Patient multi-source monitoring data fusion analysis system based on time sequence analysis

Abstract

The invention relates to the technical field of data fusion, in particular to a patient multi-source monitoring data fusion analysis system based on time sequence analysis, the system comprises a processor and a memory, and the processor executes a computer program stored in the memory to realize the following steps: obtaining a target signal segment of a physical sign data signal, obtaining a decomposition layer number corresponding to the target signal segment according to the frequency domain signal corresponding to the target signal segment, and carrying out wavelet transform on the corresponding target signal segment according to the decomposition layer number corresponding to the target signal segment to obtain all component signals corresponding to the target signal segment; and performing inverse transformation on the component signals corresponding to the target signal segments of all the physical sign data signals in the physical sign data signal set to obtain a target fusion signal. And the accuracy and the reliability during signal decomposition can be improved, so that the reference value and the reliability of a fusion result are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data fusion, and particularly to a multi-source monitoring data fusion analysis system for patients based on time series analysis. Background Art

[0002] In the modern medical field, physiological sign data of patients is monitored through medical devices or wearable devices. For example, various sign data such as electrocardiogram (ECG), blood oxygen saturation (SpO 2 )), blood pressure (BP), respiratory rate, etc. can be collected. Then, by fusing the various sign data of the monitored patients, the accuracy, timeliness, and safety of medical services are ensured. For example, by fusing the various sign data of the monitored patients, an individualized health model can be established to guide precise medication, surgical plan optimization, etc. Therefore, it is crucial to fuse the various sign data of the monitored patients.

[0003] In the prior art, the globally wavelet transform is usually performed on the monitored sign data signal using a fixed decomposition level, and then the inverse transform is performed on the signal components obtained by the wavelet transform to obtain the fusion result. However, this way of performing the globally wavelet transform on the monitored sign data signal using a fixed decomposition level will result in problems such as low accuracy and reliability of the decomposed signal components. Then, the reference value and reliability of the obtained fusion result will also be poor. For example, when a certain segment of the sign data signal to be transformed is a low-fluctuation signal, if a higher level is used during the transformation, it will cause over-filtering, resulting in the loss of key feature information. When a certain segment of the sign data signal to be transformed is a high-fluctuation signal, if a lower level is used during the transformation, it may result in the phenomenon that high-frequency information cannot be fully decomposed. Therefore, how to improve the accuracy and reliability when performing the wavelet transform on the monitored sign data signal to improve the reference value and reliability of the fusion result has become an urgent problem to be solved. Summary of the Invention

[0004] In order to solve the above problems, the present invention provides a multi-source monitoring data fusion analysis system for patients based on time series analysis, and the specific technical solutions adopted are as follows:

[0005] An embodiment of the present invention provides a multi-source monitoring data fusion analysis system for patients based on time series analysis, including a processor and a memory. The processor executes the computer program stored in the memory to implement the following steps:

[0006] Obtain a set of sign data signals of a patient;

[0007] According to the extreme points in each physiological sign data signal in the physiological sign data signal set and the time intervals between adjacent extreme points, obtain the data stability characterization value of the physiological sign data signal. According to the data stability characterization value, the total number of physiological sign data points, and the total number of extreme points on the physiological sign data signal, segment the physiological sign data signal to obtain the target signal segments of the physiological sign data signal;

[0008] According to the frequency domain signal corresponding to the target signal segment, obtain the decomposition level corresponding to the target signal segment, and perform wavelet transform on the corresponding target signal segment according to the decomposition level corresponding to the target signal segment to obtain all component signals corresponding to the target signal segment;

[0009] Perform inverse transform on the component signals corresponding to the target signal segments of all physiological sign data signals in the physiological sign data signal set to obtain the target fusion signal.

[0010] Beneficial effects: The present invention first obtains the physiological sign data signal set of the patient; then, according to the extreme points in each physiological sign data signal in the physiological sign data signal set and the time intervals between adjacent extreme points, obtains the data stability characterization value of the physiological sign data signal. According to the data stability characterization value, the total number of physiological sign data points, and the total number of extreme points on the physiological sign data signal, segments the physiological sign data signal to obtain the target signal segments of the physiological sign data signal; then, according to the frequency domain signal corresponding to the target signal segment, obtains the decomposition level corresponding to the target signal segment, and performs wavelet transform on the corresponding target signal segment according to the decomposition level corresponding to the target signal segment to obtain all component signals corresponding to the target signal segment; finally, performs inverse transform on the component signals corresponding to the target signal segments of all physiological sign data signals in the physiological sign data signal set to obtain the target fusion signal. And through the process of segmenting the physiological sign data signal and then adaptively obtaining the decomposition level corresponding to the target signal segment based on the frequency domain signal corresponding to each target signal segment, the present invention can improve the accuracy and reliability during subsequent signal decomposition, thereby improving the reference value and reliability of the fusion result. Description of the Drawings

[0011] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0012] Figure 1 It is a flowchart of a method for fusing and analyzing multi-source monitoring data of patients based on time series analysis according to the present invention. Detailed Embodiments

[0013] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention belong to the scope protected by the embodiments of the present invention.

[0014] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.

[0015] This embodiment provides a system for fusing and analyzing multi-source monitoring data of patients based on time series analysis, including a processor and a memory. The processor executes the computer program stored in the memory to implement a method for fusing and analyzing multi-source monitoring data of patients based on time series analysis, as Figure 1 shown. The method for fusing and analyzing multi-source monitoring data of patients based on time series analysis includes the following steps:

[0016] Step S001, obtaining a set of physical sign data signals of a patient.

[0017] In this embodiment, the physical sign data signals are adaptively segmented mainly by analyzing the characteristics of different physical sign data signals, and then each signal segment obtained by the division is further analyzed to obtain the decomposition layer corresponding to each signal segment. Wavelet transform is performed based on the decomposition layer corresponding to each signal segment, and an accurate fusion result is obtained based on the transformation result. And for the convenience of understanding and analysis, only the fusion process of the physical sign data of any one patient is described in this embodiment, that is, all the physical sign data that appears subsequently comes from the same patient.

[0018] Preferably, during the monitoring period, a medical device or a wearable device is used to monitor the physiological sign data of a patient, and the sign data at different monitoring times is obtained. All the sign data of the same sign data type monitored during the monitoring period is used to construct a time series sequence, which is denoted as the sign data sequence corresponding to the corresponding sign data type. That is, the sign data types in different sign data sequences are different, and the sign data types in the same sign data sequence are the same. Moreover, in this embodiment, the acquisition frequencies of all types of sign data are the same and belong to synchronous acquisition. In addition, in this embodiment, the device for monitoring the sign data of the patient can be a new type of wearable monitor or a patch-type dynamic monitor, etc. The new type of wearable monitor uses non-invasive sensors and can synchronously collect data such as electrocardiogram, respiration, blood oxygen, and pulse, and wirelessly transmit the data to the central monitoring system, which is suitable for continuous monitoring of inpatients. The patch-type dynamic monitor can continuously collect electrocardiogram waveforms, respiration frequencies, etc., and is suitable for intensive care or postoperative rehabilitation scenarios. Moreover, the length of the monitoring period and the sign data acquisition frequency when monitoring the sign data of the patient need to be set by relevant staff according to the monitoring purpose, patient status, etc.

[0019] After obtaining the sign data sequence corresponding to the sign data type, the sign data points corresponding to each sign data in the sign data sequence are obtained, and the signal formed by the sign data points corresponding to each sign data in the sign data sequence is denoted as the sign data signal corresponding to the corresponding sign data sequence. That is, the number of sign data points on the sign data signal corresponding to any sign data sequence is the same as the number of sign data in the corresponding sign data sequence. The abscissa value of the sign data point corresponding to any sign data is the acquisition time of the corresponding sign data, and the ordinate value is the corresponding sign data. The abscissa axis of the space where the sign data signal corresponding to any sign data sequence is located is time, and the ordinate axis is the sign data in the corresponding sign data sequence. In addition, the sign data types in this embodiment include but are not limited to electrocardiogram, respiration, blood oxygen, pulse, etc. As other real-time methods, the sign data types to be monitored can also be determined based on the final analysis purpose. For example, when performing autonomic function assessment and cardiopulmonary disease diagnosis, in order to ensure the reliability of the assessment and diagnosis, the respiration frequency signal and the electrocardiogram signal collected are often fused, and then the fused data is analyzed to achieve high-accuracy autonomic function assessment and cardiopulmonary disease diagnosis.

[0020] After obtaining the sign data signals corresponding to all sign data sequences, the set formed by the sign data signals corresponding to all sign data sequences is denoted as the sign data signal set of the patient, and the sign data signals are all time-domain signals.

[0021] Step S002: Obtain the data stability characterization value of the vital sign data signal based on the extreme points and the time intervals between adjacent extreme points in each vital sign data signal in the set of vital sign data signals. Segment the vital sign data signal based on the data stability characterization value, the total number of vital sign data points, and the total number of extreme points on the vital sign data signal to obtain the target signal segments of the vital sign data signal.

[0022] Since when a certain segment of the vital sign data signal to be transformed is a low-fluctuation signal, using a higher number of layers during the transformation may lead to over-filtering, resulting in the loss of key feature information. When a certain segment of the vital sign data signal to be transformed is a high-fluctuation signal, using a lower number of layers during the transformation may cause the phenomenon that high-frequency information cannot be fully decomposed. Both of the above phenomena will affect the accuracy and reliability of the wavelet transform, thus making the reference value and reliability of the finally obtained fusion result relatively low. To minimize the negative consequences caused by the above phenomena as much as possible, in the following, this embodiment will segment the vital sign data signals based on the characteristics of different vital sign data signals to obtain the respective target signal segments on each vital sign data signal. That is, in the following, this embodiment will first obtain the data stability characterization value of each vital sign data signal based on the extreme points and the time intervals between adjacent extreme points in each vital sign data signal in the set of vital sign data signals. The data stability characterization value is an important parameter for reliably segmenting the vital sign data signal in the subsequent process. Then, the specific process of obtaining the data stability characterization value of each vital sign data signal in this embodiment is as follows:

[0023] For any physiological sign data signal: First, use the derivative method to obtain all the extreme points on the physiological sign data signal, and sort all the maximum points on the physiological sign data signal in the order of acquisition time to obtain the maximum point sequence on the physiological sign data signal, and sort all the minimum points on the physiological sign data signal in the order of acquisition time to obtain the minimum point sequence on the physiological sign data signal; Then, obtain the standard deviation of the ordinates of all the maximum points in the maximum point sequence, and denote it as the first standard deviation, obtain the standard deviation of the ordinates of all the minimum points in the minimum point sequence, and denote it as the second standard deviation, and take the maximum standard deviation of the first standard deviation and the second standard deviation as the first characterization value; After that, obtain the minimum value time interval sequence and the maximum value time interval sequence of the physiological sign data signal, and denote the standard deviation of the minimum value time interval sequence as the third standard deviation, denote the standard deviation of the maximum value time interval sequence as the fourth standard deviation, and take the maximum standard deviation of the third standard deviation and the fourth standard deviation as the second characterization value, and the f-th minimum value time interval on the minimum value time interval sequence is the absolute value of the difference in the abscissa between the f-th minimum point and the (f + 1)-th minimum point on the physiological sign data signal, and the g-th maximum value time interval on the maximum value time interval sequence is the absolute value of the difference in the abscissa between the g-th maximum point and the (g + 1)-th maximum point on the physiological sign data signal; Finally, obtain the sum of the first characterization value and the second characterization value, and use the exponential function to perform a negative correlation mapping on the sum of the first characterization value and the second characterization value, and take the result of the mapping as the data stability characterization value of the physiological sign data signal, that is, the data stability characterization value of the physiological sign data signal is exp(-S), where S is the sum of the first characterization value and the second characterization value, and exp() is the exponential function with the natural constant e as the base; And when the first characterization value is larger, it indicates that the stability of the change amplitude of the physiological sign data signal in time series is worse, and when the second characterization value is larger, it indicates that the stability of the change frequency of the physiological sign data signal in time series is also worse, that is, when the first characterization value and the second characterization value are larger or the data stability characterization value of the physiological sign data signal is smaller, it indicates that the stability of the physiological sign data signal is worse, the fluctuation characteristics are more obvious, or the local change of the physiological sign data signal may have a larger difference compared with the overall change. Then, at this time, in order to improve the accuracy of subsequent signal decomposition, the physiological sign data signal should be adaptively segmented, and when the first characterization value and the second characterization value are smaller or the data stability characterization value of the physiological sign data signal is larger, it indicates that the physiological sign data signal is more stable and the fluctuation characteristics are less obvious. At this time, not segmenting the physiological sign data signal can also make the subsequent decomposition result more accurate.

[0024] Based on the above description, after obtaining the data stability characterization value of each vital sign data signal, segmentation is performed based on the data stability characterization value. That is, in the following embodiments, the target signal segments of the vital sign data signals in the vital sign data signal set will be obtained according to the data stability characterization value of the vital sign data signals in the vital sign data signal set, the total number of vital sign data points, and the total number of extreme value points on the vital sign data signals. The specific process of obtaining the target signal segments of the vital sign data signals in the vital sign data signal set is as follows:

[0025] For any vital sign data signal A in the vital sign data signal set: First, determine whether the data stability characterization value of the vital sign data signal A is greater than the preset stability threshold. If so, it indicates that there is no need to segment the vital sign data signal A at this time. However, for the convenience of subsequent description, the vital sign data signal A can be directly used as the target signal segment of the vital sign data signal A at this time. That is, if the data stability characterization value of the vital sign data signal A is greater than the preset stability threshold, then there is only one target signal segment of the vital sign data signal A at this time; otherwise, according to the data stability characterization value of the vital sign data signal A, the total number of vital sign data points, and the total number of extreme value points on the vital sign data signal A, the initial signal segments on the vital sign data signal A are obtained, and the set constructed by all the initial signal segments on the vital sign data signal A is denoted as the initial signal segment set corresponding to the vital sign data signal A. The target signal segment of the vital sign data signal A is obtained according to the clustering metric distance between any two initial signal segments in the initial signal segment set corresponding to the vital sign data signal A; and in this embodiment, the implementer needs to set the preset stability threshold according to the actual situation. For example, in this embodiment, the preset stability threshold is set to 0.65.

[0026] In this embodiment, the specific process of obtaining the initial signal segment on the vital sign data signal A according to the data stability characterization value of the vital sign data signal A, the total number of vital sign data points, and the total number of extreme points on the vital sign data signal A is as follows: First, count the number of minimum points and maximum points on the vital sign data signal A, and denote them as the first quantity value and the second quantity value respectively. Then, obtain the maximum quantity value among the first quantity value and the second quantity value, and denote it as the initial extreme value quantity value. After that, calculate the sum of the initial extreme value quantity value and a preset first constant, and denote it as the target extreme value quantity value. In this embodiment, the preset first constant is set to 0.001, and its main function is to prevent the denominator from being zero. Immediately afterwards, obtain the ratio of the total number of vital sign data points on the vital sign data signal A to the target extreme value quantity value, and denote the product of the ratio of the total number of vital sign data points on the vital sign data signal A to the target extreme value quantity value and the data stability characterization value of the vital sign data signal A as the initial change frequency characterization value. Then, according to the value range of the initial change frequency characterization value and the mapping function, perform a positive correlation mapping on the change frequency characterization value, and take the integer value obtained by multiplying the mapping result by the acquisition frequency of the vital sign data points on the vital sign data signal A as the target sliding window length corresponding to the vital sign data signal A. Take the window with a length of the target sliding window length and a width of 1 as the target sliding window of the vital sign data signal A, and use the target sliding window of the vital sign data signal A to overlap and divide the vital sign data signal A to obtain the initial signal segment on the vital sign data signal A. The specific calculation expression for the target sliding window length corresponding to the vital sign data signal A is: where C A is the target window length corresponding to the vital sign data signal A, is the upward rounding symbol, range() is the mapping function, Q A is the data stability characterization value of the vital sign data signal A, M1 is the total number of vital sign data points on the vital sign data signal A, M2 is the target extreme value quantity value, H C is the acquisition frequency of the vital sign data points on the vital sign data signal A, H C is also the acquisition frequency of the vital sign data in the vital sign data sequence corresponding to the vital sign data signal A. And if the acquisition frequency of the vital sign data points on the vital sign data signal A is larger and the number of data points it contains is more, then when sliding, a larger window is required to speed up the sliding speed. Therefore, it is used as the mapping constant term for the window size when sliding the current analysis data.

[0027] In addition, when Q A is smaller, it indicates that the overall change of the vital sign data signal A is more unstable. Then, the selected window size should be smaller, so as to facilitate the more refined extraction of abnormal signal segments or fluctuating signal segments. The smaller the value, the greater the change frequency of the physical sign data signal A. Then, the selected window size should be smaller to improve the resolution during data analysis. Conversely, when Q A is larger and is larger, a larger window is used for sliding to reduce the computational amount of the currently analyzed data and to obtain the change characteristics of sufficient data within each window, thereby improving the accuracy of subsequent analysis. And in this embodiment, the mapping range of the mapping function range() needs to be set according to the actual situation. For example, in this embodiment, range() is set to (0.5, 3), and the mapping is a positive correlation mapping, that is, when is smaller, the result mapped by the mapping function range() is also smaller. Moreover, based on the above description, it can be seen that when is smaller, the obtained window length is smaller, and when is larger, the obtained window length is larger.

[0028] In addition, it should be noted that the values of the target sliding window length and width both represent the number of physical sign data points that can be accommodated in the corresponding window. That is, if the target sliding window length and width are 3 and 1 respectively, then the maximum number of physical sign data points that can be accommodated in this target sliding window is 3.

[0029] In this embodiment, the specific process of overlapping and dividing the vital sign data signal A by using the target sliding window of the vital sign data signal A to obtain the initial signal segments on the vital sign data signal A is as follows: First, obtain the preset sliding step N of the target sliding window of the vital sign data signal A, that is, the sliding step of the target sliding window of the vital sign data signal A is N, and in this embodiment, the implementer needs to set the value of N according to the actual situation, but it is required that the value of N is less than the length of the target sliding window of the vital sign data signal A. For example, in this embodiment, the value of N is set to the ceiling value of 80% of the target sliding window of the vital sign data signal A; then, denote the vital sign data sequence corresponding to the vital sign data signal A as sequence A0, and denote the target sliding window of the vital sign data signal A as window Z. First, place the leftmost side of window Z at the leftmost side of sequence A0 and denote it as the first window on sequence A0. Then, slide the first window on sequence A0 to the right by N positions to obtain the second window on sequence A0, slide the second window on sequence A0 to the right by N positions to obtain the third window on sequence A0, and so on, until the sliding of the window stops when the last vital sign data in sequence A0 first appears in the window, and count all the windows on sequence A0; for any window on sequence A0, denote the time interval formed by the acquisition times of all the vital sign data in this window as the time interval corresponding to this window, that is, the minimum value in the time interval corresponding to this window is the acquisition time of the first vital sign data in this window, and the maximum value in the time interval corresponding to this window is the acquisition time of the last vital sign data in this window. Then, on the vital sign data signal A, obtain the signal segment with the abscissa range being the time interval corresponding to this sliding window, and denote it as an initial signal segment on the vital sign data signal A. That is, the number of initial signal segments on sequence A0 is the same as the number of windows on sequence A0. The h-th initial signal segment on sequence A0 is obtained from the h-th window on sequence A0, and the first data in the first window on sequence A0 is the first vital sign data in sequence A0.

[0030] In this embodiment, the specific process of obtaining the target signal segment of the physical sign data signal A according to the clustering metric distance between any two initial signal segments in the initial signal segment set corresponding to the physical sign data signal A is as follows: Obtain the clustering metric distance between any two initial signal segments in the initial signal segment set, and cluster all the initial signal segments in the initial signal segment set according to the clustering metric distance between any two initial signal segments in the initial signal segment set to obtain clustering clusters, and the change ranges, coefficients of variation, and fluctuation conditions of the initial signal segments belonging to the same clustering cluster are similar; then determine whether the first initial signal segment and the second initial signal segment on the physical sign data signal A belong to the same clustering cluster. If so, continue to determine whether the second initial signal segment and the third initial signal segment on the physical sign data signal belong to the same clustering cluster. If not, record the time interval corresponding to the overlapping signal segment between the second initial signal segment and the third initial signal segment as the first overlapping time interval, and use the moment corresponding to the middle position of the first overlapping time interval as the first segmentation moment on the physical sign data signal A, and record the signal segment corresponding to the starting moment of the physical sign data signal A to the first segmentation moment as the first target signal segment of the physical sign data signal A. Then continue to determine whether the third initial signal segment and the fourth initial signal segment on the physical sign data signal belong to the same clustering cluster. If so, continue to determine whether the fourth initial signal segment and the fifth initial signal segment on the physical sign data signal belong to the same clustering cluster. If not, record the time interval corresponding to the overlapping signal segment between the fourth initial signal segment and the fifth initial signal segment as the second overlapping time interval, and use the moment corresponding to the middle position of the second overlapping time interval as the second segmentation moment on the physical sign data signal A, and record the signal segment corresponding to the first segmentation moment of the physical sign data signal A to the second segmentation moment as the second target signal segment of the physical sign data signal A, and so on. Continue to traverse the initial signal segments on the physical sign data signal A until the last initial signal segment on the physical sign data signal A is traversed and then stop, and count to obtain all the target signal segments of the physical sign data signal A. And the signal segment corresponding to the starting moment of the physical sign data signal A to the first segmentation moment refers to the signal segment corresponding to the time interval formed by the starting moment of the physical sign data signal A to the first segmentation moment, that is, the time interval of the signal segment corresponding to the starting moment of the physical sign data signal A to the first segmentation moment is the time interval formed by the starting moment of the physical sign data signal A to the first segmentation moment; in addition, in this embodiment, only the first target signal segment of the physical sign data signal A is closed, and the other target signal segments of the physical sign data signal A are left-open and right-closed.

[0031] In this embodiment, the process of obtaining the clustering metric distance between any two initial signal segments in the initial signal segment set is as follows: For the b-th initial signal segment and the c-th initial signal segment in the initial signal segment set, where b is not equal to c: First, obtain the difference between the ordinate of the maximum point and the ordinate of the minimum point in the b-th initial signal segment, and denote it as the fluctuation amplitude of the b-th initial signal segment. Obtain the difference between the ordinate of the maximum point and the ordinate of the minimum point in the c-th initial signal segment, and denote it as the fluctuation amplitude of the c-th initial signal segment. Obtain the normalized value of the squared difference between the fluctuation amplitude of the b-th initial signal segment and the fluctuation amplitude of the c-th initial signal segment, and denote it as the first difference. Then, obtain the coefficient of variation of the b-th initial signal segment and the coefficient of variation of the c-th initial signal segment, and denote the normalized value of the squared difference between the coefficient of variation of the b-th initial signal segment and the coefficient of variation of the c-th initial signal segment as the second difference. The coefficient of variation of an initial signal segment is the ratio of the standard deviation to the mean of the ordinates of all data points on the corresponding initial signal segment, which can reflect the degree of dispersion. After that, obtain the normalized value of the squared difference between the number of extreme points on the b-th initial signal segment and the number of extreme points on the c-th initial signal segment, and denote it as the third difference, which can reflect the difference in change frequencies. Finally, obtain the square root of the result obtained by adding the first difference, the second difference, and the third difference, and use it as the clustering metric distance between the b-th initial signal segment and the c-th initial signal segment. The specific calculation formula is:

[0032]

[0033] where D(b, c) is the clustering metric distance between the b-th initial signal segment and the c-th initial signal segment, Norm() is the normalization function, R b is the fluctuation amplitude of the b-th initial signal segment, R c is the fluctuation amplitude of the c-th initial signal segment, Y b is the coefficient of variation of the b-th initial signal segment, Y c is the coefficient of variation of the c-th initial signal segment, N b is the number of extreme points on the b-th initial signal segment, N c is the number of extreme points on the c-th initial signal segment; and when D(b, c) is larger, it indicates that the clustering metric distance between the b-th initial signal segment and the c-th initial signal segment is larger. And when the clustering metric distance is larger, the probability of being clustered into one category later is smaller. When D(b, c) is smaller, it indicates that the clustering metric distance between the b-th initial signal segment and the c-th initial signal segment is smaller. And when the clustering metric distance is smaller, the probability of being clustered into one category later is larger.

[0034] In this embodiment, the specific process of clustering all the initial signal segments in the initial signal segment set according to the clustering metric distance between any two initial signal segments in the initial signal segment set is as follows: According to the clustering metric distance between any two initial signal segments in the initial signal segment set, use the OPTICS (Ordering Points To Identify the Clustering Structure) clustering algorithm to cluster all the initial signal segments in the initial signal segment set to obtain clustering clusters; and the OPTICS clustering algorithm is a density-based clustering algorithm, which has more advantages than the classical DBSCAN clustering algorithm when dealing with clusters of complex shapes, noise points, and clusters of different densities. The clustering process of the OPTICS clustering algorithm is a well-known technology, so it will not be described in detail here.

[0035] Therefore, through the above process, this embodiment can obtain the target signal segment of each vital sign data signal.

[0036] Step S003: Obtain the decomposition level corresponding to the target signal segment according to the frequency-domain signal corresponding to the target signal segment, and perform wavelet transform on the corresponding target signal segment according to the decomposition level corresponding to the target signal segment to obtain all component signals corresponding to the target signal segment.

[0037] After segmentation, the component signals corresponding to the target signal segment are determined based on the frequency richness of each target signal segment. That is, in this embodiment, next, the decomposition level corresponding to the target signal segment of each vital sign data signal will be obtained according to the frequency-domain signal corresponding to the target signal segment of each vital sign data signal, and wavelet transform will be performed on the corresponding target signal segment according to the decomposition level corresponding to the target signal segment to obtain all component signals corresponding to the target signal segment of each vital sign data signal; it can be seen that before obtaining all the component signals corresponding to the target signal segment, it is necessary to first obtain the decomposition level corresponding to the target signal segment. Then, the specific process of obtaining the decomposition level corresponding to the target signal segment is as follows:

[0038] For any target signal segment P: First, perform Fourier transform on the target signal segment P to obtain the frequency-domain signal corresponding to the target signal segment P. Then, according to the frequencies and the corresponding amplitudes on the frequency-domain signal corresponding to the target signal segment P, obtain the decomposition level-related factor corresponding to the target signal segment P; then, record the total number of vital sign data points on the target signal segment P as the feature quantity value, and record the integer value of the logarithm of the feature quantity value with a preset second constant as the base as the maximum decomposable level of the target signal segment P. Finally, obtain the integer value of the product of the decomposition level-related factor of the target signal segment P and the maximum decomposable level of the target signal segment P, and record it as the decomposition level corresponding to the target signal segment P; that is, the decomposition level corresponding to the target signal segment P is: where KP is the decomposition level corresponding to the target signal segment P, is the floor symbol, β is the decomposition level related factor corresponding to the target signal segment P, log 2 (M0) is the logarithm of M0 to the base 2, M0 is the characteristic quantity value, that is, the preset second constant in this embodiment is 2; and when M0 is larger and β is larger, it indicates that K P has a larger value, and K P has a larger value, which more indicates that when performing wavelet transform on the target signal segment P subsequently, more decomposition levels are required for decomposition.

[0039] In this embodiment, the specific process of obtaining the decomposition level related factor corresponding to the target signal segment P according to the frequency and the corresponding amplitude on the frequency domain signal corresponding to the target signal segment P is as follows: First, obtain the set constructed by all frequencies with non-zero amplitudes on the frequency domain signal corresponding to the target signal segment P, and denote it as the frequency set. Then, obtain the maximum amplitude and the corresponding frequency on the frequency domain signal corresponding to the target signal segment P, and denote them as the characteristic amplitude and the characteristic frequency respectively. Immediately afterwards, obtain the frequency difference and the amplitude difference corresponding to each frequency in the frequency set, and then obtain the result of multiplying the frequency difference corresponding to each frequency in the frequency set by the reciprocal of the amplitude difference corresponding to the corresponding frequency, and use it as the characteristic difference value corresponding to the corresponding frequency. Then, obtain the normalized value of the result obtained by accumulating the characteristic difference values of all frequencies in the frequency set, and use it as the decomposition level related factor corresponding to the target signal segment P. And the frequency difference corresponding to any frequency in the frequency set is the absolute value of the difference between the absolute value of this frequency and the absolute value of the characteristic frequency, and the amplitude difference corresponding to any frequency in the frequency set is the result of subtracting the amplitude corresponding to this frequency from the characteristic amplitude and then adding the preset first constant. In addition, the specific calculation expression of the decomposition level related factor corresponding to the target signal segment P is:

[0040]

[0041] where, β is the decomposition level related factor corresponding to the target signal segment P, V is the number of frequencies in the frequency set, c1 is the preset first constant, F0 is the characteristic amplitude, F v is the amplitude corresponding to the v-th frequency in the frequency set, |H0| is the absolute value of the characteristic frequency, |H v | is the absolute value of the v-th frequency in the frequency set. And when (F0 - F v ) is smaller and ||H0| - |H v||The larger it is, it indicates that there may be multiple main frequency components in the target signal segment P, or in other words, the frequency richness of the target signal segment P is higher. At this time, more decomposition levels are required to decompose it to improve the accuracy and reliability of the decomposition result. From this, it can be known that when the value of β is larger, it indicates that the frequency richness of the target signal segment P is higher, and then when performing decomposition, the target signal segment P requires more decomposition levels.

[0042] Therefore, through the above process, this embodiment can obtain the decomposition level corresponding to the target signal segment of each vital sign data signal, and then perform wavelet transform on the corresponding target signal segment according to the decomposition level corresponding to the target signal segment of each vital sign data signal, and record the result obtained by performing the transform on each target signal segment as the component signal corresponding to the corresponding target signal segment. The number of component signals corresponding to the target signal segment is the same as the decomposition level corresponding to the corresponding target signal segment. That is, when performing wavelet transform on the target signal segment, one component signal corresponding to the corresponding target signal segment can be obtained for each decomposition level corresponding to the target signal segment; and on the premise of knowing the decomposition level corresponding to a certain target signal segment, the process of performing wavelet transform on the target signal segment to obtain the component signal corresponding to the target signal segment is a well-known technology, so this embodiment will not be described in detail.

[0043] Therefore, through the above process, this embodiment can obtain all the component signals corresponding to the target signal segment of each vital sign data signal.

[0044] Step S004, perform inverse transform on the component signals corresponding to the target signal segments of all the vital sign data signals in the vital sign data signal set to obtain the target fusion signal.

[0045] In this embodiment, the specific process of performing inverse transform on the component signals corresponding to the target signal segments of all the vital sign data signals in the vital sign data signal set to obtain the target fusion signal is as follows: First, obtain the high-frequency component signal corresponding to the target signal segment and the low-frequency component signal corresponding to the target signal segment; then, in the vital sign data signal set, obtain all types of segmentation moments that appear, and record the sequence constructed by all types of segmentation moments that appear as the segmentation moment sequence; then, according to the segmentation moment sequence, obtain the signal segments to be fused on each vital sign data signal in the vital sign data signal set and the marking values of the signal segments to be fused, and according to the target signal segment to which the signal segment to be fused on the vital sign data signal belongs, the low-frequency component signal corresponding to the target signal segment, and the marking value of the signal segment to be fused, obtain the low-frequency fused vital sign data signal, and according to the target signal segment to which the signal segment to be fused on the vital sign data signal belongs, the high-frequency component signal corresponding to the target signal segment, and the marking value of the signal segment to be fused, obtain the high-frequency fused vital sign data signal, and then record both the high-frequency fused vital sign data signal and the low-frequency fused vital sign data signal as the target fusion signal.

[0046] In this embodiment, the specific process of obtaining the high-frequency component signal corresponding to the target signal segment and the low-frequency component signal corresponding to the target signal segment is as follows: For any target signal segment P: First, denote the set constructed by all component signals corresponding to the target signal segment P as the component signal set corresponding to the target signal segment P, and obtain the DTW distance between any two component signals in the component signal set. The DTW distance (Dynamic Time Warping distance) is an effective method for measuring the similarity between two unequal-length time series; Then, according to the DTW distance between any two component signals in the component signal set corresponding to the target signal segment P, use the K-means clustering algorithm to divide the component signals in the component signal set corresponding to the target signal segment P into two clustering clusters, denoted as the first clustering cluster and the second clustering cluster respectively. And the process of clustering using the K-means clustering algorithm is also a well-known technology, so it will not be described in detail here; After that, determine whether the decomposition layer of the component signals in the first clustering cluster is lower than that of the component signals in the second clustering cluster. If so, denote all the component signals in the first clustering cluster as the high-frequency component signal corresponding to the target signal segment P, and denote all the component signals in the second clustering cluster as the low-frequency component signal corresponding to the target signal segment P. Otherwise, denote all the component signals in the first clustering cluster as the low-frequency component signal corresponding to the target signal segment P, and denote all the component signals in the second clustering cluster as the high-frequency component signal corresponding to the target signal segment P. In addition, if there are two component signals in the first clustering cluster, and the two component signals in the first clustering cluster are obtained through the first decomposition layer and the second decomposition layer respectively, then at this time, the decomposition layers of the component signals in the first clustering cluster are the first decomposition layer and the second decomposition layer. If there are also two component signals in the second clustering cluster, and the two component signals in the second clustering cluster are obtained through the third decomposition layer and the fourth decomposition layer respectively, then at this time, the decomposition layers of the component signals in the second clustering cluster are the third decomposition layer and the fourth decomposition layer. Also, since the i-th decomposition layer is lower than the (i + 1)-th decomposition layer, it is determined at this time that the decomposition layer of the component signals in the first clustering cluster is lower than that of the component signals in the second clustering cluster.

[0047] In this embodiment, an example of obtaining all types of segmentation moments that appear in the set of physical sign data signals is as follows: If there are two physical sign data signals in the set of physical sign data signals, and there are two segmentation moments on the first physical sign data signal, which are 9:12 AM and 9:18 AM respectively, and there are two segmentation moments on the second physical sign data signal, which are 9:10 AM and 9:15 AM respectively, then all types of segmentation moments that appear at this time are 9:10 AM, 9:12 AM, 9:15 AM, and 9:18 AM.

[0048] In this embodiment, the specific process of obtaining the signal segments to be fused and the marking values of the signal segments to be fused on each physiological sign data signal in the physiological sign data signal set according to the segmentation time sequence is as follows: For any physiological sign data signal: First, on this physiological sign data signal, all positions with abscissas belonging to the segmentation time sequence are marked as the segmentation positions on this physiological sign data signal, and the starting position and the ending position on this physiological sign data signal are also marked as segmentation positions; Then, the local signal segment between the (r - 1)-th segmentation position and the r-th segmentation position on this physiological sign data signal is marked as the (r - 1)-th signal segment to be fused on this physiological sign data signal, where r > 1, and the marking value of the (r - 1)-th signal segment to be fused on this physiological sign data signal is marked as r - 1; In addition, except that the first signal segment to be fused on the physiological sign data signal is closed, other signal segments to be fused on the physiological sign data signal are left-open and right-closed.

[0049] In this embodiment, the specific process of obtaining the low-frequency fused physiological sign data signal according to the target signal segment to which the signal segment to be fused on the physiological sign data signal belongs, the low-frequency component signal corresponding to the target signal segment, and the marking value of the signal segment to be fused is as follows: First, obtain all the low-frequency component signal segments to be processed corresponding to each signal segment to be fused on each physiological sign data signal; Then, all the signal segments to be fused with the same marking value are divided into a set, which is denoted as the set of signal segments to be fused with each marking value; For any marking value, the set of all the low-frequency component signal segments to be processed corresponding to all the signal segments to be fused in the set of signal segments to be fused with this marking value is denoted as the set of low-frequency component signal segments to be processed corresponding to this marking value, and the inverse wavelet transform is performed on the set of low-frequency component signal segments to be processed corresponding to this marking value, and the result of the inverse transform is denoted as the low-frequency fused sub-signal corresponding to this marking value, that is, performing the inverse wavelet transform on the set of low-frequency component signal segments to be processed corresponding to this marking value means recombining all the low-frequency component signal segments to be processed in the set of low-frequency component signal segments to be processed corresponding to this marking value, and the new signal obtained by the combination is the low-frequency fused sub-signal corresponding to this marking value; Finally, the low-frequency fused sub-signals corresponding to all the marking values are spliced in ascending order of the marking values, and the new signal obtained after the splicing is denoted as the low-frequency fused physiological sign data signal.

[0050] In this embodiment, the specific process of obtaining all the low-frequency component signal segments corresponding to the signal segment to be fused is as follows: For any signal segment L to be fused: First, record the time interval corresponding to the signal segment L to be fused as the target time interval, and obtain the target signal segment to which the signal segment L to be fused belongs. Then, on all the low-frequency component signals corresponding to the target signal segment to which the signal segment L to be fused belongs, extract all the low-frequency component signal segments whose time intervals belong to the target time interval, and all of them are used as the low-frequency component signal segments to be processed corresponding to the signal segment L to be fused. And if the signal segment L to be fused is a part of the target signal segment L0 or if the signal segment L to be fused is divided from the target signal segment L0, then the target signal segment L0 is the target signal segment to which the signal segment L to be fused belongs; for example, if the target time interval is from 9:12 am to 9:18 am, then on any low-frequency component signal corresponding to the target signal segment to which the signal segment L to be fused belongs, the signal segment corresponding to 9:12 am to 9:18 am extracted is a low-frequency component signal segment to be processed corresponding to the signal segment L to be fused, that is, one low-frequency component signal segment to be processed corresponding to the signal segment L to be fused can be extracted from each low-frequency component signal corresponding to the target signal segment to which the signal segment L to be fused belongs.

[0051] In this embodiment, the method for obtaining the high-frequency fusion physical sign data signal is the same as the method for obtaining the low-frequency fusion physical sign data signal, that is, when obtaining the high-frequency fusion physical sign data signal, only need to replace the low-frequency with high-frequency in the process of obtaining the low-frequency fusion physical sign data signal. Therefore, the method for obtaining the high-frequency fusion physical sign data signal will not be described in detail.

[0052] In addition, as another implementation manner, when performing the inverse transform, different weights can also be assigned to the component signals of different physical sign data types according to the features that the implementer wants to highlight during the fusion. For example, if the implementer wants to highlight the information of the electrocardiogram data, then when performing the inverse transform, the weights of the component signals corresponding to the electrocardiogram data signal need to be larger.

[0053] So far, this embodiment has completed the fusion of the physical sign data signal set.

[0054] In summary, in this embodiment, the set of physical sign data signals of the patient is first obtained; then, according to the extreme points in each physical sign data signal in the set of physical sign data signals and the time intervals between adjacent extreme points, the data stability characterization value of the physical sign data signal is obtained. According to the data stability characterization value, the total number of physical sign data points and the total number of extreme points on the physical sign data signal, the physical sign data signal is segmented to obtain the target signal segment of the physical sign data signal; then, according to the frequency domain signal corresponding to the target signal segment, the decomposition level corresponding to the target signal segment is obtained, and wavelet transform is performed on the corresponding target signal segment according to the decomposition level corresponding to the target signal segment to obtain all component signals corresponding to the target signal segment; finally, inverse transform is performed on the component signals corresponding to the target signal segments of all the physical sign data signals in the set of physical sign data signals to obtain the target fusion signal. And in this embodiment, by segmenting the physical sign data signal and then adaptively obtaining the decomposition level corresponding to the target signal segment based on the frequency domain signal corresponding to each target signal segment, the accuracy and reliability of subsequent signal decomposition can be improved, thereby improving the reference value and reliability of the fusion result.

[0055] The above-described embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A patient multi-source monitoring data fusion analysis system based on time series analysis, comprising a processor and a memory, characterized in that: The processor executes the computer program stored in the memory to implement the following steps: Acquire a patient's vital sign data signal set; According to the time intervals between the extreme points and adjacent extreme points in each vital sign data signal in the vital sign data signal set, a data stability characterization value of the vital sign data signal is obtained; and according to the data stability characterization value and the total number of vital sign data points and the total number of extreme points on the vital sign data signal, the vital sign data signal is segmented to obtain a target signal segment of the vital sign data signal; According to the frequency domain signal corresponding to the target signal segment, the number of decomposition layers corresponding to the target signal segment is obtained, and the corresponding target signal segment is subjected to wavelet transformation according to the number of decomposition layers corresponding to the target signal segment to obtain all component signals corresponding to the target signal segment; The component signals corresponding to the target signal segments of all the vital sign data signals in the vital sign data signal set are inversely transformed to obtain a target fusion signal.

2. The patient multi-source monitoring data fusion analysis system based on time series analysis as claimed in claim 1, characterized in that: The method for obtaining the data stability characterization value of the vital sign data signal comprises: For any vital sign data signal: obtain the maximum point sequence and the minimum point sequence on the vital sign data signal, take the standard deviation of the ordinates of all the maximum points in the maximum point sequence and the maximum standard deviation of the standard deviation of the ordinates of all the minimum points in the minimum point sequence as the first characterization value, obtain the minimum time interval sequence and the maximum time interval sequence of the vital sign data signal, and take the standard deviation of the minimum time interval sequence and the maximum standard deviation of the standard deviation of the maximum time interval sequence as the second characterization value, the fth minimum time interval on the minimum time interval sequence The interval is the absolute value of the horizontal coordinate difference between the f-th minimum point and the f+1-th minimum point on the vital sign data signal, and the g-th maximum time interval on the maximum time interval sequence is the absolute value of the horizontal coordinate difference between the g-th maximum point and the g+1-th maximum point on the vital sign data signal; the vital sign data signal is composed of vital sign data points, the horizontal coordinate of the vital sign data point is the acquisition time of the vital sign data, and the vertical coordinate is the vital sign data; the result of negative correlation mapping the sum of the first characterization value and the second characterization value is used as the data stability characterization value of the vital sign data signal.

3. The patient multi-source monitoring data fusion analysis system based on time series analysis as claimed in claim 2, characterized in that: The method for acquiring the target signal segment of the vital sign data signal comprises: For any vital sign data signal: determine whether the data stability characterization value of the vital sign data signal is greater than a preset stability threshold value; if so, use the vital sign data signal directly as the target signal segment of the vital sign data signal; otherwise, obtain the initial signal segment of the vital sign data signal according to the data stability characterization value of the vital sign data signal and the total number of vital sign data points and the total number of extreme value points on the vital sign data signal, and record the set constructed by all the initial signal segments on the vital sign data signal as the initial signal segment set; obtain the target signal segment of the vital sign data signal according to the clustering metric distance between any two initial signal segments in the initial signal segment set; The method for obtaining the initial signal segment on the vital sign data signal is as follows: recording the maximum number of the number of minimum value points and the number of maximum value points on the vital sign data signal as the initial extreme value number value, recording the sum of the initial extreme value number value and the preset first constant as the target extreme value number value, recording the product of the ratio of the total number of vital sign data points on the vital sign data signal to the target extreme value number value and the data stability characterization value of the vital sign data signal as the initial change frequency characterization value; according to the value range and mapping function of the initial change frequency characterization value, performing positive correlation mapping on the change frequency characterization value, and taking the integer value obtained by multiplying the mapping result by the acquisition frequency of the vital sign data points on the vital sign data signal as the target sliding window length corresponding to the vital sign data sequence, taking a window with a length equal to the target sliding window length and a width of 1 as the target sliding window of the vital sign data signal, and using the target sliding window to perform overlapping division on the vital sign data signal, and recording the division results as the initial signal segment on the vital sign data signal.

4. The patient multi-source monitoring data fusion analysis system based on time series analysis as claimed in claim 3, characterized in that: The method for obtaining the target signal segment of the vital sign data signal according to the clustering metric distance between any two initial signal segments in the initial signal segment set comprises: Obtain the clustering metric distance between any two initial signal segments in the initial signal segment set, and cluster all the initial signal segments in the initial signal segment set according to the clustering metric distance to obtain cluster clusters; determine whether the first initial signal segment and the second initial signal segment on the vital sign data signal belong to the same cluster cluster; if so, continue to determine whether the second initial signal segment and the third initial signal segment on the vital sign data signal belong to the same cluster cluster; if not, take the middle position of the time interval corresponding to the overlapping signal segment between the second initial signal segment and the third initial signal segment as the first segmentation moment on the vital sign data signal, and record the signal segment corresponding to the starting moment to the first segmentation moment on the vital sign data signal as the first target signal segment of the vital sign data signal, and so on, continue to traverse the initial signal segments on the vital sign data signal until the last initial signal segment on the vital sign data signal is traversed and stop, and count all target signal segments of the vital sign data signal.

5. The patient multi-source monitoring data fusion analysis system based on time series analysis as claimed in claim 4, characterized in that: The method for obtaining the clustering metric distance includes: For the bth initial signal segment and the cth initial signal segment in the initial signal segment set, b is not equal to c: the normalized value of the square of the difference between the fluctuation amplitude of the bth initial signal segment and the fluctuation amplitude of the cth initial signal segment is recorded as the first difference, the normalized value of the square of the difference in the coefficient of variation between the bth initial signal segment and the cth initial signal segment is recorded as the second difference, the normalized value of the square of the difference between the number of extreme points on the bth initial signal segment and the number of extreme points on the cth initial signal segment is recorded as the third difference, and the square root of the result obtained by adding the first difference, the second difference and the third difference is used as the clustering metric distance between the bth initial signal segment and the cth initial signal segment.

6. The patient multi-source monitoring data fusion analysis system based on time series analysis according to claim 1, characterized in that: The method for obtaining the number of decomposition layers corresponding to the target signal segment includes: For any target signal segment, the decomposition layer number correlation factor corresponding to the target signal segment is obtained according to the frequency on the frequency domain signal corresponding to the target signal segment and the amplitude corresponding to the frequency; the total number of vital sign data points on the target signal segment is recorded as the characteristic quantity value, the integer value of the logarithm of the characteristic quantity value with a preset second constant as the base is recorded as the maximum decomposable number of layers of the target signal segment, and the integer value of the decomposition layer number correlation factor corresponding to the target signal segment multiplied by the maximum decomposable number of layers is recorded as the decomposition layer number corresponding to the target signal segment.

7. The patient multi-source monitoring data fusion analysis system based on time series analysis according to claim 6, characterized in that: The method for obtaining the decomposition layer number correlation factor corresponding to the target signal segment includes: The set constructed by all frequencies whose amplitudes on the frequency domain signal corresponding to the target signal segment are not 0 is recorded as a frequency set, and the maximum amplitude and the frequency corresponding to the maximum amplitude on the frequency domain signal corresponding to the target signal segment are recorded as characteristic amplitude and characteristic frequency, respectively; the frequency difference and amplitude difference corresponding to each frequency in the frequency set are obtained, and the result of multiplying the frequency difference corresponding to the frequency by the inverse of the amplitude difference corresponding to the corresponding frequency is used as the characteristic difference value of the corresponding frequency, and the normalized value of the result obtained by accumulating the characteristic difference values ​​of all frequencies in the frequency set is used as the decomposition layer number correlation factor corresponding to the target signal segment, the frequency difference corresponding to any frequency in the frequency set is the absolute value of the difference between the absolute value of the frequency and the absolute value of the characteristic frequency, and the amplitude difference corresponding to any frequency in the frequency set is the result of subtracting the amplitude corresponding to the frequency from the characteristic amplitude and adding a preset first constant.

8. The patient multi-source monitoring data fusion analysis system based on time series analysis as claimed in claim 4, characterized in that: The method of performing inverse transformation on component signals corresponding to target signal segments of all vital sign data signals in the vital sign data signal set to obtain a target fusion signal comprises: For any target signal segment: all component signals corresponding to the target signal segment are recorded as a component signal set; according to the DTW distance between any two component signals in the component signal set, the component signals in the component signal set are divided into two clusters, which are respectively recorded as the first cluster and the second cluster; it is determined whether the decomposition layer of the component signals in the first cluster is lower than the decomposition layer of the component signals in the second cluster; if so, all component signals in the first cluster are recorded as high-frequency component signals corresponding to the target signal segment, and all component signals in the second cluster are recorded as low-frequency component signals corresponding to the target signal segment; otherwise, all component signals in the first cluster are recorded as low-frequency component signals corresponding to the target signal segment, and all component signals in the second cluster are recorded as high-frequency component signals corresponding to the target signal segment; The sequence constructed by all types of segmentation moments appearing in the vital sign data signal set is recorded as a segmentation moment sequence; according to the segmentation moment sequence, the signal segments to be fused and the label values ​​of the signal segments to be fused on each vital sign data signal in the vital sign data signal set are obtained, and according to the target signal segment to which the signal segment to be fused belongs, the low-frequency component signal corresponding to the target signal segment and the label value of the signal segment to be fused, a low-frequency fused vital sign data signal is obtained; according to the target signal segment to which the signal segment to be fused belongs, the high-frequency component signal corresponding to the target signal segment and the label value of the signal segment to be fused, a high-frequency fused vital sign data signal is obtained; the method for obtaining the high-frequency fused vital sign data signal is the same as the method for obtaining the low-frequency fused vital sign data signal, and both the high-frequency fused vital sign data signal and the low-frequency fused vital sign data signal belong to the target fusion signal.

9. The patient multi-source monitoring data fusion analysis system based on time series analysis as claimed in claim 8, characterized in that: The method for acquiring the signal segments to be fused and the label values ​​of the signal segments to be fused on each vital sign data signal in the vital sign data signal set includes: For any vital sign data signal, on the vital sign data signal, all positions whose horizontal coordinates belong to the segmentation time sequence are recorded as segmentation positions, the starting position and the ending position on the vital sign data signal are also recorded as segmentation positions, and the local signal segment from the r-1th segmentation position to the rth segmentation position on the vital sign data signal is recorded as the r-1th signal segment to be fused on the vital sign data signal, r is greater than 1, and the marking value of the r-1th signal segment to be fused on the vital sign data signal is r-1.

10. The patient multi-source monitoring data fusion analysis system based on time series analysis according to claim 8, characterized in that: The method for acquiring the low-frequency fusion vital sign data signal comprises: For any signal segment to be fused, the time interval corresponding to the target signal segment to be fused is recorded as the target time interval, and all low-frequency component signal segments whose time intervals belong to the target time interval are extracted from all low-frequency component signal segments corresponding to the target signal segment to which the target signal segment to be fused belongs, and all of them are used as the low-frequency component signal segments to be processed corresponding to the signal segment to be fused; The set constructed by all the low-frequency component signal segments to be processed corresponding to all the signal segments to be fused with the same tag value is recorded as the set of low-frequency component signal segments to be processed corresponding to the corresponding tag value, the inverse transform of the wavelet transform is performed on the set of low-frequency component signal segments to be processed corresponding to the tag value, and the inverse transform result is recorded as the low-frequency fusion sub-signal corresponding to the corresponding tag value, the low-frequency fusion sub-signals corresponding to all the tag values ​​are spliced ​​in the order of the tag value from small to large, and the splicing result is recorded as the low-frequency fusion vital sign data signal.

Images