应不同的运动种类及运动强度，无需每次手动更改步长因子。
参见图7，是本发明实施例提供的一种运动脉搏波去噪装置的结构示意图，包括第一相关系数计算模块201、第一最优步长因子计算模块202和脉搏波信号去噪模块203。
其中，第一相关系数计算模块201，用于获取当前采集的传感器信号、N轴加速度计信号，并分别计算所述传感器信号与各轴加速度计信号之间的相关系数为第一相关系数；其中，N为正整数。
第一最优步长因子计算模块202，用于根据各相关系数和各步长因子的曲线拟合的关系式，获得最优步长因子为第一最优步长因子。
具体地，采集M组不同运动场景下的传感器信号和N轴加速度计信号；其中，M为正整数。
计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数；
根据所述预设自适应滤波算法，分别调节M组的步长因子，选取滤波后波形振幅最高、周期最明显的步长因子为当前运动场景下的第二最优步长因子，得到对应M组步长因子。
具体地，提取相同时间段、相同长度的传感器信号的数据和N轴加速度计数据，分别计算传感器信号的数据与N轴加速度计的每一轴数据的相关系数，得到N组相关系数为所述第二相关系数。
对M组第二相关系数和所述第二最优步长因子进行曲线拟合，得到关系式。
根据所述第一相关系数和所述关系式，获得第一最优步长因子。
脉搏波信号去噪模块203，用于根据所述第一最优步长因子和预设自适应滤波算法，对所述传感器信号进行滤波处理，获得去噪后的脉搏波信号。
优选地，所述关系式具体为：
或者其中，是当前运动场景下的相关系数，μ*是当前运动场景下的最优步长因子，abs(·)和exp(·)分别为取绝对值和取自然常数e为底的指数函数操作。
优选地，所述相关关系的具体计算公式为皮尔森相关系数计算公式：
其中，pwi和acci(j)分别是传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据，1≤j≤N；和分别是传感器的信号数据的均值和N轴加速度计的第j轴上的信号数据的均值；L是提取传感器信号数据和N轴加速度计信号数据的长度；a(j)是传感器信号和N轴加速度计的第j轴上的信号数据的相关系数；或者，
使用斯皮尔曼相关公式计算：
其中，为传感器的第i条信号数据和N轴传感器的第j(1≤j≤N)轴上的第i条信号数据的等级差。
本发明通过采用LMS算法作为自适应滤波器，通过第一最优步长因子计算模块202对各相关系数和各相关因子的曲线拟合得到的关系式，能够求得对当 前运动场景下的最优相关因子，能够使自适应滤波器的去噪能力更强，从而获得更加精准的脉搏波信号数据。
通过以上的实施方式的描述，所属领域的技术人员可以清楚地了解到本发明可借助软件加必需的通用硬件的方式来实现，当然也可以通过专用硬件包括专用集成电路、专用CPU、专用存储器、专用元器件等来实现。一般情况下，凡由计算机程序完成的功能都可以很容易地用相应的硬件来实现，而且，用来实现同一功能的具体硬件结构也可以是多种多样的，例如模拟电路、数字电路或专用电路等。但是，对本发明而言更多情况下软件程序实现是更佳的实施方式。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在可读取的存储介质中，如计算机的软盘，U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本发明各个实施例所述的方法。
以上所述是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也视为本发明的保护范围。

Claims (10)

  1. 一种运动脉搏波去噪方法，其特征在于，所述方法包括：
    获取当前采集的传感器信号、N轴加速度计信号，并分别计算所述传感器信号与各轴加速度计信号之间的相关系数为第一相关系数；其中，N为正整数；
    根据各相关系数和各步长因子的曲线拟合的关系式，获得最优步长因子为第一最优步长因子；
    根据所述第一最优步长因子和预设自适应滤波算法，对所述传感器信号进行滤波处理，获得去噪后的脉搏波信号。
  2. 如权利要求1所述的运动脉搏波去噪方法，其特征在于，所述根据各相关系数和步长因子的拟合曲线，获得最优步长因子为第一最优步长因子，包括：
    采集M组不同运动场景下的传感器信号和N轴加速度计信号；其中，M为正整数；
    计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数；
    根据所述预设自适应滤波算法，分别调节M组的步长因子，选取滤波后波形振幅最高、周期最明显的步长因子为当前运动场景下的第二最优步长因子，得到对应M组步长因子；
    对M组第二相关系数和所述第二最优步长因子进行曲线拟合，得到关系式；
    根据所述第一相关系数和所述关系式，获得第一最优步长因子。
  3. 如权利要求2所述的运动脉搏波去噪方法，其特征在于，所述计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数，包括：
    提取相同时间段、相同长度的传感器信号的数据和N轴加速度计数据，分别计算传感器信号的数据与N轴加速度计的每一轴数据的相关系数，得到N组相关系数为所述第二相关系数。
  4. 如权利要求1-2任一所述的运动脉搏波去噪方法，其特征在于，所述关系式具体为：
    或者其中，是当前运动场景下的相关系数，μ*是当前运动场景下的最优步长因子，abs(·)和exp(·)分别为取绝对值和取自然常数e为底的指数函数操作。
  5. 如权利要求1-4任一所述的运动脉搏波去噪方法，其特征在于，所述相关系数的具体计算公式为皮尔森相关系数计算公式：
    其中，pwi和acci(j)分别是传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据，1≤j≤N；和分别是传感器的信号数据的均值和N轴加速度计的第j轴上的信号数据的均值；L是提取传感器信号数据和N轴加速度计信号数据的长度；a(j)是传感器信号和N轴加速度计的第j轴上的信号数据的相关系数；或者，
    使用斯皮尔曼相关公式计算：
    其中，为传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据的等级差。
  6. 一种运动脉搏波去噪装置，其特征在于，包括：
    第一相关系数计算模块，用于获取当前采集的传感器信号、N轴加速度计信号，并分别计算所述传感器信号与各轴加速度计信号之间的相关系数为第一相关系数；其中，N为正整数；
    第一最优步长因子计算模块，用于根据各相关系数和各步长因子的曲线拟合的关系式，获得最优步长因子为第一最优步长因子；
    脉搏波信号去噪模块，用于根据所述第一最优步长因子和预设自适应滤波算法，对所述传感器信号进行滤波处理，获得去噪后的脉搏波信号。
  7. 如权利要求6所述的运动脉搏波去噪装置，其特征在于，所述第一最优步长因子计算模块具体为：
    采集M组不同运动场景下的传感器信号和N轴加速度计信号；其中，M为正整数；
    计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数；
    根据所述预设自适应滤波算法，分别调节M组的步长因子，选取滤波后波形振幅最高、周期最明显的步长因子为当前运动场景下的第二最优步长因子，得到对应M组步长因子；
    对M组第二相关系数和所述第二最优步长因子进行曲线拟合，得到关系式；
    根据所述第一相关系数和所述关系式，获得第一最优步长因子。
  8. 如权利要求7所述的运动脉搏波去噪装置，其特征在于，所述计算每组传感器信号和N轴加速度计信号的相关系数为第二相关系数，包括：
    提取相同时间段、相同长度的传感器信号的数据和N轴加速度计数据，分别计算传感器信号的数据与N轴加速度计的每一轴数据的相关系数，得到N组相关系数为所述第二相关系数。
  9. 如权利要求6-8任一所述的运动脉搏波去噪装置，其特征在于，所述关系式具体为：
    或者其中，是当前运动场景下的相关系数，μ*是当前运动场景下的最优步长因子，abs(·)和exp(·)分别为取绝对值和取自然常数e为底的指数函数操作。
  10. 如权利要求6-8任一所述的运动脉搏波去噪装置，其特征在于，所述相关系数的具体计算公式为皮尔森相关系数计算公式：
    其中，pwi和acci(j)分别是传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据，1≤j≤N；和分别是传感器的信号数据的均值和N轴加速度计的第j轴上的信号数据的均值；L是提取传感器信号数据和N轴加速度计信号数据的长度；a(j)是传感器信号和N轴加速度计的第j轴上的信号数据的相关系数；或者，
    使用斯皮尔曼相关公式计算：
    其中，为传感器的第i条信号数据和N轴加速度计的第j轴上的第i条信号数据的等级差。
PCT/CN2023/105832 2022-11-09 2023-07-05 一种运动脉搏波去噪方法及装置 WO2024098819A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211398274.1A CN115886744A (zh) 2022-11-09 2022-11-09 一种运动脉搏波去噪方法及装置
CN202211398274.1 2022-11-09

Publications (1)

Publication Number Publication Date
WO2024098819A1 true WO2024098819A1 (zh) 2024-05-16

Family

ID=86487163

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/105832 WO2024098819A1 (zh) 2022-11-09 2023-07-05 一种运动脉搏波去噪方法及装置

Country Status (2)

Country Link
CN (1) CN115886744A (zh)
WO (1) WO2024098819A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106880351A (zh) * 2015-12-15 2017-06-23 德州仪器公司 减少光体积描记(ppg)信号中运动诱发的伪影
US20180146926A1 (en) * 2015-06-02 2018-05-31 Sony Corporation Noise reduction processing circuit and method, and biological information processing device and method
CN108652609A (zh) * 2018-05-17 2018-10-16 歌尔科技有限公司 一种心率获取方法、***及可穿戴式设备
US20210290092A1 (en) * 2016-09-26 2021-09-23 Sony Corporation Vital sign processing device, vital sign processing method, and information processing device

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180146926A1 (en) * 2015-06-02 2018-05-31 Sony Corporation Noise reduction processing circuit and method, and biological information processing device and method
CN106880351A (zh) * 2015-12-15 2017-06-23 德州仪器公司 减少光体积描记(ppg)信号中运动诱发的伪影
US20210290092A1 (en) * 2016-09-26 2021-09-23 Sony Corporation Vital sign processing device, vital sign processing method, and information processing device
CN108652609A (zh) * 2018-05-17 2018-10-16 歌尔科技有限公司 一种心率获取方法、***及可穿戴式设备

Also Published As

Publication number Publication date
CN115886744A (zh) 2023-04-04

Similar Documents

Publication Publication Date Title
JP4842583B2 (ja) 多感覚音声強調のための方法および装置
CN109767783A (zh) 语音增强方法、装置、设备及存储介质
US7346504B2 (en) Multi-sensory speech enhancement using a clean speech prior
CN107004424A (zh) 噪声降低和语音增强的方法、设备和***
CN108899052B (zh) 一种基于多带谱减法的帕金森语音增强方法
CN108735229B (zh) 基于信噪比加权的幅度与相位联合补偿抗噪语音增强方法
CN110840445A (zh) 一种动态心电信号的自动降噪方法
CN114469124B (zh) 一种运动过程中异常心电信号的识别方法
CN109979476A (zh) 一种语音去混响的方法及装置
CN113349752A (zh) 一种基于传感融合的可穿戴设备实时心率监测方法
CN112370036A (zh) 基于级联型rls自适应滤波的ppg心率提取装置和方法
US20240046947A1 (en) Speech signal enhancement method and apparatus, and electronic device
CN110940409A (zh) 一种基于iceemdan与ica联合去噪的超声信号测量方法
US6216031B1 (en) Apparatus for enhancing signals in ECGs artefacts
CN114510976B (zh) 基于双变量阈值函数的呼吸信号去噪方法及装置
WO2024098819A1 (zh) 一种运动脉搏波去噪方法及装置
CN108652609B (zh) 一种心率获取方法、***及可穿戴式设备
CN111938627A (zh) 一种滤除心磁信号基线漂移的方法
CN111461206A (zh) 一种方向盘嵌入脑电传感器的基于脑电的疲劳检测方法
CN108735231B (zh) 主旋律音高序列估计方法
CN106341564A (zh) 一种信号数据的处理方法、装置及智能会议设备
Sun et al. Wavelet denoising method based on improved threshold function
CN110269608A (zh) 一种去除信号干扰的方法、装置以及可读存储介质
CN113925482A (zh) 心率计算方法、可穿戴电子设备和存储介质
JP3163570B2 (ja) 核磁気共鳴イメージングにおける動画像の画質改善方法