A wearable underwater vital signs and environmental monitoring system

Abstract

The present application relates to the technical field of vital sign measurement, in particular to a wearable underwater vital sign and environment monitoring system, which comprises a physiological signal acquisition module, a photoelectric plethysmogram sensor is fixed at a wrist position, a bioelectrical impedance electrode is attached to a chest, and a piezoelectric sensor is placed at a neck position, and the photoelectric plethysmogram signal of the wrist part, the bioelectrical impedance signal of the chest and the piezoelectric signal of the neck are acquired.In the present application, signal reconstruction is carried out based on energy proportion calculation, the position of the R wave peak of electrocardiogram and the R-R interval time sequence are extracted, the variability index is generated through frequency domain decomposition and frequency band energy statistics, the time domain and frequency domain index values are calculated and weight distribution is carried out, so as to realize comprehensive monitoring of underwater vital sign state.With the help of multi-source signal acquisition mode, the signal acquisition accuracy is improved, the noise interference is eliminated through signal trajectory matrix decomposition operation, and the signal reconstruction quality is ensured through feature vector energy proportion calculation.

Description

Technical Field

[0001] The present invention relates to the technical field of vital sign measurement, and in particular to a wearable underwater vital sign and environment monitoring system. Background Art

[0002] Vital sign measurement technology is a key branch of medical health monitoring, focusing on the acquisition, analysis, and monitoring of human vital signs. This field includes the measurement and assessment of key physiological indicators such as body temperature, heart rate, blood pressure, respiratory rate, and blood oxygen saturation. Wearable underwater vital sign monitoring systems are portable devices specifically designed to monitor human vital signs in underwater environments.

[0003] However, existing technologies have many limitations in practical applications. Relying solely on a single sensor to collect physiological signals makes it difficult to ensure data integrity and accuracy. Signal distortion and information loss are easily caused by the underwater environment, leading to large deviations in monitoring results. Furthermore, the use of a simple threshold judgment method for signal feature extraction makes it difficult to cope with signal distortion and severe fluctuations, resulting in inaccurate feature parameter extraction. In the process of assessing vital sign status, only focusing on changes in a single indicator and ignoring the joint analysis of multi-dimensional physiological parameters makes the assessment results one-sided and unreliable. Therefore, improvement is needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings existing in the prior art, and to propose a wearable underwater vital signs and environment monitoring system. In order to achieve the above purpose, the present invention adopts the following technical solutions: A wearable underwater vital signs and environment monitoring system includes: a physiological signal acquisition module, which fixes the photoelectric volume pulse wave sensor at the wrist position, attaches the bioelectrical impedance electrode to the chest, and places the piezoelectric sensor at the neck position, collects the photoelectric volume pulse wave signal at the wrist, the bioelectrical impedance signal at the chest and the piezoelectric signal at the neck, obtains multi-source physiological signals, performs time alignment and synchronous sampling on the multi-source physiological signals, and generates a physiological signal sequence; a signal denoising and reconstruction module, which constructs a signal trajectory matrix based on the physiological signal sequence, performs block decomposition operation on the signal trajectory matrix, extracts signal eigenvalues ​​and eigenvectors, obtains a set of characteristic components, calculates the energy proportion of the eigenvectors in the set of characteristic components, and generates a reconstructed signal group; A dynamic evaluation module, based on the reconstructed signal group, extracts the ECG R wave peak point position and the ECG RR interval time series, calculates the difference and fluctuation range of adjacent RR intervals, obtains the heart rate change characteristics, performs frequency domain decomposition and frequency band energy statistics on the heart rate change characteristics, and generates a heart rate variability index; a vital sign parameter fusion module, based on the heart rate variability index, respectively calculates the heart rate variability time domain index and frequency domain index values, counts the change trends and fluctuation ranges of each index, obtains vital sign fluctuation parameters, performs weight distribution on the vital sign fluctuation parameters, and generates underwater vital sign state quantities; an environmental monitoring module, for monitoring the underwater environmental temperature through a temperature sensor unit, monitoring the underwater environmental pressure through a pressure sensor unit, and monitoring the underwater environmental salinity through an inductive conductivity sensor unit, and obtaining underwater environmental monitoring data; a communication module, for communicating the underwater vital sign state quantities and underwater environmental monitoring data with a wearable display terminal and a surface monitoring device via an underwater cable or wireless communication. Preferably, the steps for acquiring the physiological signal sequence are: based on the multi-source physiological signals, perform time alignment, align each signal according to the timestamp, and adjust the signal data by interpolation or delay according to the acquisition time point of each signal to obtain a time-aligned multi-source physiological signal; based on the time-aligned multi-source physiological signals, perform data synthesis, integrate all aligned signals into a unified time point through data fusion, and generate a physiological signal sequence. Preferably, the steps for acquiring the characteristic component set are: based on the physiological signal sequence, generate a two-dimensional matrix according to the arrangement of the signal values ​​sampled at each time point, use the data at each time point as the row vector and the signal channel data as the column vector to obtain a signal trajectory matrix; based on the signal trajectory matrix, calculate the characteristic score of the signal decomposition, and the calculation formula is: in, is the feature score, are the singular values ​​of the signal trajectory matrix, is the corresponding left singular vector, is the number of blocks; according to the feature score, extract the matching components of the eigenvalues ​​and eigenvectors, aggregate the block feature scores to generate corresponding signal eigenvalues ​​and eigenvector sets, and obtain a feature component set. Preferably, the steps for obtaining the reconstructed signal group are: according to the feature component set, extract the eigenvectors and corresponding singular values, rearrange the eigenvectors and singular values ​​in matrix form to generate an eigenvector matrix; based on the eigenvector matrix, calculate the energy proportion, and the calculation formula is: in, is the energy density, To reconstruct the signal Component values, is the first feature vector in the set Component values, is the number of principal components selected, is the total number of eigenvectors; the eigenvectors with energy proportions higher than the threshold are weightedly combined with the corresponding singular values, and inverse singular value decomposition is performed to generate a reconstructed signal group. Preferably, the steps for obtaining the heart rate variation characteristics are: according to the reconstructed signal group, each ECG signal channel is processed, the peak point position of the R wave is extracted through waveform analysis and threshold determination, and the peak point positions are arranged in a time series to obtain an ECG R wave peak point position sequence; based on the ECG R wave peak point position sequence, the nonlinear fluctuation index is calculated, and the calculation formula is: in, is the nonlinear fluctuation index of adjacent RR intervals, For the The time position of the R wave peak point, and are the time positions of the previous and the previous two R wave peak points, is a dimensionless constant that is adjusted in the positive direction to avoid the denominator being zero. is the total number of R waves; based on the nonlinear fluctuation index, combined with the time series analysis of the electrocardiogram signal, the heart rate variation characteristics are generated by calculating the amplitude distribution characteristics of the R wave. Preferably, the steps for obtaining the heart rate variability index are: based on the heart rate variation characteristics, the heart rate variation characteristics are converted from the time domain to the frequency domain through Fourier transform, and different frequency components and amplitude values ​​are extracted to obtain the frequency domain heart rate characteristics; based on the frequency domain heart rate characteristics, the frequency band energy proportion of the target frequency band is calculated, and the calculation formula is: in, is the energy ratio of the frequency band, Frequency The power spectral density at and are the lower and upper frequency limits of the target frequency band, and is the minimum and maximum frequency of the entire spectrum range; combined with the main frequency component information in the frequency domain, the frequency band energy proportions of multiple frequency bands are counted and summarized into an index set to generate a heart rate variability index. Preferably, the steps for obtaining the vital sign fluctuation parameters are: classifying the heart rate variability index into two parts: time domain features and frequency domain features, performing statistics on the time domain features including the standard deviation and the root mean square interval, and classifying the frequency domain features including the frequency band energy ratio and the main frequency to obtain the classified heart rate variability index; based on the classified heart rate variability index, calculating the change trend and fluctuation range of each time domain feature and frequency domain feature, and statistically analyzing the extreme values, ranges and average changes between the indicators to obtain the statistical indicator change trend; based on the statistical indicator change trend, calculating the vital sign fluctuation parameter, the calculation formula is: in, is the physical sign fluctuation parameter, For the The change value of the time domain feature, For the The change value of the frequency domain feature of the item, For the The proportion of frequency domain features, is the total number of features. Preferably, the steps for obtaining the underwater vital sign state quantity are as follows: according to the vital sign fluctuation parameters, they are divided into time domain features and frequency domain features according to the feature type, each change value in the time domain features and each energy value in the frequency domain features are extracted respectively, and the data are normalized according to a unified range, and the normalized data are recombined to form a normalized vital sign fluctuation parameter set; based on the normalized vital sign fluctuation parameter set, the similarity between each feature is calculated through the correlation between the features, and the correlation relationship between the features is expressed in the form of a matrix, and the features are merged and recalculated to obtain a fused vital sign fluctuation parameter set; according to the fused vital sign fluctuation parameter set, weight values ​​are assigned according to the classification standards of the time domain features and the frequency domain features, and the features after the weight assignment are summed up to generate the underwater vital sign state quantity. Compared with the existing technology, the advantages and positive effects of the present invention are as follows: In the present invention, multi-source physiological signal acquisition is performed by fixing a photoplethysmography sensor on the wrist, attaching a bioelectrical impedance electrode to the chest, and placing a piezoelectric sensor on the neck. The signals are time-series aligned and synchronously sampled. A signal trajectory matrix is ​​constructed and block decomposition is performed. Eigenvalues ​​and eigenvectors are extracted. Signal reconstruction is performed based on energy weight calculation. The ECG R wave peak point position and RR interval time series are extracted. Variability indicators are generated through frequency domain decomposition and frequency band energy statistics. Time domain and frequency domain indicator values ​​are calculated and weighted, thereby achieving comprehensive monitoring of underwater vital sign status. The multi-source signal acquisition method improves signal acquisition accuracy, eliminates noise interference through signal trajectory matrix decomposition, and uses eigenvector energy weight calculation to ensure signal reconstruction quality. Based on ECG R wave feature extraction and frequency domain decomposition, accurate assessment of vital sign change trends is achieved. Combined with a weight allocation strategy, the underwater vital sign status monitoring results are objective and reliable. The invention has excellent performance in signal acquisition, noise reduction processing, feature extraction, and status assessment, providing reliable protection for vital sign monitoring of underwater workers. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0006] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Figure 1The present invention provides a technical solution: a wearable underwater vital sign and environment monitoring system includes: a physiological signal acquisition module, which fixes a photoelectric volumetric pulse wave sensor at the wrist, attaches a bioelectrical impedance electrode to the chest, and places a piezoelectric sensor at the neck, collects the photoelectric volumetric pulse wave signal at the wrist, the bioelectrical impedance signal at the chest, and the piezoelectric signal at the neck to obtain multi-source physiological signals, performs time alignment and synchronous sampling on the multi-source physiological signals, and generates a physiological signal sequence; a signal denoising and reconstruction module, which constructs a signal trajectory matrix based on the physiological signal sequence, performs block decomposition operation on the signal trajectory matrix, extracts signal eigenvalues ​​and eigenvectors, obtains a set of characteristic components, calculates the energy proportion of the eigenvectors in the set of characteristic components, and generates a reconstructed signal group; The dynamic evaluation module extracts the ECG R wave peak point position and the ECG RR interval time series based on the reconstructed signal group, calculates the difference and fluctuation range of adjacent RR intervals, obtains the heart rate change characteristics, performs frequency domain decomposition and frequency band energy statistics on the heart rate change characteristics, and generates a heart rate variability index; the vital sign parameter fusion module calculates the heart rate variability time domain index and frequency domain index values ​​based on the heart rate variability index, statistics the change trend and fluctuation range of each index, obtains the vital sign fluctuation parameter, weights the vital sign fluctuation parameter, and generates the underwater vital sign state quantity; the environmental monitoring module is used to monitor the underwater environmental temperature through the temperature sensor unit, monitor the underwater environmental pressure through the pressure sensor unit, and monitor the underwater environmental salinity through the inductive conductivity sensor unit to obtain underwater environmental monitoring data; the communication module is used to communicate the underwater vital sign state quantity and underwater environmental monitoring data with the wearable display terminal and surface monitoring equipment through underwater cables or wireless communications. The steps for acquiring a physiological signal sequence are as follows: Based on multi-source physiological signals, perform timing alignment, align each signal according to the timestamp, and adjust the signal data by interpolation or delay according to the acquisition time point of each signal to obtain a timing-aligned multi-source physiological signal; based on the timing-aligned multi-source physiological signals, perform data synthesis, integrate all aligned signals into a unified time point through data fusion, and generate a physiological signal sequence.Specifically, based on the multi-source physiological signals obtained previously, first read the timestamps marked by each signal source during the acquisition phase one by one, and convert the timestamps of different signals into millisecond level for subsequent comparison. Then, check the amplitude range of each signal source, such as the photoelectric volume pulse wave signal compared with the 0.0mV to 5.0mV range, the chest bioelectrical impedance signal compared with the 0.0Ω to 150.0Ω range, and the piezoelectric sensor signal compared with the 0.0Pa to 1000.0Pa range. The upper and lower limits of these values ​​are set empirically based on the specifications of the sensors used. If the amplitude of a record exceeds the corresponding range, it is marked as an abnormal record and is not included in the interpolation processing for the time being. Then, align the records in chronological order. If a gap with a time difference of more than 10ms is found between adjacent sampling points, fill the missing position with linear interpolation method, assuming that the adjacent known sampling points are. and , the corresponding times are and The time point that needs to be interpolated is , then the interpolated value according to Calculate, where and represents the measurement value of the adjacent known points, Between and between, if the amplitude after interpolation still falls within the aforementioned abnormal range, it is recorded as to-be-confirmed data for subsequent correction, after completing all interpolations, all time points are rearranged in ascending order and a unified time sequence is formed, and during this period, it is also checked whether the time interval of any two records is less than 1 ms, if less than 1 ms, it is determined as a repeated sampling point and mean value merging is performed, finally all records with matched time stamp and amplitude are summarized, and finally the time-aligned multi-source physiological signals are obtained. According to the multi-source physiological signals aligned as described above, first, the effective data index of each signal under the unified time reference is determined, if it is found that the photoplethysmogram signal is missing at some time and the bioimpedance signal and piezoelectric sensor signal are effective, the effective signals are retained and a marker entry is generated at the missing place to prevent omission, then all aligned signal records are sequentially combined according to time points, for multiple amplitude values appearing at the same time, average value fusion can be used for synthesis, and at the same time, whether there is a significant deviation is judged according to the amplitude difference threshold value set by experience, for example, if the amplitude difference between the photoplethysmogram and the piezoelectric sensor at the same time point exceeds 10.0 units, the threshold value is obtained by statistical observation data, then a difference entry is marked during fusion, next, all merged entries are verified again, and whether there is a new abnormality is judged by comparing the interval ranges of 0.0 to 5.0 mV, 0.0 Ω to 150.0 Ω and 0.0 Pa to 1000.0 Pa, if there is no new abnormality, the synthesis operation continues, if a new abnormality is generated, its time is recorded and the values of adjacent time points are compared briefly, if the amplitude difference after comparison is still greater than the experience threshold value, it is temporarily retained as an abnormality for subsequent correction, after completing the fusion of all time points, a time sequence can be arranged, which contains the data values of the three signals at the same time point, and finally the physiological signal sequence is generated. wherein, is the feature score, is the singular value of the signal trajectory matrix, is the corresponding left singular vector, is the number of blocks; according to the feature scores, the matching components of the eigenvalues ​​and eigenvectors are extracted, and the block feature scores are aggregated to generate the corresponding signal eigenvalue and eigenvector sets to obtain the feature component set. Specifically, based on the physiological signal sequence obtained previously, the values ​​at each sampling moment are arranged in sequence and it is ensured that each time point matches the corresponding signal amplitude data. The photoelectric volumetric pulse wave signal can be detected using 0mV to 5mV as the effective range, the chest bioelectrical impedance signal can be compared using the interval of 0Ω to 150Ω, and the neck piezoelectric signal can be checked using the interval of 0Pa to 1000Pa. These interval data are obtained by continuously collecting and recording the actual sensor data for multiple times and then sorting them out. Then, all signal amplitude values ​​are horizontally combined according to the index order of the time points, and the three signal amplitude values ​​corresponding to each timestamp are sorted into a row. In this process, the timing synchronization data needs to be re-calibrated. If individual timestamps exceed the tolerance deviation of 1ms, additional linear interpolation is performed, and then combined with the amplitude comparison detection of adjacent data, for example, when the amplitude difference between the photoelectric volumetric pulse wave and the neck piezoelectric signal at the same moment exceeds 10, the threshold of the difference is obtained by cumulative observation statistics, and it will be marked as a potential abnormality and enter the subsequent If an extreme mutation is found in subsequent analysis, the moment is recorded and preparation is made for subsequent data correction. At the same time, the bioelectrical impedance signal in the same row is checked for excessive differences in amplitude between the previous and next timestamps. If it exceeds the 20Ω threshold obtained from the previous acquisition statistics three times in a row, the row is marked as an abnormal row for further verification. Then, all sampling points are organized in the form of row vectors. The amplitude values ​​of the three channels in each row are arranged from left to right. The leftmost column is the photoplethysmography data, the middle column is the bioelectrical impedance data, and the rightmost column is the piezoelectric sensor data. After completion, the rows are numbered to correspond to the time sequence. Then, the column vectors are divided vertically according to the channel category. The number of column vectors formed is equal to the number of signal channels and the upper and lower positions are consistent. Each column in the file, from top to bottom, is the amplitude value of the same signal at different time points. Since the abnormality has been marked by threshold detection, it is only necessary to index the values ​​of each row and column. Finally, these rows and columns are stored in a matrix to obtain the signal trajectory matrix. The benefit of this formula is that it can highlight the proportion of different blocks in the signal space by multiplying and accumulating the block singular values ​​and the corresponding left singular vector amplitudes, and quantify the contribution of each block to the overall signal change into a feature score, so that in the subsequent steps, this value can be compared with the threshold to extract the corresponding effective features; The acquisition steps are as follows: first, perform singular value decomposition on the signal trajectory matrix and record each element in the singular value vector. These elements come from the decomposition results of the matrix after multiple block divisions, and are obtained by calculating the data amplitude and time index in each block. The specific block division method is to split the original matrix by row or column, keeping the sub-matrices after division non-overlapping, and then calculate the singular value vector for each sub-matrix separately, and then merge the singular values ​​of all sub-matrices according to the index position to obtain ,in The value of will be different due to different block sizes. For example, after multiple measurements and statistics, if the block size is set to 3 blocks, the record is obtained. =2.3, =1.8, =0.9; The steps to obtain it are: first decompose the left singular vectors of each submatrix Perform amplitude calculation and obtain the value by accumulating the square of the data amplitude, for example, Considered as a vector of length m, each element of which can be formed by reading the block matrix row by row and normalizing it, the corresponding , obtained in the same way and ,in The specific value of is determined by the number of rows in the submatrix after partitioning. Assuming that the value is obtained through the previous statistics =200, and calculated through measured data to obtain ; Calculation process: Step 1, list the known parameters: The second step is to multiply and add the items according to the formula: The third step is to calculate the product values ​​and add them up in sequence: The results show that the feature score of the whole block is 5.463, when the feature score of the block is higher than a certain range, it means that the fluctuation corresponding to the signal trajectory matrix may be larger, if the value falls between 2 to 4, it means that the contribution of the block in this range is relatively balanced, if it exceeds 5, it means that some blocks may have higher amplitude concentration, so in the follow-up, other indicators can be used for more detailed feature extraction. According to the feature score obtained in the foregoing, first, the corresponding index information of the singular value and the left singular vector in each block is read, and the index range corresponding to the block position in the singular value sequence recorded in the early stage is found, for example, if the number of blocks is 3, the data of the position of 1 to 3 segments in the index segmentation point of the singular value sequence is searched, and the feature score calculated in the foregoing is cross-checked to confirm whether the value of each block is within the block amplitude interval set in advance, if block 1 has a block amplitude exceeding 3.0 in the corresponding index range, it is marked as a larger value, the threshold of 3.0 is obtained by statistically summarizing a plurality of matrices collected continuously, all high-amplitude or low-amplitude index rows are found in all blocks and are compared with the amplitude square of the left singular vector one by one, if the amplitude square continuously exceeds 1.0, the position is recorded and the data difference between the front and rear rows is compared, if the continuous three rows are greater than 1.0, they are included in the candidate sequence of feature components and continuously observed in the subsequent index, the block information recorded above is extracted together with the singular value information and combined to form a matching list of feature values and feature vectors, in this process, it is necessary to ensure that each pair of feature values and left singular vectors has a unique index mark to avoid mismatch, then all the feature values and feature vectors of the blocks are combined in order and form a unified matching item set based on the same index mechanism in the foregoing, the indexes of the abnormal or high-amplitude blocks marked are reserved for later checking, and the aggregation of the feature scores of the blocks is further completed, if it is found that the singular value corresponding to a certain index segment is obviously higher than other index segments, for example, more than 2 times of the average value, the index segment is recorded and the corresponding relationship with the original sequence is maintained, and finally it is used in the subsequent steps, after completing all the blocks, the corresponding signal feature value and feature vector set are obtained, and after summarizing, the final feature component set is formed. The acquisition step of the reconstructed signal group is: according to the feature component set, the feature vector and the corresponding singular value are extracted, the feature vector and the singular value are rearranged in the form of a matrix to generate a feature vector matrix; based on the feature vector matrix, the energy proportion is calculated, and the calculation formula is: wherein, is the energy proportion, is the first component value of the reconstructed signal, is the first component value in the feature vector set, is the number of selected principal components, is the total number of eigenvectors; perform a weighted combination of the eigenvectors with energy proportions higher than the threshold and the corresponding singular values, perform inverse singular value decomposition, and generate a reconstructed signal group. Specifically, based on the previously obtained set of eigencomponents, read the corresponding vectors and singular values ​​of each eigencomponent and list them one by one, and check the arrangement relationship of these vectors in the index order one by one. If it is detected that the source of continuous eigencomponents is inconsistent with the index of the previous block result, it is necessary to re-examine the construction method of those eigenvectors to avoid dislocation, and then rearrange all corresponding eigenvectors and singular values ​​in column or row order. During the arrangement process, first record the dimension of each vector, for example, record the length of one vector as 100 and the length of the other vector as 120. These data of different lengths need to be supplemented with blank entries or truncated with redundant entries in the corresponding order of the eigenvalues. For example, if it is found that there is If the terminal index of some vectors exceeds the required range, the excess part will be removed and its specific position will be recorded. It is also necessary to align the corresponding indexes of the singular values. If a singular value does not match the index marked in front of the eigenvector, it will be rechecked before the final merge. Then, a matrix with corresponding rows and columns is constructed for all the checked eigenvectors and singular values. The eigenvector content is placed in the column direction first, and the row direction corresponds to the serial number of each singular value. At this time, the exact number of columns and rows must be determined to avoid empty columns or rows. After all the alignments are completed, these elements will be uniformly indexed and form the final eigenvector matrix. The final matrix will be used in the subsequent energy proportion calculation. The benefit of the formula is that by selecting the number of principal components The ratio operation of the sum of squares of the eigenvector components in the image to the sum of squares of all eigenvector components can directly determine the energy contribution of the currently selected principal component to the overall eigenvector set, and use this ratio to guide the subsequent reconstruction or filtering process; and The acquisition step is to read each vector element in the eigenvector matrix. The values ​​of these elements come from the amplitude data obtained after multiple signal acquisitions and decompositions. Each vector element can be regarded as an amplitude component within a sampling range. For example, in some test scenarios, the pulse wave amplitude recorded falls between 0mV and 5mV, the electrical impedance amplitude is between 0Ω and 150Ω, and the piezoelectric amplitude is between 0Pa and 1000Pa. The amplitudes of different channels are merged into the eigenvector to form the component value. After statistics, there will be n total components. The steps for obtaining are to first perform variance contribution analysis on all feature components, observe the components in descending order of variance contribution, and then select the top m components with the highest proportion for subsequent principal component processing. The specific value of m can be determined after comprehensively comparing the energy proportion curve and the recognition accuracy requirements. For example, if it is found that selecting the top three components in several groups of test data can achieve a more comprehensive coverage, m is set to 3. It can also be added or reduced later to match the new data distribution. The acquisition step is to count the total number of components in the eigenvector matrix. Its value is determined by the scale of the singular value decomposition and vector merging in the previous step. For example, in an experiment, multiple source signals were collected. Each signal generated several eigenvectors in the block processing, which were aggregated into 8 vectors. Each vector contains several components. After summing up all the components, n may reach hundreds or thousands. This value needs to be confirmed with the actual data collection volume. Calculation process: The first step is to list the values ​​of each parameter in the test case, such as the number of principal components selected. , the total number of eigenvector components , first record all component values ​​as Step 2: Calculate the denominator : The third step is to take the first three components as the principal components, the numerator Calculated as: Step 4: Substitute into the formula: The result shows that the three principal components currently selected account for about 70.9% of the overall component energy. When this ratio exceeds 80%, it may mean that the number of principal components selected is more sufficient to cover the overall energy. When the ratio is too low, the number of principal components should be increased or decreased according to the actual situation to meet the subsequent reconstruction or separation needs. Based on the energy weight value calculated previously, all eigenvectors are first matched with the corresponding singular values ​​in the order of their indexes. Then, the index of each vector with an energy weight greater than the threshold in the record is read and its position in the eigenvector matrix is ​​marked. The threshold value needs to be set in combination with the energy weight distribution obtained previously. For example, after multiple tests, it is found that when the energy weight is higher than 0.8, there is a more stable reconstruction quality. The threshold can be set to 0.8. Then, the eigenvectors corresponding to these indexes are numerically multiplied with the singular values ​​to complete the weighting operation. If the eigenvector is repeatedly included in the index traversal, the number of times is counted and the corresponding coefficient is accumulated in the subsequent weighting. The size of the coefficient can refer to the previous block result. The amplitude range of the result, for example, the stability of the photoplethysmography signal can be checked between 0mV and 5mV, and the amplitude fluctuation of the chest bioelectrical impedance signal can be monitored between 0Ω and 150Ω. These fluctuation values ​​are normalized and combined with the singular values ​​to obtain a more refined weighted result. After sorting, the inverse singular value decomposition can be performed on all the filtered weighted feature vectors. In this process, the length and position of each vector need to be kept consistent. They are arranged one by one and matrix operations are performed with the weighted singular values. After the operation, a series of time series amplitude sequences are generated. These sequences are summarized to obtain a reconstructed signal group. Finally, this reconstructed signal group can be subjected to subsequent analysis and feature extraction in subsequent steps. The steps for obtaining heart rate variation characteristics are as follows: According to the reconstructed signal group, each ECG signal channel is processed, and the peak point position of the R wave is extracted through waveform analysis and threshold judgment. The peak point positions are arranged in a time series to obtain an ECG R wave peak point position sequence; Based on the ECG R wave peak point position sequence, the nonlinear fluctuation index is calculated. The calculation formula is: in, is the nonlinear fluctuation index of adjacent RR intervals, For the The time position of the R wave peak point, and are the time positions of the previous and the previous two R wave peak points, is a dimensionless constant that is adjusted in the positive direction to avoid the denominator being zero. is the total number of R waves; according to the nonlinear wave index, combined with time series analysis of the electrocardio signal, the heart rate variation characteristics are generated by calculating the amplitude distribution characteristics of the R wave. Specifically, according to the reconstructed signal group obtained in the foregoing, first, the amplitude values at different sampling times are read for each electrocardio signal channel row by row and are indexed and compared, and if it is observed that the amplitude is continuously higher than the experience threshold, it is marked, for example, the R wave amplitude detection threshold is set to 0.5 mV, which is obtained by recording and grouping statistics of the electrocardio data collected in the actual population sample for many days, then all the amplitude records are arranged in chronological order and the adjacent amplitude difference is monitored, and if there is a sudden increase in amplitude at some time, the waveform within about 20 ms before and after the time is further scanned to determine whether there is an R wave peak with a prominent shape. In the scanning, a shape standard for judgment can be referred to, for example, the main feature of the R wave in the QRS complex is the upward sharp waveform and the peak position has a high amplitude. Each potential peak point that has been marked is further checked again in combination with the slope change rate of the waveform before and after it, and if the slope change of the peak point in the neighborhood before and after it exceeds the set slope threshold, the peak point is confirmed as the R wave peak. The slope threshold can be determined by extracting the characteristic value distribution in the daily monitoring data of multiple different subjects, for example, the slope threshold range is between 0.3 mV / ms and 0.6 mV / ms obtained through multiple inductions, and if the observed value exceeds the range, it is preliminarily determined that the peak value has sufficient distinguishability. The time positions of all peak points that pass the threshold and slope verification are recorded one by one, and the time positions are aligned to the millisecond level by software. If a small amount of time indexes of the records have a small interval, the amplitude difference before and after it can be investigated, and if it is found that the interval between adjacent peaks is less than 200 ms, the amplitude and slope are compared to select a peak value that better meets the R wave shape and eliminate the other one to avoid repetition. After the peak value identification of all channels is completed, the peak points are rearranged in chronological order to form a unified peak index sequence, and finally the electrocardio R wave peak position sequence is obtained. The advantage of the formula is that by differentiating and nonlinearly correcting the time positions of adjacent R wave peak points, the rhythm fluctuation amplitude in different time periods can be evaluated and the potential abnormal change degree can be prompted, so that the fluctuation degree is better measured when the heart rate regularity or rhythm is monitored subsequently; The acquisition step of the formula is to sequentially extract the peak time positions indexed as k from the electrocardio R wave peak position sequence obtained in the foregoing. The time positions are obtained by a millisecond timer synchronized with the electrocardio data, for example, in 24-hour sampling, the occurrence time of each R wave is recorded as a time in tens of thousands of data points; And The acquisition step of the formula is to sequentially extract the peak time positions indexed as k from the electrocardio R wave peak position sequence obtained in the foregoing. The time positions are obtained by a millisecond timer synchronized with the electrocardio data, for example, in 24-hour sampling, the occurrence time of each R wave is recorded as a time in tens of thousands of data points; The steps for obtaining are: when processing multi-day ECG data, if it is found that the denominator may have a minimum value that causes a division by zero problem, then a positive adjustment value is selected. For example, after analyzing thousands of R wave peak records, if the value is set It can balance extreme cases and avoid the denominator reaching zero; The acquisition step is to count the total number of R wave peaks marked in the batch of ECG data. For example, after obtaining 80,000 data points through 24-hour monitoring, about 90,000 sampling peaks are identified, of which there may be 85,000 valid R wave peaks. In this case, xm=85,000. Calculation process: The first step is to select the ECG R wave peak time series under a certain test scenario. For example, take out 5 moments in order. , and order , xm=5; Step 2, calculate the numerator : Step 3: Calculate the denominator : For simplicity, we can first roughly calculate: ln(801) , ln(1601) , ln(2401) , ln(3151) , ln(4051) , and 1 / 800 , 1 / 1600 , 1 / 2400 , 1 / 3150 , 1 / 4050 , adding these together gives about 38.210, and then adding Multiplying them, we get 0.038210; the denominator is ; Step 4, substitute into the formula to get: This result indicates that this R-wave peak sequence exhibits a certain degree of nonlinear fluctuation within this range. When the H value exceeds 0.6, it may indicate increased rhythm fluctuations, while when the H value is below 0.3, it indicates relatively stable fluctuations within this range. In this case, H is approximately 0.4396, indicating a moderate degree of fluctuation. Combined with the analysis of the R-wave amplitude distribution characteristics in the subsequent steps, the characteristics of heart rate variation can be further evaluated. Based on the nonlinear fluctuation index obtained previously, the actual amplitude of each peak is first read from the recorded R-wave peak list and summarized into a separate distribution sequence. When checking the amplitude range of this distribution sequence, the amplitude is compared with the range of 0.0mV to 5.0mV. Any single peak amplitude exceeding 4.5mV three times in a row is marked. This 4.5mV value is selected based on the peak conditions of more than 100 people during daily monitoring. If the peak amplitudes in the records are simultaneously below 0.1mV, it is confirmed that these peaks may be due to noise or device contact problems and are individually marked. Then, all R-wave amplitudes are sorted in order of size and their distribution frequency is counted. If it is possible to If there are high-density peak concentration sections within the visible range, it is necessary to focus on frequency comparison of these sections in subsequent operations to provide a reference for analyzing the actual amplitude changes. At the same time, the amplitude change rate of adjacent peaks should be tracked in time series. For example, it is detected whether the amplitude difference of three to five consecutive peaks exceeds the previously established 0.8mV threshold. After recording the samples that exceed this threshold, their time index is recorded and paired with the nonlinear fluctuation index obtained previously to compare whether there is a large H value fluctuation at the same time. If both are in the high value range, this period of time can be classified as a high fluctuation section. After the above recording work is completed, the heart rate change characteristics can be integrated and output according to the distribution, and finally the heart rate change characteristics are generated. The steps for obtaining the heart rate variability index are as follows: according to the heart rate change characteristics, the heart rate change characteristics are converted from the time domain to the frequency domain through Fourier transform, and different frequency components and amplitude values ​​are extracted to obtain the frequency domain heart rate characteristics; based on the frequency domain heart rate characteristics, the frequency band energy proportion of the target frequency band is calculated. The calculation formula is: in, is the energy ratio of the frequency band, Frequency The power spectral density at and are the lower and upper frequency limits of the target frequency band, and are the minimum and maximum frequencies of the entire spectrum range; combined with the main frequency component information in the frequency domain, the frequency band energy proportions of multiple frequency bands are counted and summarized into an indicator set to generate the heart rate variability index. Specifically, according to the heart rate change characteristics obtained above, the timing information of each heart rate change record is first read and recombined in an equal interval form. After the heart rate amplitudes of adjacent records are aligned at the millisecond level, discrete Fourier transforms are performed one by one to obtain the amplitude distribution corresponding to different frequency points. In the specific operation, each heart rate record can be marked according to the time index. If the sampling time of a record exceeds the preset 5ms interval compared with the previous one, the 5ms value is determined by observing the heart rate sampling distribution of multiple subjects. Then, linear interpolation is performed on the position and the amplitude after interpolation is recorded. Subsequently, the entire time series is uniformly processed during discrete Fourier transform. All sampled amplitude values ​​are arranged in order and used as transformation input. When the transformation is completed, frequency components covering the range of 0Hz to tens of Hz can be obtained. The extremely low frequency area below 0.05Hz and the ultra-high frequency area above 40Hz are eliminated or separately marked. These frequency intervals are determined based on the common analysis intervals in the study of human heart rate characteristics. For example, After multiple verifications of the heart rate data obtained from continuous one-day (24-hour) monitoring, it was found that the main information is mostly concentrated in the range of 0.05Hz to 2.0Hz. Most of the components outside this range are related to environmental interference or more extreme situations. After eliminating or marking these segments, the amplitude values ​​of the remaining parts are recorded and matched one by one with the frequency index. If the frequency points whose amplitude values ​​continuously exceed the established threshold are detected, for example, exceeding 1.2 times the maximum amplitude mean obtained in the previous statistics, these points are marked as noteworthy frequencies. Finally, all the search results are summarized to form a frequency domain data containing amplitude and corresponding frequency information. According to the multiple frequency components and amplitude sizes detected, they are classified. For example, the 0.05Hz to 0.15Hz segment is recorded as the ultra-low frequency segment, 0.15Hz to 0.4Hz is recorded as the low frequency segment, 0.4Hz to 1Hz is recorded as the medium frequency segment, and above 1Hz can be regarded as the high frequency segment. These segments are divided according to the conventional research range of human heart rate variability. After the processing is completed, the frequency domain heart rate characteristics are obtained. The formula is beneficial in that by The power spectral density within the spectrum is consistent with the overall spectrum interval The power spectral density ratio is calculated to quantify the energy proportion of the target frequency band and distinguish the relative contribution of each band; The acquisition step is to square the amplitude spectrum obtained after Fourier transform of the heart rate time series signal to obtain the power spectrum density, where each frequency point corresponds to an amplitude value. The amplitude value can be obtained by the sampling device in multiple measurements, and the power can be obtained by squaring the amplitude. and The acquisition steps are to select the target frequency band by analyzing the heart rate variation pattern. For example, when studying the low frequency band of heart rate variability, Hz, Hz, the specific value is often obtained based on clinical or experimental observation statistical methods; and The acquisition step is to count all the frequency ranges obtained after Fourier transform and take the minimum and maximum values ​​as boundaries. For example, if the sampling frequency is 100Hz and the data length is fixed, the spectrum range may be 0Hz to 50Hz. In crowd heart rate analysis, the interval of 0Hz to 2Hz is often concerned, so we can let Hz, Hz; Calculation process: The first step, in actual measurement, assume that the heart rate changes of a subject are monitored and the power spectrum covering the range of 0Hz to 2Hz is obtained, and the target frequency band is selected at the same time is [0.04, 0.15], and the The value is integrated; the second step is to integrate the The same integration operation is performed on the value (here 0Hz to 2Hz) to obtain the denominator; in the third step, the energy proportion is obtained by dividing the two integral results. ; In the given example, let (Estimated by the actual integral value of the power spectrum of a large number of heart rate samples), the integral result of the entire interval is ; Step 4, substitute into the formula: The results show that the current target frequency band (0.04Hz to 0.15Hz) accounts for 30% of the overall energy. If the energy proportion of a frequency band is higher than 50%, it may indicate that the activity in this frequency band is more active. If it is lower than 10%, it means that the frequency band contributes less to the overall activity. Based on this value, the significance of the distribution of heart rate variability in different frequency bands can be further judged. Combined with the main frequency component information previously obtained in the frequency domain, the main frequency positions are first selected from the amplitude spectrum data obtained by Fourier transform and the corresponding frequency and amplitude values ​​are recorded. The surrounding frequency segments of each main frequency position need to be checked. If the amplitude value is detected to exceed 1.2 times the average amplitude obtained previously, the segment is determined to be a high amplitude interval and marked. At the same time, the energy proportion is calculated for each segment within the overall range of 0 Hz to 2 Hz. For example, several adjacent intervals can be set, where 0.04 Hz to 0.15 Hz is the low frequency segment, 0.15 Hz to 0.4 Hz is the medium frequency segment, 0.4 Hz to 1 Hz is the high frequency segment, and 1 Hz to 2 Hz is the higher frequency segment. The energy proportion is then calculated for each interval. If the proportion in a certain interval exceeds the previously recorded historical average twice in a row, it is again marked as an active segment. After completing the statistics for all frequency bands, these proportion values ​​are collected and combined with the corresponding main frequency components to form an indicator set for this segment data. Finally, the multi-band energy proportion and main frequency position data contained in this indicator set are sorted and output as a heart rate variability index. The steps for obtaining the vital sign fluctuation parameters are as follows: classify the heart rate variability index into two parts: time domain characteristics and frequency domain characteristics. Statistically analyze the time domain characteristics, including standard deviation and root mean square interval, and classify the frequency domain characteristics, including frequency band energy ratio and main frequency, to obtain the classified heart rate variability index; based on the classified heart rate variability index, calculate the change trend and fluctuation range of each time domain characteristic and frequency domain characteristic, and calculate the extreme value, range and average change between the statistical indicators to obtain the statistical indicator change trend; based on the statistical indicator change trend, calculate the vital sign fluctuation parameters. The calculation formula is: in, is the physical sign fluctuation parameter, For the The change value of the time domain feature, For the The change value of the frequency domain feature of the item, For the The proportion of frequency domain features, is the total number of features. Specifically, the heart rate variability index is classified into two parts: time domain features and frequency domain features. First, the time domain features are confirmed to include the statistical methods required for the two indicators, standard deviation and root mean square interval. The continuously recorded heart rate values ​​or RR intervals are extracted from the recorded data set and arranged in chronological order. The time domain features are read one by one and the standard deviation is calculated. When calculating, the difference between all heart rate values ​​and their average values ​​can be squared and then averaged. Finally, the square root of the result can be obtained to obtain the standard deviation. The root mean square interval can also be obtained by performing a secondary operation on the difference between adjacent heart rate values ​​or RR intervals, averaging, and then taking the square root. If any record lacks a valid value, it is marked as missing data and summarized in subsequent comparisons. If it is found that some time domain features have extremely large or extremely small deviations, additional checks are performed and the number of threshold violations is recorded. These thresholds are determined by the statistical results of continuous monitoring of the same population for several weeks. For example, a standard deviation greater than 70ms or a root mean square interval greater than 100ms is recorded as a significant deviation. Then, the frequency domain information in the heart rate data is compared to obtain frequency domain features. For frequency domain features, several key intervals can be selected from the power spectrum data obtained above, such as the low frequency band of 0.04Hz to 0.15Hz and the high frequency band of 0.15Hz to 0.4Hz, and the power ratios in these intervals are counted. Then, the amplitude information of the sampling points is read for the main frequency position. If the power ratio observed in a specific interval exceeds 1.3 times the historical mean, it is recorded as a frequency band with large fluctuations. The low-frequency power ratio and high-frequency power ratio obtained above are summarized together with the main frequency value as frequency domain features. When summarizing the results, it is necessary to ensure that each feature item has a corresponding time index for subsequent comparison. If individual intervals lack valid frequency components, the interval is marked as a null value and is not included in the mean calculation. Through this process, features can be collected in both the time domain and the frequency domain. The obtained standard deviation, root mean square interval, frequency band energy ratio and main frequency are recorded and classified in turn to form a set of classified heart rate variability indicators.Based on the classified heart rate variability indicators, first read the standard deviation and root mean square interval in the time domain features one by one and record the maximum and minimum values in multiple sampling stages, if the standard deviation exceeds the set 150ms upper limit in a certain period, it is marked as the maximum value record, this 150ms is determined by grouping a large amount of data for multiple times, at the same time, if the root mean square interval is lower than 10ms in the continuous monitoring hours, it is marked as the minimum value interval, make incremental statistics of these extreme values in the software and combine the time range of each sampling period to obtain the feature change curve, then the same operation is performed on the frequency domain features, the maximum or minimum value of each frequency band energy ratio at different time is extracted, if the energy ratio reaches or exceeds 50% in any period and exceeds twice the average value obtained by continuous observation, it is recorded as a high energy paragraph, and the time index is marked, if the main frequency position appears continuous rise or fall in several detections and the amplitude exceeds 0.1Hz, it also needs to be recorded separately, wherein the threshold of 0.1Hz is obtained according to the regular fluctuation range found in multiple human body monitoring, then the above indicators are compared horizontally, the peak and valley range and average change value experienced by them in the sampling period are counted, if it is found that the root mean square interval of the time domain features and the high frequency band energy of the frequency domain features both appear large fluctuations, the change amplitude and time period are recorded and the extreme value is selected, finally, according to the maximum value, minimum value, range and average value fluctuation of all time domain features and frequency domain features, the overall index change trend is drawn, and the statistical index change trend is obtained after the arrangement is completed. The advantage of the formula is that the change value of the time domain feature and the change value of the frequency domain feature are considered at the same time. , the change value of the frequency domain feature , and the proportion of the frequency domain feature , the linkage between each feature is comprehensively measured and the frequency domain change is modified by the logarithmic function, which can more finely reflect the status of different types of features in the overall fluctuation; The acquisition step is to select the difference value of the first time domain feature (such as standard deviation or root mean square interval) in several monitoring periods, the specific value can be obtained by subtracting the measurement results of different periods and counting, for example, in a certain observation, the first time domain feature is 30ms in the early monitoring stage and 50ms in the later monitoring stage, then ms; The acquisition step is to extract the difference before and after from multiple measurements for the corresponding frequency domain feature (such as low frequency band energy ratio or high frequency band energy ratio), if the low frequency band energy ratio increases from 20% to 32% in a certain period, then , it is necessary to ensure that the data comes from the same feature item and the difference of the monitoring results of the same sequence; The acquisition step is to read the proportion of the frequency domain feature, for example, the overall energy proportion of a certain frequency band is obtained from the previous frequency spectrum integration result, if the frequency band proportion is 25%, then , this value can be regarded as the relative weight of the feature in the frequency domain; The acquisition step is to count all the features that have been included in the analysis and get the total number. For example, if it contains two time domain features (standard deviation, root mean square interval) and two frequency domain features (low frequency band energy ratio, high frequency band energy ratio), then ; Calculation process: The first step is to list the parameters obtained from actual observations: Let , assuming that this is the difference and proportion obtained for the two time domain features of standard deviation and root mean square interval and the two frequency domain features of low frequency band and high frequency band; the second step is to calculate The specific value of hour: right hour: The third step is to square the two and add them together and then take the square root: The results show that the above two time domain features and two frequency domain features are combined, the fluctuation parameter of the sign is about 6.151, if the value is higher than 10, it means that the change of multiple features is large and the weight of frequency domain is high, if it is lower than 2, it means that the overall change amplitude is limited and most of the feature difference is small, in the actual scene, the threshold can be further formulated to determine whether the sign enters the high fluctuation interval or remains stable. The acquisition steps of the underwater sign state quantity are: according to the sign fluctuation parameter, the features are divided into time domain features and frequency domain features according to the feature type, the change value of each item in the time domain feature and the energy value of each item in the frequency domain feature are extracted, the data is normalized according to the unified range, the normalized data is recombined to form a normalized sign fluctuation parameter set; based on the normalized sign fluctuation parameter set, the similarity between each feature is calculated through the correlation between the features, and the correlation between the features is represented in the form of matrix, the features are merged and recalculated to obtain the fused sign fluctuation parameter set; according to the fused sign fluctuation parameter set, the weight value is distributed according to the classification standard of time domain features and frequency domain features, and the features after weight distribution are summed and calculated to generate the underwater sign state quantity. Specifically, according to the sign fluctuation parameter obtained in the foregoing, first, the numerical range of each time domain feature and frequency domain feature is counted and the upper and lower limits of these ranges are recorded, these upper and lower limits can be obtained by grouping and screening the continuous multi-day sampling results and taking the minimum and maximum values, then the features are divided into time domain features and frequency domain features according to the feature type, each item change value in the time domain feature and each item energy value in the frequency domain feature are read separately and normalized according to the unified reference interval, for example, the data in the 0 to minimum value region is uniformly mapped to 0, and the part from the maximum value to the higher value is linearly extended or removed if it exists, while ensuring that extreme over-limit data is specially marked in the normalization process, for example, when a time domain feature appears to be higher than the previously counted 200ms upper limit, the value is uniformly set to 1 in the normalization process and marked as an over-limit point, the 200ms upper limit is obtained by quantile analysis based on continuous monitoring of hundreds of hours of heart rate sampling records, after normalization, the final data of each feature item is indexed and checked to avoid corresponding disorder, if some features are missing in a sampling stage, the corresponding empty record is reserved in the normalization list and marked with the time index, finally, the processed time domain features and frequency domain features are merged into a new sequence in the same time order, if the number of time domain features and frequency domain features is inconsistent during merging, interpolation or zero padding is performed according to the previously recorded time index to ensure that the final time domain and frequency domain features can correspond one by one, after this step, the normalized data is recombined to form a normalized sign fluctuation parameter set.Based on the normalized set of vital sign fluctuation parameters, we first compare the numerical similarity between each feature item one by one and select a suitable measurement method to calculate the similarity. For example, for the change value of the time domain feature, the ratio of the absolute value of the numerical difference can be used for comparison. For the energy value of the frequency domain feature, the difference and its proportion can be used for comprehensive measurement. If the difference of the same feature in multiple adjacent time periods is lower than the 0.05 threshold set by the statistics, it is considered to have a high similarity. The 0.05 is obtained by summarizing the variance distribution of multiple measurements after normalization and selecting a standard with a deviation of about 5% from the mean. Then, each pair of features is compared and the obtained similarity is recorded in the corresponding position in the matrix. If the similarity of some feature pairs is high in multiple samplings, If the average is lower than 0.1, they can be regarded as relatively independent. The 0.1 is also set with reference to the average difference of the previous multiple groups of heart rate samples. Then, feature merging is carried out according to the similarity matrix, and features with high similarity are classified into the same category or their results are smoothly combined. In this process, if a feature contains multiple numerical components and the distribution span is large, this feature needs to be split into multiple sub-items and then compared separately. After the comparison is completed, the merged result is recalculated. For example, multiple time domain features with high similarity are uniformly weighted or the weighted average is taken. The same steps are performed on the frequency domain features and the weighting coefficient is corrected with the previous proportion information. Finally, these re-merged and calculated results are recorded one by one and assembled into a fused set of vital sign fluctuation parameters. According to the fused set of vital sign fluctuation parameters, first read the category corresponding to each feature in turn according to the classification standards of time domain features and frequency domain features and assign corresponding weight values ​​to them. The weight assignment can be implemented in combination with the importance scores of time domain and frequency domain indicators in historical monitoring data. For example, the weight of time domain features can be set between 0.4 and 0.6 and slightly increased or decreased according to the previous degree of deviation from the standard deviation and root mean square interval. The weight of frequency domain features can be set in the range of 0.3 to 0.5 and an increasing or decreasing factor is used for the actual proportion of low frequency bands and high frequency bands. The specific values ​​of these weights are obtained by statistically comparing multiple measurements and integrating expert experience. It is determined by the test method. If it is found in multiple batches of data that the low-frequency energy has a greater impact on the overall heart rate state change, the weight of the low-frequency band can be increased accordingly. Finally, each eigenvalue after the weight is assigned is summed up at the same time index. When summing up, the combined results of the time domain and frequency domain are read one by one and multiplied by their corresponding weights one by one, and then all the products are summed up to obtain a representative value. If any eigenvalue is missing at a certain moment, the moment will be marked as blank during weighting and the summation will be skipped or its weight will be assigned to other features. After the calculation of all moments is completed, the summation results in the time series can be further recorded, and the underwater vital sign state quantity is generated.

Claims

1. A wearable underwater vital signs and environment monitoring system, characterized in that: The system comprises: A physiological signal acquisition module fixes a photoplethysmography sensor on the wrist, attaches a bioelectrical impedance electrode to the chest, and places a piezoelectric sensor on the neck. The module collects photoplethysmography signals from the wrist, bioelectrical impedance signals from the chest, and piezoelectric signals from the neck to obtain multi-source physiological signals. The module then performs time alignment and synchronous sampling on the multi-source physiological signals to generate a physiological signal sequence. A signal denoising and reconstruction module constructs a signal trajectory matrix based on the physiological signal sequence, performs a block decomposition operation on the signal trajectory matrix, extracts signal eigenvalues ​​and eigenvectors to obtain a set of characteristic components, calculates the energy proportion of the eigenvectors in the set of characteristic components, and generates a reconstructed signal group; A dynamic evaluation module extracts the ECG R wave peak point position and ECG RR interval time series based on the reconstructed signal group, calculates the difference and fluctuation range of adjacent RR intervals, obtains heart rate variation characteristics, performs frequency domain decomposition and frequency band energy statistics on the heart rate variation characteristics, and generates a heart rate variability index; The vital sign parameter fusion module calculates the heart rate variability time domain index and frequency domain index values ​​based on the heart rate variability index, calculates the change trend and fluctuation range of each index, obtains the vital sign fluctuation parameter, and performs weight distribution on the vital sign fluctuation parameter to generate the underwater vital sign state quantity; Environmental monitoring module, used to monitor water temperature through temperature sensors; a communication module, used for communicating with the outside world via an underwater cable or wireless communication; The steps for acquiring the heart rate variation characteristics are: According to the reconstructed signal group, each ECG signal channel is processed, the peak point position of the R wave is extracted by waveform analysis and threshold determination, and the peak point positions are arranged in a time series to obtain an ECG R wave peak point position sequence; Based on the ECG R wave peak point position sequence, the nonlinear fluctuation index is calculated using the following formula: Among them, H is the nonlinear fluctuation index of adjacent RR intervals, t k is the time position of the kth R wave peak point, t k-1 and t k-2 are the time positions of the previous and the previous two R-wave peaks, ∈ is a dimensionless constant adjusted in the positive direction to avoid the denominator being zero, and xm is the total number of R waves; Based on the nonlinear fluctuation index and combined with the time series analysis of the ECG signal, the heart rate variation characteristics are generated by calculating the amplitude distribution characteristics of the R wave; The steps for obtaining the heart rate variability index are: According to the heart rate variation characteristics, the heart rate variation characteristics are converted from the time domain to the frequency domain through Fourier transform, and different frequency components and amplitude values ​​are extracted to obtain the frequency domain heart rate characteristics; Based on the frequency domain heart rate characteristics, the frequency band energy ratio of the target frequency band is calculated using the following formula: Among them, P is the energy proportion of the frequency band, |X(f)| 2 is the power spectral density at frequency f, f l and f h are the lower and upper frequency limits of the target frequency band, f min and f max are the minimum and maximum frequencies of the entire spectrum; Combined with the main frequency component information in the frequency domain, the frequency band energy proportions of multiple frequency bands are counted and summarized into an index set to generate the heart rate variability index; The steps for obtaining the physical sign fluctuation parameters are: The heart rate variability index is classified into two parts: time domain features and frequency domain features. The time domain features including standard deviation and root mean square interval are statistically analyzed, and the frequency domain features including frequency band energy ratio and main frequency are classified to obtain the classified heart rate variability index. Based on the classified heart rate variability indicators, calculating the change trend and fluctuation range of each time domain feature and frequency domain feature, and the extreme value, range and average change between the statistical indicators to obtain the change trend of the statistical indicators; Based on the changing trend of the statistical indicators, the physical sign fluctuation parameters are calculated using the following formula: Among them, T is the fluctuation parameter of physical signs, ΔD i is the change value of the time domain feature of the i-th item, ΔF i is the change value of the i-th frequency domain feature, P i is the proportion of the i-th frequency domain feature, and k is the total number of features.

2. The wearable underwater vital signs and environment monitoring system according to claim 1, characterized in that: The steps of acquiring the physiological signal sequence are: Based on the multi-source physiological signals, performing timing alignment, aligning each signal according to the timestamp, and adjusting each signal data by interpolation or delay according to the acquisition time point of each signal to obtain a time-aligned multi-source physiological signal; Data synthesis is performed based on the time-series aligned multi-source physiological signals, and all aligned signals are integrated into a unified time point through data fusion to generate a physiological signal sequence.

3. The wearable underwater vital signs and environment monitoring system according to claim 1, characterized in that: The steps of obtaining the feature component set are: According to the physiological signal sequence, a two-dimensional matrix is ​​generated according to the signal values ​​sampled at each time point, the data at each time point is used as a row vector, and the signal channel data is used as a column vector to obtain a signal trajectory matrix; According to the signal trajectory matrix, the characteristic score of the signal decomposition is calculated using the following formula: Among them, S is the feature score, λ i is the singular value of the signal trajectory matrix, u i is the corresponding left singular vector, zk is the number of blocks; According to the feature scores, matching components of the feature values ​​and feature vectors are extracted, and the block feature scores are aggregated to generate corresponding signal feature value and feature vector sets to obtain a feature component set.

4. The wearable underwater vital signs and environment monitoring system according to claim 1, characterized in that: The steps for obtaining the reconstructed signal group are: Extracting eigenvectors and corresponding singular values ​​according to the set of eigencomponents, and rearranging the eigenvectors and singular values ​​in a matrix form to generate an eigenvector matrix; Based on the eigenvector matrix, the energy proportion is calculated using the following formula: Where E is the energy density, q j is the jth component value of the reconstructed signal, q i is the i-th component value in the eigenvector set, m is the number of principal components selected, and n is the total number of eigenvectors; The eigenvectors whose energy proportions are higher than a threshold are weightedly combined with corresponding singular values, and inverse singular value decomposition is performed to generate a reconstructed signal group.

5. The wearable underwater vital signs and environment monitoring system according to claim 1, characterized in that: The steps for obtaining the underwater vital sign state quantity are: According to the vital sign fluctuation parameters, the characteristics are divided into time domain characteristics and frequency domain characteristics according to the characteristic type, each change value in the time domain characteristics and each energy value in the frequency domain characteristics are extracted respectively, the data are normalized according to a unified range, and the normalized data are recombined to form a normalized vital sign fluctuation parameter set; Based on the normalized vital sign fluctuation parameter set, the similarity between each feature is calculated through the correlation between features, and the correlation relationship between the features is expressed in the form of a matrix. The features are merged and recalculated to obtain a fused vital sign fluctuation parameter set; According to the fused set of vital sign fluctuation parameters, weight values ​​are assigned according to the classification standards of time domain features and frequency domain features, and the features after weight assignment are summed up to generate underwater vital sign state quantities.

Images