Respiratory rhythm classification method and system

Abstract

The application discloses a respiration rhythm classification method and system, and the method comprises the following steps: providing a respiration rhythm classification method, acquiring and segmenting an effective respiration signal in a unit sampling period, and determining a corresponding respiration frequency difference sequence; calculating an average value corresponding to the respiration frequency difference sequence, and calculating a dynamic classification threshold feature corresponding to the respiration frequency difference sequence according to the average value; calculating a maximum value and a standard deviation corresponding to the respiration frequency difference sequence, and determining a respiration rhythm type corresponding to the respiration signal according to the maximum value and the standard deviation and the dynamic classification threshold feature. The method dynamically determines a suitable classification threshold based on the average value, can adapt to the respiration frequency changes of different people, realizes intelligent classification of the respiration type, improves the robustness, accuracy and reliability of the classification method, has strong adaptability, and can also provide a diagnosis basis for doctors.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to a method and system for classifying respiratory rhythms. Background Technology

[0002] In the medical field, respiration is one of the fundamental processes of life, maintaining the body's gas exchange balance by supplying oxygen and removing carbon dioxide. However, respiration is not merely a mechanical physiological process; it is also regulated and influenced by a variety of factors, including the nervous system, cardiovascular system, metabolic processes, and psychological state. A normal respiratory rhythm is a complex physiological phenomenon, usually characterized by a regular and stable breathing pattern, but it can change under different circumstances.

[0003] Currently, the method involves acquiring real-time respiratory waveform data, extracting respiratory features from the data, and inputting these features into a pre-trained classification model to determine the corresponding classification type. However, this method requires a large amount of sample data, and the classification result largely depends on the selected respiratory signal features, resulting in low classification accuracy. Furthermore, it fails to adequately consider individual differences, making it difficult to establish a unified model to adapt to the respiratory rhythm changes of different populations, thus exhibiting poor adaptability. Summary of the Invention

[0004] One of the objectives of this invention is to provide a method for classifying respiratory rhythm types, in order to solve the technical problem in the prior art that it lacks the ability to accurately distinguish respiratory types by monitoring the rhythm of respiratory signals, mining and extracting respiratory health information and related characteristics from changes in respiratory rhythm.

[0005] One of the objectives of this invention is to provide a respiratory rhythm type classification system.

[0006] To achieve one of the above-mentioned objectives, the present invention provides a respiratory rhythm classification method, comprising: acquiring and segmenting effective respiratory signals within a unit sampling period, and determining the corresponding respiratory frequency difference sequence; calculating the average value corresponding to the respiratory frequency difference sequence, and calculating a dynamic classification threshold feature corresponding to the respiratory frequency difference sequence based on the average value; wherein the dynamic classification threshold feature is used to represent the fluctuation range and fluctuation amplitude of the respiratory frequency within the unit sampling period; calculating the maximum value and standard deviation corresponding to the respiratory frequency difference sequence, and determining the respiratory rhythm type corresponding to the respiratory signal based on the maximum value, standard deviation, and the dynamic classification threshold feature.

[0007] As a further improvement of one embodiment of the present invention, the respiratory frequency differential sequence is used to characterize the change in respiratory frequency between two adjacent signal segments.

[0008] As a further improvement of one embodiment of the present invention, the step of "acquiring and segmenting the effective respiratory signal within a unit sampling period and determining the corresponding respiratory frequency difference sequence" specifically includes: controlling the signal acquisition device to acquire respiratory signals within the unit sampling period, and performing data processing operations on the respiratory signals to filter and obtain the effective respiratory signals; wherein, the data processing operations include at least one of respiratory signal conversion operations and respiratory signal filtering operations; performing signal segmentation operations on the effective respiratory signals based on a respiratory signal segmentation algorithm to obtain multiple sets of respiratory signal segments; wherein, each set of respiratory signal segments contains the same sampling time length; calculating and determining the respiratory frequency difference sequence based on multiple sets of respiratory frequencies corresponding to the multiple sets of respiratory signal segments.

[0009] As a further improvement of one embodiment of the present invention, the step of "performing data processing operations on the respiratory signal and filtering to obtain the effective respiratory signal" specifically includes: performing an analog-to-digital conversion operation on the respiratory signal to obtain a discrete digital sequence corresponding to the respiratory signal; and using a signal filtering algorithm to perform a filtering operation on the discrete digital sequence to filter and obtain an effective respiratory signal that meets the first preset respiratory frequency threshold range; wherein, the first preset respiratory frequency threshold range is 0.1Hz-0.5Hz.

[0010] As a further improvement of one embodiment of the present invention, the step of "calculating and determining the respiratory frequency difference sequence based on multiple respiratory frequencies corresponding to the multiple sets of respiratory signal segments" specifically includes: performing a Fourier transform operation on the multiple sets of respiratory signal segments based on the Fourier transform method to obtain the corresponding multiple sets of spectral energy distributions; calculating and statistically summing the spectral energy of all respiratory signal segments corresponding to the multiple sets of spectral energy distributions to obtain the total spectral energy; determining whether each set of respiratory frequencies is within the range of a second preset respiratory frequency threshold; wherein, the range of the second preset respiratory frequency threshold is 0.1Hz-1Hz; if so, then using a pseudo-detection algorithm, multiple sets of pseudo-respiratory signal segments are obtained by filtering based on the single spectral energy of each set of respiratory signal segments and the total spectral energy, and multiple sets of respiratory frequencies corresponding to the multiple sets of pseudo-respiratory signal segments are calculated to obtain the respiratory frequency difference sequence.

[0011] As a further improvement of one embodiment of the present invention, the step of "selecting multiple sets of pseudo-respiratory signal segments based on the single-spectrum energy of each corresponding group of respiratory signal segments and the total spectral energy" specifically includes: calculating and determining whether the ratio of the single-spectrum energy to the total spectral energy is greater than a preset ratio threshold; if so, retaining the respiratory signal segment corresponding to the single spectral energy from the multiple sets of pseudo-respiratory signal segments; if not, deleting the respiratory signal segment corresponding to the single spectral energy from the multiple sets of pseudo-respiratory signal segments and updating the multiple sets of pseudo-respiratory signal segments.

[0012] As a further improvement of one embodiment of the present invention, the step of "calculating multiple sets of respiratory frequencies corresponding to the multiple sets of pseudo-respiratory signal segments to obtain the respiratory frequency difference sequence" specifically includes: using the multiple sets of respiratory frequencies corresponding to the multiple sets of pseudo-respiratory signal segments as sequence elements to construct a respiratory frequency sequence; calculating the difference between two adjacent elements in the respiratory frequency sequence respectively, and generating the respiratory frequency difference sequence based on all the differences.

[0013] As a further improvement of one embodiment of the present invention, the step of "calculating the dynamic classification threshold feature corresponding to the respiratory rate difference sequence based on the average value" specifically includes: obtaining optimal decision parameters; wherein, the optimal decision parameters include a first optimal decision parameter and a second optimal decision parameter; calculating a first dynamic classification threshold feature based on the first optimal decision parameter and the average value; and calculating the dynamic classification threshold feature based on the first dynamic classification threshold feature and the second optimal decision parameter; wherein, the first dynamic classification threshold feature is equal to the product of the first optimal decision parameter and the average value, and the dynamic classification threshold feature is equal to the sum of the first dynamic classification threshold feature and the second optimal decision parameter.

[0014] As a further improvement of one embodiment of the present invention, the "obtaining optimal decision parameters" specifically includes: constructing a classification model and a decision function; acquiring and calculating statistical features of several respiratory sample data; inputting the statistical features into the classification model based on the decision function; and calculating the optimal decision parameters of the classification model; wherein the statistical features include at least the average value of the several respiratory sample data.

[0015] As a further improvement of one embodiment of the present invention, the dynamic classification threshold feature includes a dynamic standard deviation threshold; the "construction of classification model and decision function" specifically includes: constructing a first classification model and a first decision function; the "acquiring and calculating statistical features of several respiratory sample data, and inputting the statistical features into the classification model based on the decision function to calculate the optimal decision parameters of the classification model" specifically includes: acquiring and calculating the mean and standard deviation of several respiratory signal sample data, and inputting the mean and standard deviation of the respiratory signal sample data into the first classification model for training to obtain a first trained classification model; calculating the decision boundary when the first decision function is equal to 0 according to the first trained classification model, iteratively optimizing the first decision function according to the decision boundary to obtain the first optimal decision parameters and the second optimal decision parameters.

[0016] As a further improvement of one embodiment of the present invention, the dynamic classification threshold feature includes a dynamic maximum value threshold; the "construction of classification model and decision function" specifically includes: constructing a second classification model and a second decision function; the "acquiring and calculating statistical features of several respiratory sample data, and inputting the statistical features into the classification model based on the decision function to calculate the optimal decision parameters of the classification model" specifically includes: acquiring and calculating the average and maximum values ​​of several respiratory signal sample data, and inputting the average and maximum values ​​of the respiratory signal sample data into the second classification model for training to obtain a second trained classification model; calculating the decision boundary when the second decision function is equal to 0 according to the second trained classification model, iteratively optimizing the second decision function according to the decision boundary to obtain the first optimal decision parameters and the second optimal decision parameters.

[0017] As a further improvement of one embodiment of the present invention, the breathing rhythm type includes at least one of normal breathing, irregular breathing, and shallow breathing; wherein, normal breathing is: a breathing rate of 12-20 times per minute, with consistent time intervals between exhalation and inhalation; irregular breathing is: an unstable breathing rate and breathing depth within a unit sampling period, with inconsistent time intervals between exhalation and inhalation; and shallow breathing is: a breathing rate exceeding 20 times per minute, with minimal changes in lung volume.

[0018] As a further improvement of one embodiment of the present invention, the dynamic classification threshold feature includes at least one of a dynamic standard deviation threshold and a dynamic maximum value threshold; the step of "determining the respiratory rhythm type corresponding to the respiratory signal based on the maximum value, standard deviation, and the dynamic classification threshold feature" specifically includes: determining whether the maximum value is greater than the dynamic maximum value threshold, and / or whether the standard deviation is greater than the dynamic standard deviation threshold; if not, then determining that the respiratory rhythm type corresponding to the respiratory signal is normal breathing; if yes, then calculating the corresponding Poincaré map based on the respiratory frequency difference sequence, extracting and determining the respiratory rhythm type corresponding to the respiratory signal based on the Poincaré features of the Poincaré map.

[0019] As a further improvement to one embodiment of the present invention, the step of "calculating the corresponding Poincaré map based on the respiratory frequency difference sequence" specifically includes: obtaining and determining several sets of position coordinate points, with two adjacent elements in the respiratory frequency difference sequence as the start and end points of the vector; performing polar coordinate transformation on each position coordinate point to obtain multiple sets of corresponding polar angles and polar radii; wherein, the polar angle is the angle between the corresponding position coordinate point and the positive horizontal axis, and the polar radii is the distance from the corresponding position coordinate point to the origin; and drawing multiple vector line segments centered on the origin based on the multiple sets of corresponding polar angles and polar radii to form the Poincaré map.

[0020] As a further improvement of one embodiment of the present invention, the Poincaré feature includes the length feature of the vector line segment and the angle feature of the vector line segment.

[0021] As a further improvement of one embodiment of the present invention, the step of "extracting and determining the respiratory rhythm type corresponding to the respiratory signal based on the Poincaré features corresponding to the Poincaré map" specifically includes: acquiring and calculating the length and angle of all vector line segments in the Poincaré map to obtain multiple sets of vector length values ​​and multiple sets of vector angle values; calculating the grid proportion of all vector line segments based on the multiple sets of vector length values ​​and the multiple sets of vector angle values ​​to determine the angle characteristics of the vector line segments; wherein, the grid proportion is used to characterize the angle distribution of the vector line segments in the Poincaré map; calculating the proportion of the number of vector line segments that satisfy a preset vector length based on the multiple sets of vector length values ​​to determine the length characteristics of the vector line segments; wherein, the proportion of the number is used to characterize the length distribution of the vector line segments in the Poincaré map; and determining the respiratory rhythm type corresponding to the respiratory signal based on the length characteristics and the angle characteristics of the vector line segments.

[0022] As a further improvement to one embodiment of the present invention, the step of "calculating the grid proportion of all vector segments based on the multiple sets of vector length values ​​and the multiple sets of vector angle values, and determining the angular characteristics of the vector segments" specifically includes: obtaining grid division parameters corresponding to the Poincaré map; wherein, the grid division parameters include the radius of the central circle of the Poincaré map, the maximum radius, the length step of the Poincaré map, and the angle step of the Poincaré map; performing a spider web-like grid division operation on the Poincaré map according to the grid division parameters to obtain a spider web-like Poincaré map; obtaining and counting the total number of all grids in the spider web-like Poincaré map, and the total number of grids occupied by the endpoint of each vector segment in the spider web-like Poincaré map; wherein, the endpoint of the vector segment does not include the vector endpoint located on the central circle of the spider web-like Poincaré map; calculating the ratio of the total number of grids to the total number of all grids to obtain the angular characteristics of the vector segments.

[0023] As a further improvement of one embodiment of the present invention, the step of "calculating the proportion of vector segments that satisfy the preset vector length based on the multiple sets of vector length values ​​and determining the length characteristics of the vector segments" specifically includes: counting and determining the maximum value of the multiple sets of vector length values, determining a vector length threshold based on the maximum vector length value; counting the total number of the multiple sets of vector length values ​​and the number of vector segments less than the vector length threshold to obtain the total number of vectors and the number of vector segments; and calculating the ratio of the number of vector segments to the total number of vectors to obtain the length characteristics of the vector segments.

[0024] As a further improvement of one embodiment of the present invention, the step of "determining the respiratory rhythm type corresponding to the respiratory signal based on the length characteristics and angle characteristics of the vector line segment" specifically includes: determining whether the length characteristics of the vector line segment are less than a preset vector length threshold and whether the angle characteristics of the vector line segment are less than a preset vector angle threshold; if yes, then the respiratory rhythm type corresponding to the respiratory signal is determined to be normal breathing; if no, then determining whether the length characteristics of the vector line segment are greater than or equal to the preset vector length threshold and whether the angle characteristics of the vector line segment are greater than or equal to the preset vector angle threshold; if the length characteristics of the vector line segment are greater than or equal to the preset vector length threshold and the angle characteristics of the vector line segment are greater than or equal to the preset vector angle threshold, then the respiratory rhythm type corresponding to the respiratory signal is determined to be irregular breathing; if the length characteristics of the vector line segment are less than the preset vector length threshold and the angle characteristics of the vector line segment are less than the preset vector angle threshold, then the respiratory rhythm type corresponding to the respiratory signal is determined to be shallow breathing.

[0025] To achieve one of the above-mentioned objectives, one embodiment of the present invention provides a respiratory rhythm classification system, the respiratory rhythm classification system comprising: a memory and a processor, the memory having a computer program capable of running on the processor, the processor executing the program to implement the steps of the respiratory rhythm classification method described in any one of the above-mentioned embodiments.

[0026] Compared with the prior art, the embodiments of the present invention have at least one of the following beneficial effects:

[0027] This invention employs a respiratory rhythm type classification method. By segmenting the respiratory signal within a unit sampling period, continuous respiratory signals can be analyzed, and the respiratory signal is divided into several data sequence segments, facilitating the calculation of the average, maximum, and standard deviation of the subsequent respiratory frequency difference sequence. Furthermore, by calculating the average value of the respiratory frequency difference sequence, the overall trend of respiratory frequency variation can be reflected, avoiding classification errors caused by random errors. Based on this, a suitable classification threshold can be dynamically determined, adapting to the respiratory frequency variations of different populations, achieving intelligent classification of respiratory types, improving the robustness, accuracy, and reliability of the classification method, and exhibiting strong adaptability. It can also provide diagnostic basis for doctors. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the steps of a respiratory rhythm classification method according to an embodiment of the present invention.

[0029] Figure 2 This is a schematic diagram of the steps of the respiratory rhythm classification method in the first embodiment of the present invention.

[0030] Figure 3This is a detailed schematic diagram of step S13 of the respiratory rhythm classification method in one embodiment of the present invention.

[0031] Figure 4 This is a schematic diagram of some steps in the respiratory rhythm classification method in the second embodiment of the present invention.

[0032] Figure 5(a) is a schematic diagram of the standard deviation statistics of the respiratory rate difference sequence of the respiratory rhythm classification method in one embodiment of the present invention.

[0033] Figure 5(b) is a statistical diagram showing the relationship between the average and maximum values ​​of the respiratory rate difference sequence in a respiratory rhythm classification method according to an embodiment of the present invention.

[0034] Figure 5(c) is a statistical diagram showing the relationship between the mean and standard deviation of the respiratory rate difference sequence in a respiratory rhythm classification method according to an embodiment of the present invention.

[0035] Figure 5(d) is a statistical diagram showing the relationship between the average and maximum values ​​of the respiratory rate difference sequence in a respiratory rhythm classification method according to an embodiment of the present invention.

[0036] Figure 6 This is a schematic diagram of the steps of the respiratory rhythm classification method in the third embodiment of the present invention.

[0037] Figure 7 This is a schematic diagram of step S32B of the respiratory rhythm classification method in one embodiment of the present invention.

[0038] Figure 8 This is a schematic diagram of step S32B of the respiratory rhythm classification method in one embodiment of the present invention.

[0039] Figure 9 This is a schematic diagram of step S32B22 of the respiratory rhythm classification method in one embodiment of the present invention.

[0040] Figure 10 This is a schematic diagram of step S32B23 of the respiratory rhythm classification method in one embodiment of the present invention.

[0041] Figure 11(a) is a schematic diagram of the Poincaré mapping corresponding to the normal respiratory rhythm in a respiratory rhythm classification method according to an embodiment of the present invention.

[0042] Figure 11(b) is a schematic diagram of the Poincaré map corresponding to the irregular respiratory rhythm in a respiratory rhythm classification method according to an embodiment of the present invention.

[0043] Figure 11(c) is a schematic diagram of the Poincaré mapping corresponding to occasional respiratory rhythms in a respiratory rhythm classification method according to an embodiment of the present invention.

[0044] Figure 11(d) is a schematic diagram of the Poincaré mapping corresponding to shallow breathing rhythms in a breathing rhythm classification method according to an embodiment of the present invention.

[0045] Figure 12 This is a flowchart illustrating a preferred embodiment of the respiratory rhythm classification method according to one aspect of the present invention. Detailed Implementation

[0046] The present invention will now be described in detail with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the scope of protection of the present invention.

[0047] It should be noted that the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the description of specific embodiments of the present invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0048] Different physiological and pathological conditions can lead to changes in respiratory rhythm. For example, breathing patterns may change significantly during exercise, sleep, stress response, or when suffering from respiratory diseases or cardiovascular problems. Therefore, research on respiratory rhythm classification methods is crucial for understanding normal breathing and breathing patterns associated with disease states, and has important clinical and scientific value in promoting medical science and improving patient care.

[0049] Based on this, the present invention provides a method for classifying respiratory rhythms, such as... Figure 1 As shown, the respiratory rhythm classification method specifically includes the following steps:

[0050] Step S1: Acquire and segment the effective respiratory signal within a unit sampling period, and determine the corresponding respiratory rate differential sequence;

[0051] Step S2: Calculate the average value corresponding to the respiratory rate difference sequence, and calculate the dynamic classification threshold feature corresponding to the respiratory rate difference sequence based on the average value;

[0052] Step S3: Calculate the maximum value and standard deviation corresponding to the respiratory rate difference sequence, and determine the respiratory rhythm type corresponding to the respiratory signal based on the maximum value, standard deviation and dynamic classification threshold feature.

[0053] Thus, by calculating the average value of the respiratory rate difference sequence, the overall trend of respiratory rate change can be reflected, and a suitable classification threshold can be dynamically determined accordingly to achieve intelligent classification of respiratory types and improve the robustness and reliability of the classification method.

[0054] The respiratory rate difference sequence characterizes the change in respiratory rate between two adjacent signal segments within a unit sampling period; in other words, it reflects the trend of respiratory rate change over time. The dynamic classification threshold feature can be used to represent the fluctuation range and amplitude of respiratory rate within the unit sampling period.

[0055] The respiratory rhythm refers to the rhythmic pattern of breathing, including the ratio of inhalation time to exhalation time and the variation pattern of respiratory depth. The respiratory rhythm can reflect not only the speed of breathing, but also the depth or proportion of breathing.

[0056] Based on this, in this invention, the breathing type may include at least one of normal breathing, irregular breathing, and shallow breathing. Specifically, normal breathing may refer to a respiratory rate of 12-20 breaths per minute, with consistent time intervals between exhalation and inhalation, that is, regular respiratory depth and rhythm within a unit sampling period; irregular breathing may refer to unstable respiratory rate and depth within a unit respiratory period, with inconsistent time intervals between exhalation and inhalation, that is, intermittent changes in respiratory rate and depth within a unit respiratory period, and irregular respiratory rhythm; shallow breathing may refer to a respiratory rate exceeding 20 breaths per minute with minimal changes in lung volume, that is, shallow respiratory depth within a unit sampling period.

[0057] To improve classification accuracy and reduce computational costs, data processing operations need to be performed on the collected respiratory signals, combined with... Figure 1 and Figure 2 As shown, in the first embodiment, the respiratory rhythm classification method may specifically include the following steps:

[0058] Step S11: The control signal acquisition device acquires respiratory signals within the unit sampling period, performs data processing operations on the respiratory signals, and filters out the valid respiratory signals;

[0059] Step S12: Based on the respiratory signal segmentation algorithm, perform signal segmentation operation on the effective respiratory signal to obtain multiple sets of respiratory signal segments;

[0060] Step S13: Calculate and determine the respiratory frequency difference sequence based on the multiple respiratory frequencies corresponding to the multiple sets of respiratory signal segments.

[0061] Thus, by performing signal segmentation, noise and outliers in the respiratory signal can be removed, making subsequent frequency analysis more accurate and reliable.

[0062] Each respiratory signal segment contains the same sampling time length; in other words, each respiratory signal segment includes the same number of sampling points. In one embodiment, the real-time acquired data can be buffered in fixed time segments, thus obtaining several respiratory signal segments of equal length. In another embodiment, a respiratory signal segment is cut out every preset number of sampling points according to the sampling frequency, also resulting in several respiratory signal segments of equal length. Furthermore, the segment length of the respiratory signal can be adjusted as needed, and this invention does not impose specific limitations.

[0063] Furthermore, the data processing operations include at least one of respiratory signal conversion operations and respiratory signal filtering operations. Specifically, in one embodiment, a signal conversion operation is performed on the acquired respiratory signal; in another embodiment, a filtering operation may be performed on the acquired respiratory signal; in a preferred embodiment, the two embodiments described above may be used in combination.

[0064] In a preferred embodiment, the part of step S12 that "performs signal segmentation operation on the effective respiratory signal to obtain multiple sets of respiratory signal segments" may specifically include the following steps:

[0065] Step S121: Perform an analog-to-digital conversion operation on the respiratory signal to obtain a discrete digital sequence corresponding to the respiratory signal;

[0066] Step S122: A signal filtering algorithm is used to perform a filtering operation on the discrete digital sequence to obtain an effective respiratory signal that meets the first preset respiratory frequency threshold range.

[0067] Thus, performing digital-to-analog-to-digital conversion and filtering on the respiratory signal can filter out noise components in the respiratory signal, improving the signal-to-noise ratio and quality of the respiratory signal.

[0068] The normal respiratory signal frequency is approximately 12-20 times per minute, corresponding to a frequency range of 0.2-0.33 Hz. Preferably, the first preset respiratory frequency threshold range can be 0.1 Hz-0.5 Hz, thus retaining respiratory signal segments with respiratory frequencies within the first preset respiratory frequency threshold range and filtering out irrelevant high-frequency noise, thereby obtaining an effective respiratory signal (i.e., corresponding to "filtering out effective respiratory signals that meet the first preset respiratory frequency threshold range" as described in step S122).

[0069] The analog-to-digital conversion operation refers to converting the acquired analog respiratory signal (i.e., the respiratory signal corresponding to step S121) into a digital signal (i.e., the discrete digital sequence described in step S121), and then performing a filtering operation on the converted digital signal based on a filtering algorithm to obtain a valid respiratory signal. Optionally, the filtering algorithm may be a bandpass digital filter based on FIR (Finite Impulse Response), with its passband frequency set to 0.1Hz-0.5Hz, i.e., set to the first preset respiratory frequency threshold range. Of course, this invention does not exclude other filtering algorithms.

[0070] Combination Figure 2 and Figure 3 As shown, in one embodiment, the present invention provides a refined step for step S13, which may specifically include:

[0071] Step S131: Based on the Fourier transform method, perform Fourier transform operation on the multiple sets of respiratory signal segments to obtain the corresponding multiple sets of spectral energy distributions;

[0072] Step S132: Calculate and sum the spectral energies of all respiratory signal segments corresponding to the multiple sets of spectral energy distributions to obtain the total spectral energy;

[0073] Step S133: Determine whether the respiratory rate of each group is within the range of the second preset respiratory rate threshold.

[0074] If so, proceed to step S134, employ a pseudo-detection algorithm, and filter out multiple sets of pseudo-respiratory signal segments based on the single-spectrum energy and the total spectral energy of each corresponding set of respiratory signal segments, calculate multiple sets of respiratory frequencies corresponding to the multiple sets of pseudo-respiratory signal segments, and obtain the respiratory frequency difference sequence.

[0075] Thus, by using the artifact detection algorithm, the influence of artifact signals on the recognition results can be eliminated, improving measurement accuracy and respiratory signal quality. At the same time, this method is simple to implement and simplifies the subsequent processing of respiratory rate differential sequences.

[0076] The second preset respiratory rate threshold range can preferably be set to 0.1Hz-1Hz. Noise generated by human shaking or movement is below 0.1Hz, while the main energy of a real respiratory signal is concentrated within the second preset respiratory rate threshold range of 0.1Hz-1Hz. Therefore, this range can encompass the main frequency components of both normal and abnormal respiratory states, making it suitable for analyzing respiratory signals.

[0077] The aforementioned artifact detection algorithm is a method for detecting outliers or abnormal behaviors in data. It can model and predict data, and identify anomalies by comparing the differences between actual observations and predicted values. It should be noted that the artifact detection algorithm processes respiratory signal segments that meet certain respiratory rate criteria, thus reducing computational load and ensuring accurate and reliable calculations.

[0078] Furthermore, in one embodiment, the present invention provides a detailed step for the part in step S134 described as "selecting multiple sets of pseudo-respiratory signal segments based on the single-spectral energy of each corresponding group of respiratory signal segments and the total spectral energy," which may specifically include:

[0079] Step S13411: Calculate and determine whether the ratio of the single-spectrum energy to the total spectrum energy is greater than a preset ratio threshold.

[0080] If so, proceed to step S13412A to retain the respiratory signal segment corresponding to the single spectral energy from the multiple sets of pseudo-respiratory signal segments;

[0081] If not, proceed to step S13412B, delete the respiratory signal segment corresponding to the single spectral energy from the multiple sets of pseudo-respiratory signal segments, and update the multiple sets of pseudo-respiratory signal segments.

[0082] In this way, effective respiratory signal segments are obtained through screening, simplifying the subsequent calculation process of the respiratory rate difference sequence and improving the reliability of the data. Preferably, the preset ratio threshold may include 0.1.

[0083] In another embodiment, the present invention provides a detailed step for the part of step S134 described as "screening multiple sets of pseudo-respiratory signal segments based on the single-spectral energy of each corresponding group of respiratory signal segments and the total spectral energy," which may specifically include:

[0084] Step S13421: Use the multiple sets of respiratory frequencies corresponding to the multiple sets of pseudo-respiratory signal segments as sequence elements to construct a respiratory frequency sequence;

[0085] Step S13422: Calculate the difference between two adjacent elements in the respiratory frequency sequence, and generate the respiratory frequency difference sequence based on all the differences.

[0086] Thus, by calculating the difference, the dynamic change pattern of respiratory rate can be reflected, improving the ability to express respiratory characteristics in a simple and efficient way.

[0087] For example, assuming that real-time collected data is cached into a respiratory signal segment every 120 seconds within a unit sampling period, this can be divided into M respiratory signal segments. The respiratory frequency, single-spectrum energy, and total spectral energy of each of the M respiratory signal segments are calculated. Based on a pseudo-detection algorithm, respiratory signal segments whose respiratory frequencies do not fall within the 0.1-1Hz range are filtered out, resulting in N respiratory signal segments. The ratio of the single-spectrum energy to the total spectral energy of each respiratory signal segment is calculated, and signal segments with a ratio less than or equal to 0.1 are selected from the N respiratory signal segments to obtain the corresponding respiratory frequency sequence, which can be defined as RR. The difference between the respiratory frequency sequences RR is calculated to obtain the respiratory frequency difference sequence, which can be defined as ΔRR. Here, M is greater than or equal to N.

[0088] Combination Figure 1 and Figure 4 As shown, in the second embodiment, the respiratory rhythm classification method may further include the following steps:

[0089] Step S21: Obtain the optimal decision parameters;

[0090] Step S22: Calculate the first dynamic classification threshold feature based on the first optimal decision parameter and the average value; calculate the dynamic classification threshold feature based on the first dynamic classification threshold feature and the second optimal decision parameter.

[0091] In this way, by calculating the dynamic classification threshold features, the system can adaptively adjust according to the real-time changes in respiratory signals, thereby enhancing the adaptive ability of classification and improving the robustness of the algorithm.

[0092] The optimal decision parameters include a first optimal decision parameter and a second optimal decision parameter; the first dynamic classification threshold feature is equal to the product of the first optimal decision parameter and the average value, and the dynamic classification threshold feature is equal to the sum of the first dynamic classification threshold feature and the second optimal decision parameter.

[0093] The optimal decision parameters can be obtained based on human experience or statistical analysis of multiple trials. Therefore, step S21 can specifically include the following steps:

[0094] Step S21: Construct the classification model and decision function;

[0095] Step S22: Obtain and calculate the statistical features of several respiratory sample data, and input the statistical features into the classification model based on the decision function to calculate the optimal decision parameters of the classification model.

[0096] In this way, extracting the average statistical features of respiratory signals can represent the central trend of the overall sample, thereby reducing training costs and retaining key feature information, and improving the model's generalization ability.

[0097] The decision function refers to a function in machine learning and statistical models that uses input variables to calculate predicted output values. It is implemented based on training data, maps input vectors or features to different classification or regression results, and serves as the basis for judging the results.

[0098] Furthermore, the statistical characteristics include at least the average value of the plurality of respiratory sample data. In one embodiment, a fixed-size classification threshold can be set, and the data can be classified according to the standard deviation ΔRR of the respiratory rate difference sequence. std and the maximum value ΔRR max Statistical features were used to extract and distinguish the characteristics of normal and irregular respiratory rhythms, as shown in Figures 5(a) and 5(b).

[0099] To improve the flexibility and accuracy of classification, in one embodiment, the dynamic classification threshold feature may include at least one of a dynamic standard deviation threshold and a dynamic maximum value threshold. Specifically, in a first embodiment, the dynamic classification threshold feature may include a dynamic standard deviation threshold; the statistical feature may include the mean and standard deviation of the plurality of respiratory sample data; based on this, step S21 may specifically include the following steps:

[0100] Step S21': Construct the first classification model and the first decision function;

[0101] Step S22 may specifically include the following steps:

[0102] Step S2211: Obtain and calculate the average value and standard deviation of several respiratory signal sample data, and input the average value and standard deviation of the respiratory signal sample data into the first classification model for training to obtain the first trained classification model;

[0103] Step S2212: Based on the first trained classification model, calculate the decision boundary when the first decision function equals 0, and iteratively optimize the first decision function based on the decision boundary to obtain the first optimal decision parameters and the second optimal decision parameters.

[0104] Thus, using effective statistical features (i.e., mean and standard deviation) can accelerate the training process and prevent overfitting; in addition, by continuously adjusting the parameters through the decision boundary, the parameters converge to the global optimum.

[0105] For ease of description, the average value of the respiratory rate difference sequence is defined as ΔRR. meanDefine the dynamic standard deviation threshold as STD_th, and define the first optimal decision parameter and the second optimal decision parameter as a1 and b1, respectively. Based on the above, the dynamic standard deviation threshold STD_th can at least satisfy formula (1):

[0106] STD th =a1*ΔRR mean +b1 (1).

[0107] In the second embodiment, the dynamic classification threshold feature may include a dynamic maximum value threshold; the statistical feature may include the average and maximum values ​​of the plurality of respiratory sample data; based on this, step S21 may specifically include the following steps:

[0108] Step S21”: Construct the second classification model and the second decision function;

[0109] Step S22 may specifically include the following steps:

[0110] Step S2221: Obtain and calculate the average and maximum values ​​of several respiratory signal sample data, and input the average and maximum values ​​of the respiratory signal sample data into the second classification model for training to obtain the second trained classification model;

[0111] Step S2222: Based on the second trained classification model, calculate the decision boundary when the second decision function equals 0, and iteratively optimize the second decision function based on the decision boundary to obtain the first optimal decision parameters and the second optimal decision parameters.

[0112] Thus, using effective statistical features (i.e., average and maximum values) can accelerate the training process and prevent overfitting; in addition, by continuously adjusting the parameters through the decision boundary, the parameters converge to the global optimum.

[0113] For ease of description, the dynamic maximum threshold is defined as MAX_th, and the first optimal decision parameter and the second optimal decision parameter are defined as a2 and b2, respectively. Based on the above, the dynamic maximum threshold MAX_th can at least satisfy formula (2):

[0114] MAX th =a2*ΔRR mean +b2 (2).

[0115] In a preferred embodiment, the two embodiments described above can be used in combination, that is, the classification criteria for respiratory rhythm are determined by jointly using the dynamic standard deviation threshold and the dynamic maximum value threshold, so that the classification results are more accurate and have better effects.

[0116] It should be noted that the first optimal decision parameter and the second optimal decision parameter in the first embodiment, and the first optimal decision parameter and the second optimal decision parameter in the second embodiment are two completely different decision parameters. Furthermore, steps S21' and S21" can be understood as derived steps of step S21; steps S2211 to S2212 and steps S2221 to S2222 can be understood as derived steps of step S22; steps S22' to S2212 and steps S22" to S2222 do not have a logical order and can be executed concurrently.

[0117] For example, as shown in Figures 5(c) and 5(d), assuming a respiratory signal segment is buffered every 120 seconds, and 20 valid respiratory signal segments are buffered within a unit sampling period, the respiratory rate difference sequence ΔRR can be calculated. The average value ΔRR of this respiratory rate difference sequence ΔRR is then calculated and statistically analyzed. mean Standard deviation ΔRR std and the maximum value ΔRR max According to formulas (1) and (2) respectively, ΔRR mean Plotting the graph with the dynamic standard deviation threshold STD_th or the dynamic maximum threshold MAX_th as the vertical axis can distinguish the regional distribution of normal respiratory rate and irregular respiratory rhythm. The dynamic thresholds are the linear functions shown by the dashed lines in Figures 5(c) and 5(d) (i.e., formulas (1) and (2)).

[0118] It should be noted that, because the Poincaré model scatter plot can intuitively reveal the dynamic relationship between points and reveal the essential characteristics and differences of different respiratory rhythms in a relatively short time, the respiratory rate difference sequence ΔRR can be mapped onto the Poincaré plot to further extract the relevant features of respiratory rhythms and achieve more accurate rhythm type classification.

[0119] Based on this, combined Figure 1 and Figure 6 As shown, in the third embodiment, the respiratory rhythm classification method may further include the following steps:

[0120] Step S31: Determine whether the maximum value is greater than the dynamic maximum value threshold, and / or whether the standard deviation is greater than the dynamic standard deviation threshold;

[0121] If not, proceed to step S32A to determine that the respiratory rhythm type corresponding to the respiratory signal is normal breathing;

[0122] If so, proceed to step S32B, calculate the corresponding Poincaré map based on the respiratory frequency difference sequence, extract and determine the respiratory rhythm type corresponding to the respiratory signal based on the Poincaré features of the Poincaré map.

[0123] Thus, by employing the Poincaré map, hidden respiratory features can be detected, increasing the accuracy and intelligence of respiratory rhythm classification.

[0124] The Poincaré map, also known as a Poincaré section, is a graphical representation method used to study dynamical systems. It is created by projecting the trajectory of the dynamical system onto a low-dimensional cross-section, specifically a two-dimensional plane. The Poincaré features refer to the characteristics or phenomena observed in the Poincaré map. In this invention, by analyzing and extracting Poincaré features, respiratory signals can be further analyzed and predicted, hidden respiratory features can be uncovered, and thus more accurate classification of respiratory rhythm types can be achieved.

[0125] In one embodiment, if the maximum value ΔRR of the respiratory rate difference sequence max The value is greater than the dynamic maximum value threshold MAX_th, and / or the standard deviation ΔRR of the respiratory rate difference sequence. std If the respiratory rate is greater than the dynamic standard deviation threshold STD_th, then the respiratory rhythm type corresponding to the respiratory signal is determined to be normal breathing.

[0126] Specifically, in the first embodiment, the maximum value of the respiratory rate difference sequence is greater than the dynamic maximum value threshold (i.e., ΔRR). max >MAX_th). In this way, normal respiratory rhythm types can be distinguished.

[0127] In the second embodiment, the standard deviation of the respiratory rate difference sequence is greater than the dynamic standard deviation threshold (i.e., ΔRR). std >STD_th). In this way, normal respiratory rhythm types can be distinguished.

[0128] In the third embodiment, the maximum value of the respiratory rate difference sequence is greater than the dynamic maximum value threshold, and the standard deviation of the respiratory rate difference sequence is greater than the dynamic standard deviation threshold (i.e., ΔRR). max MAX th ,ΔRR std >STD_th). This allows for more accurate differentiation of normal respiratory rhythm types. The third embodiment is also a preferred embodiment, offering even better classification results.

[0129] Furthermore, such as Figure 7As shown, in one embodiment, the part of step S32B that "calculates the corresponding Poincaré map based on the respiratory rate difference sequence" may specifically include the following steps:

[0130] Step S32B11: Obtain and determine several sets of position coordinate points, taking the two adjacent elements in the respiratory frequency difference sequence as the start and end points of the vector;

[0131] Step S32B12: Perform polar coordinate transformation on each position coordinate point to obtain multiple sets of corresponding polar angles and polar radii;

[0132] Step S32B13: Centered on the origin, draw multiple vector line segments according to the multiple sets of corresponding polar angles and polar radii to form the Poincaré map.

[0133] This allows for the conversion of respiratory rate difference sequences to Poincaré maps, facilitating further analysis and extraction of respiratory features.

[0134] Wherein, the polar angle is the angle between the corresponding position coordinate point and the positive horizontal axis, and the polar radius is the distance from the corresponding position coordinate point to the origin. Specifically, based on the respiratory rate difference sequence ΔRR, a polar coordinate transformation operation is performed to define the horizontal coordinate x = ΔRR. n y = ΔRR n+1 Then, the polar radius (or radius) corresponding to the nth point in the Poincaré map can be calculated by formula (3):

[0135]

[0136] Based on the knowledge of mathematical trigonometric transformation, the corresponding phase information, i.e., the corresponding polar angle θ, is extracted according to formula (4). Formula (4) is defined as follows:

[0137]

[0138] Furthermore, the Poincaré feature may include the length feature and the angle feature of the vector line segment. Specifically, the length feature of the vector line segment may refer to the Euclidean distance between the endpoints of the vector line segment, which can represent the amplitude of signal change. In this invention, its unit can be set to seconds, meaning that the difference in respiratory frequency between two adjacent sampling points within a certain time period reaches the corresponding differential amplitude. The angle feature may refer to the angle between adjacent line segments, which can represent the change of respiratory signal in different directions.

[0139] Similarly, as Figure 8 As shown, in one embodiment, the part of step S32B that "extracts and determines the respiratory rhythm type corresponding to the respiratory signal based on the Poincaré features corresponding to the Poincaré map" may specifically include the following steps:

[0140] Step S32B21: Obtain and calculate the length and angle of all vector line segments in the Poincaré map to obtain multiple sets of vector length values ​​and multiple sets of vector angle values;

[0141] Step S32B22: Calculate the grid proportion of all vector line segments based on the multiple sets of vector length values ​​and the multiple sets of vector angle values, and determine the angular characteristics of the vector line segments;

[0142] Step S32B23: Based on the multiple sets of vector length values, calculate the proportion of vector line segments that satisfy the preset vector length, and determine the length characteristics of the vector line segments;

[0143] Step S32B24: Determine the respiratory rhythm type corresponding to the respiratory signal based on the length characteristics and angle characteristics of the vector line segment.

[0144] Thus, by using the Poincaré mapping to convert respiratory signals into visual features, the complexity of directly analyzing the original signals is avoided. In addition, by calculating the length and angle features of vector line segments, the geometric shape of the graph is fully reflected, improving the stability of the analysis and enabling automatic classification of respiratory rhythm types.

[0145] The grid proportion can be used to characterize the angular distribution of the vector line segments in the Poincaré map; the number proportion can be used to characterize the length distribution of the vector line segments in the Poincaré map.

[0146] To simplify the calculation process, in a preferred embodiment, a meshing operation can be performed on the Poincaré map, thus allowing for quick and accurate calculation of the corresponding mesh proportions and number proportions. Based on this, in one implementation, such as... Figure 9 As shown, for step S32B22, the present invention provides a refined step, which may specifically include:

[0147] Step S32B221: Obtain the grid division parameters corresponding to the Poincaré map;

[0148] Step S32B222: Perform a spider meshing operation on the Poincaré map according to the meshing parameters to obtain a spider mesh Poincaré map;

[0149] Step S32B223: Obtain and count the total number of all grids in the spider web Poincaré map, and the total number of grids occupied by the endpoint of each vector line segment in the spider web Poincaré map;

[0150] Step S32B224: Calculate the ratio of the total number of grids to the total number of all grids to obtain the angular characteristics of the vector line segment.

[0151] Thus, grid division can provide more refined and comprehensive angular distribution statistics with high computational efficiency; in addition, grid proportion features can provide richer, more intuitive and quantifiable visual features, which is beneficial to improving the classification ability of Poincaré map graphics.

[0152] The meshing parameters may include the central circle radius, maximum radius, length step size, and angular step size of the Poincaré map. The central circle radius and maximum radius of the Poincaré map determine the circular regions to be divided, while the length step size and angular step size determine the mesh density.

[0153] Specifically, the length step of the Poincaré map determines the radial density of the grid; a smaller length step results in a denser radial distribution. Similarly, the angular step of the Poincaré map determines the angular density of the grid; a smaller angular step results in a denser distribution along a fixed radial direction. Referring to Figure 11(a), the corresponding central circle radius can be set to 0.5s, the maximum radius to 2s, the length step to 0.5s, and the angular step to 30°. These four parameters can be determined through statistical analysis of prior data, but other methods of obtaining them are also acceptable.

[0154] Furthermore, the endpoints of the vector segments do not include the endpoints of the vectors located in the central circle of the spider-web Poincaré map; in other words, the central circle of the spider-web Poincaré map can filter out points with stable signals or filter out repetitive information, while points outside the central circle can contain information about changes in the respiratory signal state, thereby allowing the extraction of useful state change features.

[0155] In one implementation, such as Figure 10 As shown, for step S32B23, the present invention provides a refined step, which may specifically include:

[0156] Step S32B231: Statistically determine the maximum value of the multiple sets of vector length values, and determine the vector length threshold based on the maximum vector length value;

[0157] Step S32B232: Count the total number of the multiple sets of vector length values ​​and the number of vector segments that are less than the vector length threshold to obtain the total number of vectors and the number of vector segments.

[0158] Step S32B233: Calculate the ratio of the number of vector line segments to the total number of vectors to obtain the length characteristics of the vector line segments.

[0159] Thus, by calculating the proportion of vector line segment length distribution, the characteristics of vector line segment length in the Poincaré map can be quickly and effectively analyzed and determined. The calculation method is simple and easy to implement.

[0160] Specifically, the maximum value of the multiple sets of vector length values ​​can be defined as `max_len`, and the vector length threshold can be defined as `len_thr`. Preferably, `len_thr` = 0.25 * `max_len`. The total number of all vector segments in the spiderweb-like Poincaré map is defined as `N`, and the number of vector segments with length values ​​less than `len_thr` in the multiple sets of vector length values ​​is defined as `n`. Based on this, the length feature of the vector segment is defined as `vector_len_ratio`, and the angle feature of the vector segment is defined as `grid_ratio`; wherein, the length feature of the vector segment can be `vector_len_ratio = n / N`.

[0161] It should be noted that the human body may experience abnormal breathing during exercise or emergency response. In such cases, the abnormal breathing is sporadic and not caused by disease. Therefore, it is not essentially a type of abnormal breathing rhythm. Thus, such sporadic abnormal breathing can be identified and excluded from irregular breathing and classified as normal breathing.

[0162] Based on this, in one embodiment, the present invention provides a refined step for step S32B24, which may specifically include:

[0163] Step S32B241: Determine whether the length feature of the vector line segment is less than a preset vector length threshold and whether the angle feature of the vector line segment is less than a preset vector angle threshold.

[0164] If so, proceed to step S32B242A to determine that the respiratory rhythm type corresponding to the respiratory signal is normal breathing;

[0165] If not, proceed to step S32B242B to determine whether the length feature of the vector line segment is greater than or equal to a preset vector length threshold and whether the angle feature of the vector line segment is greater than or equal to a preset vector angle threshold.

[0166] Step S32B243: If the length characteristic of the vector line segment is greater than or equal to the preset vector length threshold, and the angle characteristic of the vector line segment is greater than or equal to the preset vector angle threshold, then the respiratory rhythm type corresponding to the respiratory signal is determined to be irregular breathing.

[0167] Step S32B244: If the length characteristic of the vector line segment is less than the preset vector length threshold and the angle characteristic of the vector line segment is less than the preset vector angle threshold, then the respiratory rhythm type corresponding to the respiratory signal is determined to be shallow breathing.

[0168] By comprehensively considering both vector direction and vector length, respiratory rhythm types can be identified more comprehensively and accurately, thereby improving the accuracy and reliability of classification results.

[0169] For example, suppose several normal breathing segments and several irregular breathing segments are cached within a unit sampling period. Based on the method described above, the Poincaré map corresponding to different breathing rhythm types is calculated and identified, as shown in Figures 11(a), 11(b), 11(c), and 11(d). In these figures, the number of grid cells occupied by the vector line segments marked with hexagonal stars reflects the angular characteristics of the vector line segments.

[0170] Referring to Figure 11(a), because the normal breathing rhythm is relatively uniform and regular, most of the vector line segments are relatively short. That is, the vector length of normal breathing is concentrated within less than 0.5s. The presence of the central circle of the Poincaré map will ignore the vector line segments located inside the central circle. Therefore, the angular characteristics of its vector line segments (i.e., the proportion of the number of grids occupied by the vector endpoint of the vector line segment) do not exceed the range of the central circle. Therefore, the corresponding hexagonal star-shaped grids are few or even non-existent.

[0171] Referring to Figure 11(b), irregular breathing is characterized by irregular fluctuations in the respiratory rhythm. As a result, the vector length of the vector line segment in the Poincaré map corresponding to the irregular respiratory rhythm is much greater than that of the normal respiratory rhythm. That is, the vector length corresponding to the irregular respiratory rhythm is mostly greater than 0.5s. Furthermore, the angular characteristics of the corresponding vector line segment are also chaotic and discrete. Therefore, the number of grids of hexagonal star markers corresponding to the vector endpoints of the vector line segments is much greater than that of the normal respiratory rhythm.

[0172] Referring to Figure 11(c), occasional respiratory abnormalities are random respiratory rhythm abnormalities caused by occasional events. Therefore, in the corresponding Poincaré map, a small portion of the vector lengths of the vector segments exceed 0.5s (corresponding to occasional respiratory rhythm abnormalities, such as exercise or stress responses), but most of the vector lengths are less than 0.5s (corresponding to normal respiratory rhythms). Consequently, the number of hexagonal star-shaped grids corresponding to the vector endpoints of the vector segments is relatively small. This allows for further differentiation between irregular respiratory rhythms and occasional abnormal respiratory rhythms, making the classification results of irregular and normal breathing more accurate and consistent with reality.

[0173] Referring to Figure 11(d), shallow breathing is a type of irregular breathing, such as in asthma. During an asthma attack, the frequency of shallow breathing increases significantly and breathing becomes rapid. Therefore, the respiratory rhythm of asthma shows a special distribution characteristic in the Poincaré map. Specifically, most of the vector line segments are concentrated in the second and fourth quadrants, and the vector endpoints of the vector line segments are all within two length steps. Their corresponding vector angle characteristics are similar (i.e., the vector direction and vector length corresponding to multiple sets of vector line segments are consistent). Therefore, the number of grids with hexagonal star markers corresponding to the vector endpoints is also relatively small. Thus, irregular respiratory rhythms can be further classified as shallow breathing.

[0174] like Figure 12 A flowchart of a preferred embodiment of a respiratory rhythm type classification method is shown. The processing procedure of this preferred embodiment will be summarized below with reference to Figure 11.

[0175] Respiratory signal data is collected within a unit sampling period, and the respiratory signal data is processed to obtain a valid respiratory signal; wherein, the data processing includes analog-to-digital conversion and filtering operations.

[0176] Based on the valid respiratory signal, the corresponding respiratory rate sequence RR is calculated and cached. The respiratory rate sequence RR is then differentiated to obtain a respiratory rate difference sequence ΔRR, and the average value ΔRR of the respiratory rate difference sequence ΔRR is calculated. max Standard deviation ΔRR std and the maximum value ΔRR max .

[0177] Based on the average value of the respiratory rate difference sequence ΔRR, ΔRR max The corresponding dynamic standard deviation threshold STD_th and dynamic maximum value threshold MAX_th are calculated; when the standard deviation ΔRR std When the dynamic standard deviation threshold STD_th is less than or equal to the dynamic standard deviation threshold STD_th, or the maximum value is less than or equal to the dynamic maximum value threshold MAX_th, the rhythm type of the respiratory signal is determined to be normal breathing.

[0178] When the standard deviation ΔRR std When the value is greater than the dynamic standard deviation threshold STD_th and the maximum value is greater than the dynamic maximum value threshold MAX_th, the corresponding Poincaré map is calculated based on the respiratory rate difference sequence ΔRR; the length and angle of all vector line segments in the Poincaré map are counted to obtain multiple sets of vector length values ​​and multiple sets of vector angle values; and the corresponding vector angle feature grid_ratio and vector length feature vector_len_ratio are calculated based on the multiple sets of vector length values ​​and multiple sets of vector angle values.

[0179] If the length feature vector_len_ratio of the vector line segment is greater than or equal to the preset vector length threshold th1, and the angle feature rid_ratio of the vector line segment is greater than or equal to the preset vector angle threshold th2, then the respiratory rhythm type corresponding to the respiratory signal is determined to be irregular breathing; if the length feature vector_len_ratio of the vector line segment is less than the preset vector length threshold th1, and the angle feature rid_ratio of the vector line segment is less than the preset vector angle threshold th2, then the respiratory rhythm type corresponding to the respiratory signal is determined to be shallow breathing.

[0180] The present invention also provides a respiratory rhythm classification system, comprising: a memory and a processor, wherein the memory has a computer program that can run on the processor, and the processor implements the above-described respiratory rhythm classification method when executing the program.

[0181] In summary, the respiratory rhythm classification method provided by this invention can reflect the dynamic changes in respiratory frequency by calculating the respiratory frequency difference sequence, thus improving the expressive ability of respiratory characteristics and being simple and efficient. Furthermore, by calculating the mean, standard deviation, and maximum value of the respiratory frequency difference sequence, it can reflect the statistical characteristics and overall trend of respiratory signals, avoiding classification errors caused by random errors. A suitable classification threshold is dynamically determined based on the mean, adapting to respiratory frequency changes in different populations, achieving intelligent classification of respiratory types, improving the robustness, accuracy, and reliability of the classification method, and demonstrating strong adaptability. It can also provide diagnostic basis for doctors.

[0182] It should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0183] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for classifying respiratory rhythms, characterized in that, include: The control signal acquisition device acquires respiratory signals within the unit sampling period, performs data processing operations on the respiratory signals, and filters out valid respiratory signals; Acquire and segment the effective respiratory signal within a unit sampling period, and determine the corresponding respiratory rate differential sequence; Calculate the average value corresponding to the respiratory rate difference sequence, and calculate the dynamic classification threshold feature corresponding to the respiratory rate difference sequence based on the average value; wherein, the dynamic classification threshold feature is used to represent the fluctuation range and fluctuation amplitude of the respiratory rate within the unit sampling period; Calculate the maximum value and standard deviation corresponding to the respiratory rate difference sequence, and determine the respiratory rhythm type corresponding to the respiratory signal based on the maximum value, standard deviation and dynamic classification threshold feature.

2. The respiratory rhythm classification method according to claim 1, characterized in that, The respiratory rate differential sequence is used to characterize the change in respiratory rate between two adjacent signal segments.

3. The respiratory rhythm classification method according to claim 1, characterized in that, The data processing operation includes at least one of respiratory signal conversion operation and respiratory signal filtering operation; the step of "acquiring and segmenting the effective respiratory signal within a unit sampling period, and determining the corresponding respiratory frequency difference sequence" specifically includes: Based on the respiratory signal segmentation algorithm, a signal segmentation operation is performed on the effective respiratory signal to obtain multiple sets of respiratory signal segments; wherein each set of respiratory signal segments contains the same sampling time length; The respiratory frequency difference sequence is determined by calculating and based on multiple sets of respiratory frequencies corresponding to the multiple sets of respiratory signal segments.

4. The respiratory rhythm classification method according to claim 3, characterized in that, The phrase "performing data processing operations on the respiratory signal to filter and obtain the valid respiratory signal" specifically includes: An analog-to-digital conversion operation is performed on the respiratory signal to obtain a discrete digital sequence corresponding to the respiratory signal; A signal filtering algorithm is used to perform a filtering operation on the discrete digital sequence to obtain effective respiratory signals that meet the first preset respiratory frequency threshold range; wherein, the first preset respiratory frequency threshold range is 0.1Hz-0.5Hz.

5. The respiratory rhythm classification method according to claim 3, characterized in that, The phrase "calculating and determining the respiratory frequency difference sequence based on multiple sets of respiratory frequencies corresponding to the multiple sets of respiratory signal segments" specifically includes: Based on the Fourier transform method, Fourier transform operations are performed on the multiple sets of respiratory signal segments to obtain the corresponding multiple sets of spectral energy distributions; The total spectral energy is obtained by calculating and summing the spectral energies of all respiratory signal segments corresponding to the multiple sets of spectral energy distributions. Determine whether the respiratory rate of each group is within the range of the second preset respiratory rate threshold; wherein the range of the second preset respiratory rate threshold is 0.1Hz-1Hz; If so, a pseudo-difference detection algorithm is used to filter out multiple sets of pseudo-difference respiratory signal segments based on the single-spectrum energy and the total spectral energy of each corresponding set of respiratory signal segments, and to calculate multiple sets of respiratory frequencies corresponding to the multiple sets of pseudo-difference respiratory signal segments to obtain the respiratory frequency difference sequence.

6. The respiratory rhythm classification method according to claim 5, characterized in that, The phrase "selecting multiple sets of pseudo-respiratory signal segments based on the single-spectral energy of each corresponding group of respiratory signal segments and the total spectral energy" specifically includes: Calculate and determine whether the ratio of the single-spectrum energy to the total spectrum energy is greater than a preset ratio threshold. If so, then the respiratory signal segment corresponding to the single-spectral energy is retained from the multiple sets of pseudo-respiratory signal segments; If not, then delete the respiratory signal segment corresponding to the single-spectral energy from the multiple sets of pseudo-respiratory signal segments, and update the multiple sets of pseudo-respiratory signal segments.

7. The respiratory rhythm classification method according to claim 5, characterized in that, The phrase "calculating multiple sets of respiratory frequencies corresponding to the multiple sets of pseudo-respiratory signal segments to obtain the respiratory frequency difference sequence" specifically includes: The respiratory frequencies corresponding to the multiple sets of pseudo-respiratory signal segments are used as sequence elements to construct a respiratory frequency sequence. Calculate the difference between any two adjacent elements in the respiratory rate sequence, and generate the respiratory rate difference sequence based on all the differences.

8. The respiratory rhythm classification method according to claim 1, characterized in that, The phrase "calculating the dynamic classification threshold feature corresponding to the respiratory rate difference sequence based on the average value" specifically includes: Obtain the optimal decision parameters; wherein, the optimal decision parameters include a first optimal decision parameter and a second optimal decision parameter; Based on the first optimal decision parameter and the average value, the first dynamic classification threshold feature is calculated, and based on the first dynamic classification threshold feature and the second optimal decision parameter, the dynamic classification threshold feature is calculated. Wherein, the first dynamic classification threshold feature is equal to the product of the first optimal decision parameter and the average value, and the dynamic classification threshold feature is equal to the sum of the first dynamic classification threshold feature and the second optimal decision parameter.

9. The respiratory rhythm classification method according to claim 8, characterized in that, The "obtaining optimal decision parameters" specifically includes: Construct classification models and decision functions; Statistical features of several respiratory sample data are acquired and calculated. Based on the decision function, the statistical features are input into the classification model to calculate the optimal decision parameters of the classification model. The statistical features include at least the average value of the several respiratory sample data.

10. The respiratory rhythm classification method according to claim 9, characterized in that, The dynamic classification threshold feature includes a dynamic standard deviation threshold; the "construction of classification model and decision function" specifically includes: Construct the first classification model and the first decision function; The phrase "acquiring and calculating statistical features of several respiratory sample data, inputting the statistical features into the classification model based on the decision function, and calculating the optimal decision parameters of the classification model" specifically includes: Acquire and calculate the mean and standard deviation of several respiratory signal sample data, and input the mean and standard deviation of the respiratory signal sample data into the first classification model for training to obtain the first trained classification model; Based on the first trained classification model, the decision boundary when the first decision function equals 0 is calculated, and the first decision function is iteratively optimized based on the decision boundary to obtain the first optimal decision parameters and the second optimal decision parameters.

11. The respiratory rhythm classification method according to claim 9, characterized in that, The dynamic classification threshold feature includes a dynamic maximum threshold; the "construction of classification model and decision function" specifically includes: Construct a second classification model and a second decision function; The phrase "acquiring and calculating statistical features of several respiratory sample data, inputting the statistical features into the classification model based on the decision function, and calculating the optimal decision parameters of the classification model" specifically includes: The average and maximum values ​​of several respiratory signal sample data are obtained and calculated, and the average and maximum values ​​of the respiratory signal sample data are input into the second classification model for training to obtain the second trained classification model. Based on the second trained classification model, the decision boundary when the second decision function equals 0 is calculated, and the second decision function is iteratively optimized based on the decision boundary to obtain the first optimal decision parameter and the second optimal decision parameter.

12. The respiratory rhythm classification method according to claim 1, characterized in that, The breathing rhythm type includes at least one of normal breathing, irregular breathing, and shallow breathing; wherein, normal breathing is defined as a breathing rate of 12-20 breaths per minute with consistent time intervals between exhalation and inhalation; irregular breathing is defined as an unstable breathing rate and depth within a unit sampling period with inconsistent time intervals between exhalation and inhalation; and shallow breathing is defined as a breathing rate exceeding 20 breaths per minute with minimal changes in lung volume.

13. The respiratory rhythm classification method according to claim 1, characterized in that, The dynamic classification threshold feature includes at least one of a dynamic standard deviation threshold and a dynamic maximum value threshold; the step of "determining the respiratory rhythm type corresponding to the respiratory signal based on the maximum value, standard deviation, and dynamic classification threshold feature" specifically includes: Determine whether the maximum value is greater than the dynamic maximum value threshold, and / or whether the standard deviation is greater than the dynamic standard deviation threshold; If not, the respiratory rhythm type corresponding to the respiratory signal is determined to be normal breathing; If so, the corresponding Poincaré map is calculated based on the respiratory frequency difference sequence, and the respiratory rhythm type corresponding to the respiratory signal is determined by extracting and using the Poincaré features of the Poincaré map.

14. The respiratory rhythm classification method according to claim 13, characterized in that, The phrase "calculating the corresponding Poincaré map based on the respiratory rate difference sequence" specifically includes: Obtain and determine several sets of position coordinate points, using two adjacent elements in the respiratory rate difference sequence as the start and end points of the vector; Perform polar coordinate transformation on each position coordinate point to obtain multiple sets of corresponding polar angles and polar radii; wherein, the polar angle is the angle between the corresponding position coordinate point and the positive horizontal axis, and the polar radii is the distance from the corresponding position coordinate point to the origin; Centered on the origin, multiple vector line segments are drawn based on the corresponding polar angles and polar radii to form the Poincaré map.

15. The respiratory rhythm classification method according to claim 13, characterized in that, The Poincaré features include the length features of the vector line segments and the angle features of the vector line segments.

16. The respiratory rhythm classification method according to claim 13, characterized in that, The phrase "extracting and determining the respiratory rhythm type corresponding to the respiratory signal based on the Poincaré features corresponding to the Poincaré map" specifically includes: Obtain and calculate the length and angle of all vector line segments in the Poincaré map to obtain multiple sets of vector length values ​​and multiple sets of vector angle values; Based on the multiple sets of vector length values ​​and the multiple sets of vector angle values, the grid proportion of all vector line segments is calculated to determine the angular characteristics of the vector line segments; wherein, the grid proportion is used to characterize the angular distribution of the vector line segments in the Poincaré map; Based on the multiple sets of vector length values, the proportion of vector line segments that satisfy the preset vector length is calculated to determine the length characteristics of the vector line segments; wherein, the proportion of the number is used to characterize the length distribution of the vector line segments in the Poincaré map; Based on the length and angle characteristics of the vector line segment, the respiratory rhythm type corresponding to the respiratory signal is determined.

17. The respiratory rhythm classification method according to claim 16, characterized in that, The phrase "calculating the grid proportion of all vector line segments based on the multiple sets of vector length values ​​and the multiple sets of vector angle values, and determining the angular characteristics of the vector line segments" specifically includes: Obtain the meshing parameters corresponding to the Poincaré map; wherein, the meshing parameters include the center circle radius, maximum radius, length step size, and angle step size of the Poincaré map; Based on the meshing parameters, a spider meshing operation is performed on the Poincaré map to obtain a spider mesh Poincaré map; Obtain and count the total number of all grids in the spider-web Poincaré map, and the total number of grids occupied by the endpoint of each vector line segment in the spider-web Poincaré map; wherein, the endpoint of the vector line segment does not include the endpoint of the vector located in the central circle of the spider-web Poincaré map; The angular characteristics of the vector line segment are obtained by calculating the ratio of the total number of grids to the total number of all grids.

18. The respiratory rhythm classification method according to claim 16, characterized in that, The phrase "calculating the percentage of vector segments that satisfy a preset vector length based on the multiple sets of vector length values, and determining the length characteristics of the vector segments" specifically includes: Statistically determine the maximum value of the multiple sets of vector length values, and determine the vector length threshold based on the maximum vector length value; The total number of vector length values ​​and the number of vector segments less than the vector length threshold are counted separately to obtain the total number of vectors and the number of vector segments. The length characteristics of the vector segments are obtained by calculating the ratio of the number of vector segments to the total number of vectors.

19. The respiratory rhythm classification method according to claim 16, characterized in that, The phrase "determining the respiratory rhythm type corresponding to the respiratory signal based on the length and angle characteristics of the vector line segment" specifically includes: Determine whether the length characteristic of the vector line segment is less than a preset vector length threshold, and whether the angle characteristic of the vector line segment is less than a preset vector angle threshold; If so, the respiratory rhythm type corresponding to the respiratory signal is determined to be normal breathing; If not, then determine whether the length feature of the vector line segment is greater than or equal to a preset vector length threshold, and whether the angle feature of the vector line segment is greater than or equal to a preset vector angle threshold; If the length characteristic of the vector line segment is greater than or equal to the preset vector length threshold, and the angle characteristic of the vector line segment is greater than or equal to the preset vector angle threshold, then the respiratory rhythm type corresponding to the respiratory signal is determined to be irregular breathing; If the length characteristic of the vector line segment is less than the preset vector length threshold, and the angle characteristic of the vector line segment is less than the preset vector angle threshold, then the respiratory rhythm type corresponding to the respiratory signal is determined to be shallow breathing.

20. A respiratory rhythm classification system, characterized in that, The respiratory rhythm classification system includes: A memory and a processor, the memory having a computer program executable on the processor, the processor executing the program to implement the steps of the respiratory rhythm classification method according to any one of claims 1-19.

Images