Multi-parameter fusion pulse wave signal quality assessment methodTechnical Field
The invention relates to the technical field of medical diagnosis, in particular to a pulse wave signal quality evaluation method based on multi-parameter fusion.
Background
Blood oxygen saturation (SpO 2) is the percentage of the volume of oxyhemoglobin (HbO 2) bound by oxygen in the blood to the volume of total hemoglobin (Hb) that can be bound, i.e. the concentration of blood oxygen in the blood, which is an important physiological parameter of the respiratory cycle. The ratio of HbO2 concentration to HbO2+Hb concentration is different from the percentage of oxyhemoglobin, so monitoring arterial oxygen saturation can estimate the oxygenizing and hemoglobin oxygen carrying capacity of the lung. The blood oxygen saturation of normal human arterial blood is 98% and venous blood is 75%.
The metabolism of human body is biological oxidation process, and the oxygen needed in the metabolism is introduced into human blood through respiratory system, and combined with hemoglobin in blood erythrocytes to form oxyhemoglobin, which is then transferred into tissue cells of each part of human body, and the oxygen carrying capacity of blood is measured by the oxygen saturation of blood.
Pulse oximeters are monitors that measure pulse rate and blood oxygen saturation noninvasively, however, pulse oximeters are susceptible to noise interference, negatively affect the accuracy of the output pulse rate and blood oxygen saturation, and when the output pulse rate is uncorrelated with the electrocardio, clinicians are prone to challenge, so that it is necessary to evaluate the reliability of the output result of the pulse oximeter;
The signal quality evaluation in the existing scheme mainly comprises signal quality evaluation based on acceleration and signal quality evaluation based on feature extraction and machine learning or deep learning, body movement judgment is carried out based on triaxial acceleration extraction signal combination vector, the mode depends on triaxial acceleration amplitude envelope extraction and detection threshold values, detection is easy to miss, interference caused by body movement is detected, interference of other factors such as ambient light, weak perfusion and the like actually exists, and a method based on feature extraction and machine learning or deep learning, the method depends on accuracy of an actual dataset label and generalization of a model, labeling time of the dataset is long, most of evaluation results only divide signals into usable or unusable, detailed classification is not carried out, specific signal anomaly types are given, practical use is limited, and in view of the fact, the method for evaluating the pulse wave signal quality by multi-parameter fusion is provided.
Disclosure of Invention
The invention aims to provide a multi-parameter fusion pulse wave signal quality evaluation method, which aims to solve the technical problems that the prior pulse oximeter is inconvenient to refine and classify when the output signal quality is evaluated, specific signal abnormality types are given, and practical use is limited.
In order to solve the technical problems, the invention provides a multi-parameter fusion pulse wave signal quality evaluation method, which comprises the following steps:
S1, collecting PPG signals and triaxial acceleration data;
synchronously collecting the red light, infrared light and triaxial acceleration raw data of pulse waves;
S2, PPG signal end point detection;
aiming at the signal saturation problem caused by the difference between an individual and a collecting part, endpoint detection is set, and a label is set according to the relation between the signal amplitude and the endpoint;
S3, extracting non-pulsation component characteristics of the PPG signals;
extracting non-pulsation component with large difference between wearing and unworn by using filter, and judging and setting label according to threshold value;
S4, extracting pulse component characteristics of the PPG signals;
extracting and processing the time window to obtain basic amplitude signalAnd amplitude characteristics, the label is set according to the threshold range
S5, detecting abnormal pulse components of the PPG signal;
at the basic amplitude signalOn the basis of (1) extracting a dynamic threshold toIn order to be a time window in which,Extracting the sum of the mean value and 3 times of standard deviation of data between upper and lower quartiles in each time window as the threshold value in each time window for the window moving length, traversing the whole signal, outputting pulse component amplitude threshold value, and setting a label according to the relation between the amplitude and the threshold value
S6, extracting PPG signal segmentation autocorrelation characteristics;
selecting time window segmentation signal according to pulse wave cycle characteristics, calculating a characteristic value according to autocorrelation of the time window segmentation signal, and calibrating a label according to a threshold value;
S7, extracting a Pearson correlation coefficient of the PPG signal template;
Detecting key points and outputting templates, calculating correlation coefficients based on the key point templates, and setting labels according to threshold values after processing;
S8, detecting three-axis acceleration body movement;
Extracting first order partial derivatives for each axis of triaxial dataCalculating first-order partial derivative vectorFiltering to obtain a composite vector envelopeSetting a label according to a threshold value;
S9, detecting the three-axis acceleration posture state change;
judging the state of the equipment according to the amplitude and the direction of the triaxial acceleration;
S10, detecting the three-axis acceleration body position change frequency;
Calculating frequency according to body position change time and scaling label according to threshold value;
S11, displaying signal abnormality;
When (when)When the signal abnormality type is prompted to be saturated in amplitude, the signal abnormality prompt priority is 1, whenWhen the signal abnormality type is not worn, the signal abnormality prompt priority is 2, whenWhen the signal abnormality type is indicated as low amplitude and priority is 3, whenWhen the signal abnormality type is indicated as that the amplitude is larger, the signal abnormality indication priority is 4, whenWhen the signal abnormality type is indicated as poor signal periodicity, the signal abnormality indication priority is 5, whenWhen the signal abnormality type is signal morphological abnormality, the signal abnormality prompt priority is 6, whenWhen the signal abnormality type is that the body movement amplitude is abnormal, the signal abnormality prompt priority is 7, whenWhen the signal abnormality type prompt is abnormal due to equipment state change, the signal abnormality prompt priority is 8, and the signal abnormality prompt priority displays the signal abnormality type with low priority.
The pulse wave and triaxial acceleration characteristics are deeply fused by 8 evaluation modes, so that various conditions such as signal saturation, equipment wearing abnormality, signal amplitude abnormality, periodicity abnormality, morphological characteristic abnormality, body movement amplitude abnormality and equipment state change frequency abnormality are comprehensively covered, and a comprehensive and accurate evaluation system is provided for the signal quality of the pulse oximeter. In the signal saturation detection, the signal effectiveness is accurately screened according to the set endpoint value of the AD sampling maximum value, the wearing state is judged according to the individual scene threshold value by extracting the characteristics through a specific filter during the non-pulsation component analysis, the signal quality is accurately judged in a multi-dimensional cooperation mode, a clinician is strongly assisted in accurately evaluating the output reliability of the pulse oximeter, potential problems are effectively eliminated, and the monitoring accuracy and efficiency are improved.
Preferably, in S1, red light, infrared light and triaxial accelerationA shaft(s),A shaft(s),The axis raw data are respectively expressed as、、、、Representing the PPG signal inThe representation is made of a combination of a first and a second color,Acceleration of three axesThe representation is made of a combination of a first and a second color,。
Preferably, in S2, the pulse wave signal acquired in S1 is subjected to end point detection, and an end point value based on the AD sampling maximum value is set、Outputting the label when the signal amplitude value is equal to the end point value0, Otherwise1, As shown in the following formula:
。
Preferably, in S3, a 4-order zero-phase IIR digital low-pass filter is adopted to filter interference components below 0.5Hz and above 5Hz, and red light and infrared light non-pulsating components are extracted, wherein the cut-off frequency of the zero-phase IIR digital low-pass filter is 0.1Hz;
Low frequency component between 0.5Hz and 5Hz is used as two characteristics in S3 toThe representation is made of a combination of a first and a second color,Setting threshold according to individual scene difference、When (when)Or (b)Output tag0, Otherwise1, As shown in the following formula:
。
Preferably, in S4, 1 second is taken as a fixed time window, the difference between the maximum value and the minimum value of the signals in each time window is extracted, and a basic amplitude signal is output,Wherein, the method comprises the steps of, wherein,AndRespectively representing the basic amplitude signals of red light and infrared light, then convoluting the basic amplitude signals with a rectangular window with the length of 5, and outputting the amplitude characteristics of the pulse components,Wherein, the method comprises the steps of, wherein,AndRespectively representing the amplitude characteristics of the pulse components of the red light and the infrared light, when the amplitude characteristics areAt the experimental thresholdWithin a range of, orAt the experimental thresholdOutput tag within rangeSet to 0, otherwise1, As shown in the following formula:
。
Preferably, in S5,Is that,Wherein, the method comprises the steps of, wherein,AndRepresenting the pulse component amplitude threshold value of the red light and the infrared light respectively, if the basic amplitude of the red light and the infrared light is within the pulse component amplitude threshold valueOtherwise, the current second signal is considered to be abnormal,Set to 0, otherwise1, As shown in the following formula:
。
Preferably, in S6, the signal is divided by taking 5 seconds as a time window and taking 1 second as a window moving length, the signal in each time window is subjected to autocorrelation, and the signal in a certain time window is set as,,For the index value of the signal,For the time window length, the autocorrelation calculation formula is as follows:
, is an autocorrelation signalIndependent variables of (2);
normalizing the formula to obtain a normalized autocorrelation valueThe following formula is shown:
Extracting autocorrelation characteristic values,,Quantitatively indicating the degree of periodicity of the pulse signal, the smaller the value, the better the periodicity, wherein、To normalize the post-autocorrelation peak points,The representation takes the maximum value in B, C, D;
the pulse wave signal periodicity is evaluated byGreater than a threshold value,Set to 0, otherwise1, As shown in the following formula:
;
The weighting strategy in the calculation strengthens the periodic component according to the dynamic weighting of the signal characteristics, improves the evaluation precision, and the weighted autocorrelation function is as followsWeighting ofLocal energy in dependence on signalAnd frequency characteristicsCalculations, e.g.Optimizing feature value extraction and periodic evaluation, wherein,、In order to adjust the parameters of the device,As a hysteresis parameter in the autocorrelation function,Representing pulse wave signals at discrete momentsIs used for the sampling value of (a),For the discrete-time index,To calculate the local energy of the signalAnd frequency characteristicsIs set for the window length parameter of the (c) window,Is pulse wave signalIs a function of the autocorrelation of (a).
Preferably, in S7, four key points of the pulse wave A, B, C, D are detected, the data centered on the four key point positions are averaged to output four key point templates, then Pearson correlation coefficients are calculated by using the key point templates and the data extracted centered on the key points, respectively, the Pearson correlation coefficients are used as independent variables, and nonlinear interpolation is performed by using the key point positions as dependent variables, so that the sampling rate of the interpolated Pearson correlation coefficients is the same as that of the original signal, and then downsampling processing is performed to finally output Pearson correlation coefficientsThe sampling rate being 1Hz, i.e. outputting a correlation coefficient value every 1 second, if,Set to 0, otherwise1, As shown in the following formula:
Wherein the method comprises the steps ofA decision threshold representing signal quality assessment using Pearson correlation coefficients;
In S8, the first order partial derivative extraction formula is as follows:
Wherein,Is the three-axis discrete data independent variable index value,Data representing each axis of the triaxial dataIs used for the first-order deflection guide of the (a),Representing the three-axis acceleration raw data inThe data on the axis of the device,、、Respectively representA shaft(s),Shaft and method for producing the sameA shaft;
Calculating the resultant vector of first-order partial derivativesThe calculation formula is as follows:
For a pair ofTaking 1 second as a filtering window length and 0.5 second as a window moving length, carrying out moving average filtering, and outputting the envelope of the combined vectorWhen (when)In the time-course of which the first and second contact surfaces,Set to 0, otherwise1, As shown in the following formula:
;
Wherein,The body movement determination threshold value is indicated,、AndRespectively is、、First order partial derivatives of the axis data.
Preferably, in S9, the device state is determined according to the magnitude and direction of the triaxial acceleration, and if the device deflects rightward, it is determined that:
;
Wherein,Indicating a decision that the device is deflected to the right,、AndRespectively represent the triaxial acceleration in、、Amplitude of the shaft;
In S10, at the first posture change time pointIf the time interval between the current position change time point and the previous position change time point is more than 20s as the starting point, the previous position change time point is taken as the position continuous change ending point and is recorded asNumber of changes of body position change starting point and ending point is usedRepresentation, body position change frequencyCalculated by the following formula:
;
The body position change frequency is greater than a certain threshold valueThenSet to 0, otherwise1, As shown in the following formula:
;
The multi-sensor data are fused to construct a dynamic attitude model, and a Kalman filtering algorithm is used for fusing gyroscope and accelerometer data into an attitude transfer matrixUpdating, improving the accuracy of body movement interference discrimination, wherein,,For the discrete-time index,The function is updated for the fusion algorithm,To be at the momentThe data of the gyroscope which is collected,To be at the momentAccelerometer data collected.
Compared with the prior art, the invention has the beneficial effects that:
1. According to the invention, through 8 signal quality evaluation modes, signal saturation, equipment wearing abnormality, signal amplitude abnormality, signal periodicity abnormality, signal morphological characteristic abnormality, body movement signal amplitude abnormality and equipment state change frequency abnormality are respectively related, according to the abnormality types, various characteristics of pulse waves and triaxial acceleration are fused, so that the pulse wave signal quality can be effectively evaluated, a clinician is assisted in evaluating the reliability of an output result of the pulse oximeter, and according to the corresponding signal abnormality types, the clinician can be helped to directly eliminate the corresponding problems, the monitoring efficiency is improved, and the problems that the existing pulse oximeter is inconvenient to refine classification, the specific signal abnormality types are given and the practical use is limited when the output signal quality of the pulse oximeter is evaluated are solved.
2. When the signal periodicity is determined, the signal periodicity is determined by normalizing the autocorrelation characteristic value, and compared with the signal periodicity determined by the duration of the peak interval, the signal periodicity determining method can effectively avoid the problem of periodicity determining abnormality caused by abnormal peak detection.
3. When the signal morphological characteristic abnormality judgment is carried out, the method extracts the plurality of feature point templates and Pearson correlation coefficients on a plurality of feature points of the pulse wave, increases the resolution of the extracted Pearson correlation coefficients, and can effectively reduce the false detection probability caused by the template extraction abnormality.
4. According to the method, the total vector is calculated on the data after the first-order difference of the three-axis original data, so that the influence of the baseline drift of the three-axis sensor on the body movement amplitude can be reduced, and the body movement detection accuracy is improved. And on the condition that whether the body movement interference exists, the frequency parameter of the change of the position state of the equipment is introduced, and whether the body movement interference exists is judged by changing the frequency, so that the interference to the pulse signal caused by the fact that the body movement amplitude is smaller but the position state of the equipment is changed can be effectively avoided.
Drawings
FIG. 1 is a flow chart of signal quality evaluation of a multi-parameter fusion pulse oximeter of the instant invention;
FIG. 2 is a diagram of the red and infrared light wearing and not wearing waveforms of the present invention;
FIG. 3 is a graph of a noise detection waveform of the infrared light in the range of the infrared light amplitude exceeding the normal threshold value;
FIG. 4 is a waveform diagram of the pulse wave autocorrelation and normalized autocorrelation of the present invention;
FIG. 5 is a waveform diagram of key points of pulse wave according to the present invention;
FIG. 6 is a schematic diagram of a three-axis accelerometer body motion detection according to the present invention;
fig. 7 is a schematic diagram of a three-axis acceleration posture state according to the present invention.
Detailed Description
In order to facilitate the understanding of the technical scheme of the present invention by those skilled in the art, the technical scheme of the present invention will be further described with reference to the accompanying drawings.
Example 1
The invention provides a pulse wave signal quality evaluation method with multi-parameter fusion, which comprises the following steps:
S1, collecting PPG signals and triaxial acceleration data;
synchronously collecting the red light, infrared light and triaxial acceleration raw data of pulse waves;
S2, PPG signal end point detection;
aiming at the signal saturation problem caused by the difference between an individual and a collecting part, endpoint detection is set, and a label is set according to the relation between the signal amplitude and the endpoint;
S3, extracting non-pulsation component characteristics of the PPG signals;
extracting non-pulsation component with large difference between wearing and unworn by using filter, and judging and setting label according to threshold value;
S4, extracting pulse component characteristics of the PPG signals;
extracting and processing the time window to obtain basic amplitude signalAnd amplitude characteristics, the label is set according to the threshold range
S5, detecting abnormal pulse components of the PPG signal;
at the basic amplitude signalOn the basis of (1) extracting a dynamic threshold toIn order to be a time window in which,Extracting the sum of the mean value and 3 times of standard deviation of data between upper and lower quartiles in each time window as the threshold value in each time window for the window moving length, traversing the whole signal, outputting pulse component amplitude threshold value, and setting a label according to the relation between the amplitude and the threshold value
S6, extracting PPG signal segmentation autocorrelation characteristics;
selecting time window segmentation signal according to pulse wave cycle characteristics, calculating a characteristic value according to autocorrelation of the time window segmentation signal, and calibrating a label according to a threshold value;
S7, extracting a Pearson correlation coefficient of the PPG signal template;
Detecting key points and outputting templates, calculating correlation coefficients based on the key point templates, and setting labels according to threshold values after processing;
S8, detecting three-axis acceleration body movement;
Extracting first order partial derivatives for each axis of triaxial dataCalculating first-order partial derivative vectorFiltering to obtain a composite vector envelopeSetting a label according to a threshold value;
S9, detecting the three-axis acceleration posture state change;
judging the state of the equipment according to the amplitude and the direction of the triaxial acceleration;
S10, detecting the three-axis acceleration body position change frequency;
Calculating frequency according to body position change time and scaling label according to threshold value;
S11, displaying signal abnormality;
When (when)When the signal abnormality type is prompted to be saturated in amplitude, the signal abnormality prompt priority is 1, whenWhen the signal abnormality type is not worn, the signal abnormality prompt priority is 2, whenWhen the signal abnormality type is indicated as low amplitude and priority is 3, whenWhen the signal abnormality type is indicated as that the amplitude is larger, the signal abnormality indication priority is 4, whenWhen the signal abnormality type is indicated as poor signal periodicity, the signal abnormality indication priority is 5, whenWhen the signal abnormality type is signal morphological abnormality, the signal abnormality prompt priority is 6, whenWhen the signal abnormality type is that the body movement amplitude is abnormal, the signal abnormality prompt priority is 7, whenWhen the signal abnormality type prompt is abnormal due to equipment state change, the signal abnormality prompt priority is 8, and the signal abnormality prompt priority displays the signal abnormality type with low priority.
In the embodiment of the invention, in S1, red light, infrared light and triaxial accelerationA shaft(s),A shaft(s),The axis raw data are respectively expressed as、、、、Representing the PPG signal inThe representation is made of a combination of a first and a second color,Acceleration of three axesThe representation is made of a combination of a first and a second color,。
In the embodiment of the invention, in S2, in practice, the alternating current and direct current components of different individual pulse waves are different, and the acquired signal intensities of different parts are also different, so that the numerical range of an actual output signal is fixed when the device performs AD sampling, automatic dimming is not involved in part of device firmware, signal saturation is easy to cause, and the acquired signals are invalid values when the signals are saturated. In order to avoid signal abnormality caused by the difference between an individual with larger AC/DC component or an acquisition part and an acquisition part in hardware design, the pulse wave signal acquired in S1 is subjected to end point detection, and an end point value based on the AD sampling maximum value is set、Outputting the label when the signal amplitude value is equal to the end point value0, Otherwise1, As shown in the following formula:
。
in the embodiment of the invention, as shown in fig. 2, in S3, the wearing and the unworn of the non-pulsating component have larger difference in numerical distribution range, the frequency of the non-pulsating component is lower, the frequency range of the actual pulse wave is 30-240 times/min, therefore, a 4-order, zero-phase IIR digital low-pass filter is adopted to filter the interference component below 0.5Hz and above 5Hz, the interference of the lowest respiratory frequency to the non-pulsating component is effectively filtered, and the non-pulsating components of red light and infrared light are extracted, wherein the cut-off frequency of the zero-phase IIR digital low-pass filter is 0.1Hz;
Low frequency component between 0.5Hz and 5Hz is used as two characteristics in S3 toThe representation is made of a combination of a first and a second color,Setting threshold according to individual scene difference、When (when)Or (b)Output tag0, Otherwise1, As shown in the following formula:
。
The threshold value selection is determined according to the difference between different scenes of different individuals and different scenes of different individuals, wherein the minimum value of the non-pulsation component is shown in the table 1 when the different individuals wear, the maximum value of the non-pulsation component is shown in the table 2 when the different scenes are not worn, the simulation scenes mainly comprise scenes such as sensor backlight, shading, direct illumination of backlight mobile phone light, direct illumination of shading mobile phone light and the like, the minimum value of the actual individuals do not overlap with the maximum value of the different scenes when the different scenes are not worn, the determination threshold value is determined, and interference when the individuals do not wear is eliminated, and the following table is attached:
in the embodiment of the present invention, in S4, 1 second is taken as a fixed time window, the difference between the maximum and minimum values of the signals in each time window is extracted, and a basic amplitude signal is output,Wherein, the method comprises the steps of, wherein,AndRespectively representing the basic amplitude signals of red light and infrared light, then convoluting the basic amplitude signals with a rectangular window with the length of 5, and outputting the amplitude characteristics of the pulse components,Wherein, the method comprises the steps of, wherein,AndRespectively representing the amplitude characteristics of the pulse components of the red light and the infrared light, wherein the amplitude of the normal pulse wave signal is obviously different from that of the noise signal when the pulse wave signal is not worn, and the pulse wave signal is distinguished from the noise signal when the pulse wave signal is not wornAt the experimental thresholdWithin a range of, orAt the experimental thresholdOutput tag within rangeSet to 0, otherwise1, As shown in the following formula:
。
In an embodiment of the present invention, in S5,Is that,Wherein, the method comprises the steps of, wherein,AndRepresenting the pulse component amplitude threshold value of the red light and the infrared light respectively, if the basic amplitude of the red light and the infrared light is within the pulse component amplitude threshold valueOtherwise, the current second signal is considered to be abnormal,Set to 0, otherwise1, As shown in the following formula:
。
In the embodiment of the present invention, in S6, according to the frequency range of the actual pulse wave signal, the longest period of the signal is 2S, so as to ensure that at least 2 pulse periods exist in each time window, the signal is divided by taking 5 seconds as the time window and 1 second as the window moving length, the signal in each time window is subjected to autocorrelation, and the signal in a certain time window is set as the signal in the certain time window,,For the index value of the signal,For the time window length, the autocorrelation calculation formula is as follows:
, is an autocorrelation signalIndependent variables of (2);
The autocorrelation within a certain time window is shown in FIG. 4, and the signal has four effective peak points, respectively;
Normalizing the formula to obtain a normalized autocorrelation valueThe following formula is shown:
The normalization method can effectively avoid the difference of autocorrelation amplitude caused by the difference of zero padding number, and the normalized autocorrelation waveform in a certain time window is shown in figure 4, in the figureIs the normalized autocorrelation peak point;
Extracting autocorrelation characteristic values,Quantitatively indicating the degree of periodicity of the pulse signal, the smaller the value, the better the periodicity, wherein、To normalize the post-autocorrelation peak points,The representation takes the maximum value in B, C, D;
the pulse wave signal periodicity is evaluated byGreater than a threshold value,Set to 0, otherwise1, As shown in the following formula:
;
The weighting strategy in the calculation strengthens the periodic component according to the dynamic weighting of the signal characteristics, improves the evaluation precision, and the weighted autocorrelation function is as followsWeighting ofLocal energy in dependence on signalAnd frequency characteristicsCalculations, e.g.Optimizing feature value extraction and periodic evaluation, wherein,、In order to adjust the parameters of the device,As a hysteresis parameter in the autocorrelation function,Representing pulse wave signals at discrete momentsIs used for the sampling value of (a),For the discrete-time index,To calculate the local energy of the signalAnd frequency characteristicsIs set for the window length parameter of the (c) window,Is pulse wave signalIs a function of the autocorrelation of (a).
In the embodiment of the invention, the key points of the pulse wave mainly have 4 positions, as shown in fig. 5, in S7, four key points of the pulse wave A, B, C, D are detected, the point a is the start point of the pulse wave, the point B is the main peak point, the point C is the down-center isthmus point, the point D is the dead-beat point, the data centered on the four key point positions are averaged, four key point templates are output, then the Pearson correlation coefficient is calculated by the key point templates and the data extracted centered on the key points respectively, the key point positions are taken as independent variables, the Pearson correlation coefficient is taken as dependent variable, nonlinear interpolation is performed to ensure that the sampling rate of the Pearson correlation coefficient after interpolation is the same as the sampling rate of the original signal, and then downsampling processing is performed to ensure that the Pearson correlation coefficient finally outputThe sampling rate being 1Hz, i.e. outputting a correlation coefficient value every 1 second, if,Set to 0, otherwise1, As shown in the following formula:
Wherein the method comprises the steps ofA decision threshold representing signal quality assessment using Pearson correlation coefficients;
In general, signal quality evaluation is carried out through a template, namely, a single pulse wave key position point is mainly used for extracting the template, the template is easy to extract abnormality, the sampling rate of extracting a correlation coefficient is low, the error is large, and the false detection probability table 3 caused by the abnormality of template extraction can be effectively reduced by adding the key point position and extracting the correlation coefficient;
In S8, the first order partial derivative extraction formula is as follows:
Wherein,Is the three-axis discrete data independent variable index value,Data representing each axis of the triaxial dataIs used for the first-order deflection guide of the (a),Representing the three-axis acceleration raw data inThe data on the axis of the device,、、Respectively representA shaft(s),Shaft and method for producing the sameA shaft;
Calculating the resultant vector of first-order partial derivativesThe calculation formula is as follows:
For a pair ofTaking 1 second as a filtering window length and 0.5 second as a window moving length, carrying out moving average filtering, and outputting the envelope of the combined vectorWhen (when)In the time-course of which the first and second contact surfaces,Set to 0, otherwise1, As shown in the following formula:
;
The body movement determination threshold value is indicated,、AndRespectively is、、The first-order deviation guide of the axis data can reduce the influence of the baseline drift of the triaxial sensor on the body movement amplitude by calculating the sum vector of the data after the first-order difference of the triaxial original data, improve the body movement detection accuracy and realize the body movement detection output as shown in figure 6.
In the embodiment of the invention, in S9, in sleep monitoring, the sleeping postures mainly include five sleeping postures of supine position, prone position, left side lying, right side lying and standing, and when the sleeping postures change, the pulse wave signals are necessarily disturbed. Therefore, according to the magnitude and the direction of each axis of the triaxial acceleration, the state of the equipment is judged, and if the equipment deflects rightwards, the judgment is that:
;
Wherein,Indicating a decision that the device is deflected to the right,、AndRespectively represent the triaxial acceleration in、、Amplitude of the shaft;
FIG. 7 is a three axis acceleration body position state, S, P, L, R, up/Down in the figure corresponding to supine, prone, left lateral, right lateral and upright positions, respectively;
In S10, at the first posture change time pointIf the time interval between the current position change time point and the previous position change time point is more than 20s as the starting point, the previous position change time point is taken as the position continuous change ending point and is recorded asNumber of changes of body position change starting point and ending point is usedRepresentation, body position change frequencyCalculated by the following formula:
;
The body position change frequency is greater than a certain threshold valueThenSet to 0, otherwise1, As shown in the following formula:
;
as shown in FIG. 7, the frequency of change is 7.38 times/second between 2617.53 seconds and 2630 seconds, so if the threshold is determinedIf the value is set to be 5, judging that the body movement interference exists in the time period;
The multi-sensor data are fused to construct a dynamic attitude model, and a Kalman filtering algorithm is used for fusing gyroscope and accelerometer data into an attitude transfer matrixUpdating, improving the accuracy of body movement interference discrimination, wherein,,For the discrete-time index,The function is updated for the fusion algorithm,To be at the momentThe data of the gyroscope which is collected,To be at the momentAccelerometer data collected.
The device outputs the signal quality label according to the algorithm, displays the corresponding signal abnormality type, and as shown in table 3, the clinician can judge whether the current output value is credible or not according to the displayed signal abnormality type, and can adjust the monitoring state through the output signal abnormality type so as to acquire a signal with better quality. The signal prompt priority in the table refers to that when the signal meets both the condition of the signal quality evaluation 1 and the condition of the signal quality evaluation 3, the signal abnormality prompt displays the signal abnormality type in the signal quality evaluation 1, namely, the signal abnormality type with smaller signal quality evaluation value.
The embodiments of the present invention are disclosed as preferred embodiments, but not limited thereto, and those skilled in the art will readily appreciate from the foregoing description that various modifications and variations can be made without departing from the spirit of the present invention.