A child oral breathing function monitoring and breathing efficiency evaluation system

Abstract

The application discloses a kind of child oral breathing function monitoring and respiratory efficiency evaluation system, it is related to medical care technical field.The system includes: signal splitting module, for responding to the oral-nasal airflow signal of the child to be monitored collected, oral-nasal airflow signal is split into breathing cycle term signal and swallowing residual error term signal;Feature extraction module is used to based on the signal main window of oral-nasal airflow signal and each signal sub-window included in signal main window, the breathing cycle term signal and swallowing residual error term signal are carried out feature extraction, and the first comprehensive feature vector of signal main window is constructed;Type identification module is used to the first comprehensive feature vector is input into pre-trained respiratory identification model, and the breathing cycle type of oral-nasal airflow signal in signal main window is obtained;Efficiency evaluation module is used to based on the breathing cycle type of oral-nasal airflow signal in each signal main window, determines the target oral breathing efficiency of the child to be monitored.The application can accurately evaluate the oral breathing function and respiratory efficiency of child.

Description

Technical Field

[0001] This invention relates to the field of healthcare technology, specifically to a system for monitoring children's oral breathing function and assessing their respiratory efficiency. Background Technology

[0002] Mouth breathing in children is often accompanied by conditions such as sleep apnea. Accurate monitoring and assessment of children's mouth breathing function and respiratory efficiency are of great significance for clinical diagnosis and timely intervention. Sleep monitoring can capture key information such as the frequency of airflow interruptions through the mouth and nose and fluctuations in blood oxygen saturation during sleep, distinguishing between simple snoring and sleep apnea requiring intervention, and providing objective evidence for clinical practice.

[0003] Currently, sleep monitoring in children primarily utilizes polysomnography (PSG) devices, which analyze and assess children's mouth breathing by collecting airflow signals from the mouth and nose. The basic idea is to perform routine processing on the collected airflow signals from the mouth and nose, attempting to extract breathing-related information to evaluate children's mouth breathing function and respiratory efficiency.

[0004] However, swallowing during sleep in children can cause momentary blockages and fluctuations in the airflow signals from the mouth and nose. These changes are highly similar to the airflow instability caused by mouth breathing, and children swallow more frequently during sleep. Existing PSG devices struggle to accurately distinguish between swallowing interference in the airflow signals from actual mouth breathing events, making it impossible to accurately assess children's mouth breathing function and respiratory efficiency. Summary of the Invention

[0005] This invention provides a system for monitoring children's mouth breathing function and assessing their breathing efficiency, which can accurately assess children's mouth breathing function and breathing efficiency.

[0006] A first aspect of the present invention provides a system for monitoring oral breathing function and assessing respiratory efficiency in children, comprising: The signal splitting module is used to split the oral and nasal airflow signals of the child to be monitored into respiratory cycle signal reflecting the respiratory cycle and swallowing residual signal capturing swallowing pulse in response to the acquisition of the oral and nasal airflow signals. The feature extraction module is used to extract features from the respiratory cycle term signal and the swallowing residual term signal based on the main signal window of the oral and nasal airflow signal and the various signal sub-windows included in the main signal window, and to construct the first comprehensive feature vector of the main signal window; the main signal window is adapted to the respiratory cycle of the child to be monitored, and the signal sub-windows are adapted to the swallowing duration of the child to be monitored; The type recognition module is used to input the first comprehensive feature vector of the main signal window into the pre-trained respiratory recognition model to obtain the respiratory cycle type of the oral and nasal airflow signal in the main signal window; the respiratory cycle type includes pure respiratory type and swallowing type. The efficiency assessment module is used to determine the target mouth breathing efficiency of the child to be monitored based on the respiratory cycle type of the oral and nasal airflow signals in each signal main window.

[0007] Furthermore, the present invention also proposes that, before constructing the first comprehensive feature vector of the signal main window based on the nasal and oral airflow signals and the various signal sub-windows included within the signal main window, the system further includes: The sequence construction module is used to perform adaptive peak detection on the respiratory cycle signal to construct the respiratory cycle sequence of the child to be monitored; the respiratory cycle sequence includes the duration of each respiratory cycle of the child to be monitored; The sequence analysis module is used to determine the median, mean, and standard deviation of the respiratory cycle duration of the child under monitoring based on the respiratory cycle sequence of the child under monitoring. The sequence analysis module is also used to determine the coefficient of variation of the respiratory cycle based on the average respiratory cycle duration and the standard deviation of the respiratory cycle duration. The window determination module is used to determine the size of the main window of the signal based on the respiratory cycle variation coefficient and the median respiratory cycle duration.

[0008] Furthermore, the present invention also proposes that, before constructing the first comprehensive feature vector of the signal main window based on the nasal and oral airflow signals and the various signal sub-windows included within the signal main window, the system further includes: The signal analysis module is used to determine the median and standard deviation of swallowing duration for the child under monitoring based on the swallowing pulse signals of the swallowing residual term within a preset time period. The window determination module is also used to determine the sub-window size of the signal sub-window by summing twice the standard deviation of swallowing duration and the median of swallowing duration.

[0009] Furthermore, the present invention also proposes a feature extraction module, comprising: The first feature extraction unit is used to divide the sum of the energy of the respiratory cycle items in the main signal window by the sum of the energy of the respiratory cycle items in the main signal window and the sum of the energy of the swallowing residual items in the main signal window to obtain the energy ratio of the cycle items in the main signal window. The second feature extraction unit is used to perform adaptive peak detection on the respiratory cycle item signal in the main signal window to obtain the completeness of the cycle item in the main signal window. The third feature extraction unit is used to determine whether each signal sub-window included in the main signal window belongs to a strong pulse sub-window, and to obtain the strong pulse continuity of the main signal window; the strong pulse sub-window is a signal sub-window in the swallowing residual signal whose swallowing pulse peak value is greater than N times the stable signal threshold, where N is a positive integer; The fourth feature extraction unit is used to subtract the energy percentage of the periodic items in the previous signal main window from the energy percentage of the periodic items in the main signal window to obtain the energy fluctuation value in the main signal window. The vector construction unit is used to construct the first comprehensive feature vector of the main signal window based on the energy ratio of the periodic term, the completeness of the periodic term, the continuity of the strong pulse, and the energy fluctuation value.

[0010] Furthermore, the present invention also proposes that, before determining whether each signal sub-window included in the main signal window belongs to a strong pulse sub-window and obtaining the strong pulse continuity of the main signal window, the feature extraction module further includes: The respiratory segment screening unit is used to screen out stable respiratory segments from the main signal window based on the energy percentage of the periodic items in the main signal window and the completeness of the periodic items in the main signal window. The threshold determination unit is used to determine the average signal value of the swallowing residual signal within the steady breathing segment as the steady signal threshold.

[0011] Furthermore, the present invention also proposes a type identification module, comprising: The distance determination unit is used to determine the sample Euclidean distance between the signal main window and the labeled training samples based on the first comprehensive feature vector of the signal main window and the second comprehensive feature vector of the labeled training samples. The type identification unit is used to determine the respiratory cycle type of the nasal and oral airflow signal in the main signal window based on the sample Euclidean distance between the main signal window and each labeled training sample, as well as the respiratory cycle type label of each labeled training sample.

[0012] Furthermore, the present invention also proposes a distance determination unit for: The first comprehensive feature vector and the second comprehensive feature vector are standardized respectively to obtain the first standardized feature vector and the second standardized feature vector. The squared feature differences between the first and second standardized feature vectors in each feature dimension are multiplied by the discriminant index of the corresponding feature dimension to obtain the feature difference between the main signal window and the labeled training samples in each feature dimension. Based on the feature differences between the main signal window and the labeled training samples in each feature dimension, the Euclidean distance between the main signal window and the labeled training samples is determined.

[0013] Furthermore, this invention also proposes that, before multiplying the squared feature differences between the first and second standardized feature vectors in each feature dimension by the corresponding discriminant index of the feature dimension to obtain the feature difference between the main signal window and the labeled training samples in each feature dimension, the type recognition module further includes: The information acquisition unit is used to acquire the first mean and first standard deviation of each labeled training sample with a respiratory cycle type label of pure respiratory type label in the target feature dimension, and to acquire the second mean and second standard deviation of each labeled training sample with a respiratory cycle type label containing swallowing type label in the target feature dimension; the target feature dimension can be any feature dimension; The discrimination determination unit is used to determine the discrimination index of the target feature dimension based on the first mean, the first standard deviation, the second mean, and the second standard deviation.

[0014] Furthermore, the present invention also proposes an efficiency evaluation module, comprising: The participation determination unit is used to determine the oral breathing participation of the child to be monitored in the respiratory cycle corresponding to each signal main window based on the respiratory cycle type of the oral and nasal airflow signals in each signal main window. The window filtering unit is used to filter out target signal main windows from each signal main window based on the oral breathing participation of the child under monitoring in the respiratory cycle corresponding to each signal main window. The efficiency assessment unit is used to determine the target mouth breathing efficiency of the child to be monitored based on the window duration percentage of the target signal main window.

[0015] Furthermore, the present invention also proposes a participation degree determination unit, used for: Based on the respiratory cycle type of the oral and nasal airflow signal in the main signal window, obtain the upper and lower energy threshold values ​​corresponding to the respiratory cycle type. Based on the energy percentage of the periodic items, the upper energy threshold, and the lower energy threshold within the main signal window, the oral breathing participation of the child under monitoring within the corresponding respiratory cycle of the main signal window is determined. The energy percentage of the periodic items is used to characterize the sum of the energy of the respiratory cycle items within the main signal window, divided by the sum of the energy of the respiratory cycle items within the main signal window and the sum of the energy of the swallowing residual items.

[0016] The present invention has the following beneficial effects: In the pediatric mouth breathing function monitoring and respiratory efficiency assessment system provided in this invention, the signal decomposition module can decompose the nasal and oral airflow signals into respiratory cycle signals and swallowing residual signals, distinguishing the signals at their source; the feature extraction module extracts features based on a main signal window adapted to the respiratory cycle and a sub-signal window adapted to the swallowing duration, constructing a first comprehensive feature vector, which can more accurately capture signal features; the type recognition module uses a pre-trained respiratory recognition model to identify the respiratory cycle type of the main signal window, effectively distinguishing between pure breathing and swallowing situations; finally, the efficiency assessment module determines the target mouth breathing efficiency based on the respiratory cycle type of each main signal window. Thus, through the synergistic effect of these modules, swallowing interference and actual mouth breathing events in the nasal and oral airflow signals can be accurately distinguished, thereby accurately assessing children's mouth breathing function and respiratory efficiency. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of a system for monitoring oral breathing function and assessing respiratory efficiency in children, provided in one embodiment of the present invention. Figure 2 This is a schematic diagram of an oral and nasal airflow signal provided in one embodiment of the present invention; Figure 3 This is a schematic diagram of a respiratory cycle term signal provided in one embodiment of the present invention; Figure 4 This is a schematic diagram of a swallowing residual signal provided in one embodiment of the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a child mouth breathing function monitoring and respiratory efficiency assessment system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] In traditional pediatric sleep monitoring technologies, the instantaneous blockage and fluctuation of airflow signals from the mouth and nose caused by swallowing behavior are highly similar to the airflow instability characteristics of mouth breathing events. This makes it impossible for existing monitoring systems to effectively distinguish between swallowing interference and real breathing events. The high frequency of swallowing behavior further exacerbates the complexity of signal analysis, resulting in systematic biases in the assessment results of mouth breathing function and respiratory efficiency, which affects the objectivity of clinical judgment.

[0022] For example, during polysomnography of a preschool child, the child frequently swallowed during sleep. The airflow signals from the mouth and nose showed a pattern of instantaneous decline and recovery similar to apnea at the moment of swallowing. The monitoring equipment misidentified this swallowing interference as a respiratory event, resulting in an incorrect judgment of the respiratory cycle type. Consequently, the respiratory efficiency calculation results were distorted and could not accurately reflect the true respiratory state.

[0023] If the above problems are not addressed, the unreliability of monitoring results will make it difficult for clinicians to distinguish between simple snoring and sleep apnea requiring intervention, which may lead to diagnostic delays and increase health risks. In particular, the deviation in respiratory efficiency assessment directly affects the timeliness and accuracy of intervention decisions.

[0024] In this regard, such as Figure 1 As shown in the diagram, this invention provides a structural schematic of a child mouth breathing function monitoring and respiratory efficiency assessment system 100, which includes: The signal splitting module 110 is used to split the oral and nasal airflow signal into a respiratory cycle term signal reflecting the respiratory cycle and a swallowing residual term signal capturing swallowing pulses in response to the acquisition of the oral and nasal airflow signal of the child to be monitored. The feature extraction module 120 is used to extract features from the respiratory cycle term signal and the swallowing residual term signal based on the main signal window of the oral and nasal airflow signal and the various signal sub-windows included in the main signal window, and to construct the first comprehensive feature vector of the main signal window; the main signal window is adapted to the respiratory cycle of the child to be monitored, and the signal sub-windows are adapted to the swallowing duration of the child to be monitored. The type recognition module 130 is used to input the first comprehensive feature vector of the main signal window into the pre-trained respiratory recognition model to obtain the respiratory cycle type of the oral and nasal airflow signal in the main signal window; the respiratory cycle type includes pure respiratory type and swallowing type. The efficiency assessment module 140 is used to determine the target mouth breathing efficiency of the child to be monitored based on the respiratory cycle type of the oral and nasal airflow signals in each signal main window.

[0025] For ease of understanding, the following explains some key terms in this embodiment: The oral and nasal airflow signal refers to the raw signal reflecting changes in airflow through a child's mouth and nose, collected by sensors. This signal contains information about the child's physiological activities such as breathing and swallowing. The respiratory cycle signal is the portion separated from the oral and nasal airflow signal that primarily characterizes the periodic changes in a child's breathing. This signal is used to analyze a child's respiratory characteristics, such as respiratory rate and depth. The swallowing residual signal is the portion separated from the oral and nasal airflow signal that primarily captures the instantaneous pulses generated by a child's swallowing behavior. This signal is used to identify and quantify swallowing events.

[0026] The main signal window refers to a time period set when analyzing airflow signals from the mouth and nose. The length of this time period matches the respiratory cycle of the child being monitored, and is used to capture signal characteristics within a complete respiratory cycle. The sub-signal window refers to a shorter time period further subdivided within the main signal window. The length of this sub-window matches the swallowing duration of the child being monitored, and is used to accurately capture instantaneous events such as swallowing pulses. The first comprehensive feature vector is a multidimensional data structure constructed after feature extraction from the respiratory cycle signal and the swallowing residual signal. This vector contains quantitative features of breathing and swallowing behavior within the main signal window, used for subsequent type identification.

[0027] A breathing recognition model refers to a trained machine learning model or pattern recognition method. This breathing recognition model can receive a first comprehensive feature vector as input and output the corresponding breathing cycle type.

[0028] Respiratory cycle type refers to the classification information of the airflow signals from the mouth and nose within the main signal window. This type is used to distinguish whether the main signal window contains pure respiratory activity or swallowing activity. Pure respiratory type refers to one of the respiratory cycle types, indicating that within the corresponding main signal window, the airflow signals from the mouth and nose mainly represent respiratory activity, and no significant swallowing interference is detected. Swallowing-involved type refers to another of the respiratory cycle types, indicating that within the corresponding main signal window, the airflow signals from the mouth and nose contain both respiratory activity and significant swallowing interference.

[0029] Target mouth breathing efficiency refers to the parameter output by the system that quantifies the degree of mouth breathing in the child being monitored. This parameter reflects the extent or efficiency of the child's mouth breathing participation over a period of time.

[0030] The following is a detailed description of the pediatric mouth breathing function monitoring and respiratory efficiency assessment system 100 of this embodiment. The pediatric mouth breathing function monitoring and respiratory efficiency assessment system 100 includes a signal splitting module 110, a feature extraction module 120, a type recognition module 130, and an efficiency assessment module 140.

[0031] The signal decomposition module 110 is used to process the collected oral and nasal airflow signals of the child to be monitored. Specifically, the signal decomposition module 110 decomposes the original oral and nasal airflow signals into two main parts: a respiratory cycle term signal reflecting the periodic changes in respiration, and a swallowing residual term signal capturing the instantaneous pulses generated by swallowing behavior. As one implementation, the signal decomposition module 110 can use seasonal and trend decomposition using Loess (STL) based on local weighted regression to decompose the original oral and nasal airflow signals into a respiratory cycle term signal and a swallowing residual term signal. For example... Figure 2 The diagram illustrates an oral and nasal airflow signal. STL decomposition utilizes the difference between the regular, repetitive fluctuations of respiration and the sudden impulses of swallowing, allowing the periodic terms of the decomposed oral and nasal airflow signal to correspond to the respiratory signal and the residual terms to correspond to the swallowing impulses. For example... Figure 3 As shown, a schematic diagram of a respiratory cycle term signal is provided; as Figure 4 The diagram shown provides a schematic representation of a swallowing residual signal.

[0032] The feature extraction module 120 is responsible for extracting effective features from the split respiratory cycle signal and swallowing residual signal. Based on the main signal window of the nasal and oral airflow signal and the various sub-windows contained within it, the feature extraction module 120 extracts features from the respiratory cycle signal and swallowing residual signal, and constructs a first comprehensive feature vector for the main signal window. The length of the main signal window is set to match the respiratory cycle of the child being monitored, while the length of each sub-window is set to match the swallowing duration of the child. For example, the feature extraction module 120 can calculate the average amplitude of the respiratory cycle signal within the main signal window, and simultaneously calculate the maximum pulse amplitude, pulse duration, and pulse number of the swallowing residual signal within each sub-window. These calculated values ​​are combined into the first comprehensive feature vector. As another implementation, the feature extraction module 120 can perform Fourier transform on the respiratory cycle term signal to extract its main frequency components and their corresponding energy, and perform time-domain analysis on the swallowing residual term signal to extract its instantaneous power and waveform steepness, thereby constructing the first comprehensive feature vector.

[0033] The type recognition module 130 is used to classify the first comprehensive feature vector output by the feature extraction module 120. The type recognition module 130 inputs the first comprehensive feature vector of the main signal window into a pre-trained respiratory recognition model to obtain the respiratory cycle type of the nasal and oral airflow signal in the main signal window. The respiratory cycle type includes pure breathing and swallowing-related types. Specifically, the respiratory recognition model can be a classifier based on fixed rules. For example, when a certain feature value (such as the number of swallowing pulses) in the first comprehensive feature vector exceeds a preset threshold, the respiratory cycle type is identified as swallowing-related; otherwise, it is identified as pure breathing. Alternatively, the respiratory recognition model can be a pre-trained decision tree model, which performs hierarchical judgments based on multiple feature values ​​in the first comprehensive feature vector and finally outputs the corresponding respiratory cycle type.

[0034] The efficiency assessment module 140 is used to determine the target mouth breathing efficiency for the child being monitored based on the identified respiratory cycle type. This module 140 quantifies the degree of mouth breathing in the child based on the respiratory cycle type of the nasal and oral airflow signals within each signal master window. For example, the module 140 can count the number of signal master windows identified as pure breathing types within a monitoring period and calculate the ratio of this number to the total number of signal master windows to obtain the target mouth breathing efficiency. Alternatively, the module 140 can accumulate the durations of all signal master windows identified as pure breathing types and calculate the ratio of this sum to the total monitoring duration to reflect the proportion of mouth breathing in time, which can then be used as the target mouth breathing efficiency.

[0035] As an example, suppose a child is undergoing nighttime sleep monitoring to assess their mouth breathing function and respiratory efficiency. During monitoring, sensors continuously collect airflow signals from the child's mouth and nose. First, the raw airflow signals are transmitted to a signal decomposition module 110. The raw airflow signals are processed and decomposed into two independent signal streams: a respiratory cycle signal and a swallowing residual signal. For example, when the child is breathing steadily, the respiratory cycle signal will exhibit periodic fluctuations, while the swallowing residual signal remains stable. When the child swallows, the swallowing residual signal will show transient pulses, while the respiratory cycle signal may simultaneously exhibit short-term disturbances. Through this decomposition, respiratory and swallowing activities are initially identified at the signal level.

[0036] Subsequently, the feature extraction module 120 receives the split respiratory cycle signal and swallowing residual signal. The feature extraction module 120 segments the signal according to a preset main signal window and sub-windows. For example, a main signal window may cover approximately 5 seconds of the child's breathing activity, while multiple sub-windows within the main signal window may each cover approximately 0.5 seconds to capture swallowing pulses. Within the main signal window, the feature extraction module 120 calculates features such as the average amplitude and waveform integrity of the respiratory cycle signal, and analyzes features such as the pulse intensity and number of the swallowing residual signal within each sub-window. These quantified features are combined into a first comprehensive feature vector, which describes the breathing and swallowing activity within the current main signal window. For example, if significant swallowing pulses exist within the main window, the feature value reflecting swallowing intensity in the first comprehensive feature vector will be higher.

[0037] Next, the type recognition module 130 receives the first comprehensive feature vector. This module inputs the first comprehensive feature vector into a pre-trained respiratory recognition model. Based on the features in the first comprehensive feature vector, the respiratory recognition model determines the respiratory cycle type of the current signal main window. For example, if the first comprehensive feature vector shows a stable respiratory cycle signal and no significant pulses in the swallowing residual signal, the type recognition module 130 will identify it as a pure breathing type. Conversely, if the first comprehensive feature vector shows a perturbation in the respiratory cycle signal and strong pulses in the swallowing residual signal, the respiratory recognition model will identify it as a swallowing type. In this way, the system can distinguish which respiratory cycles are pure breathing and which are disturbed by swallowing.

[0038] Finally, the efficiency assessment module 140 summarizes the respiratory cycle types of all signal master windows. Based on this type information, the efficiency assessment module 140 determines the target mouth breathing efficiency for the child being monitored. For example, if the system identifies a large number of signal master windows containing swallowing types throughout the nighttime monitoring process, it indicates that the child frequently swallows during sleep. The entire process, through the collaborative work of the modules, achieves effective separation, feature extraction, type identification, and final efficiency assessment of breathing and swallowing events in the nasal and oral airflow signals, solving the problem of inaccurate mouth breathing assessment caused by swallowing interference in existing technologies.

[0039] In this embodiment, the signal splitting module 110 can split the nasal and oral airflow signals into respiratory cycle signals and swallowing residual signals, distinguishing the signals at their source. The feature extraction module 120 extracts features based on a main signal window adapted to the respiratory cycle and a sub-signal window adapted to the swallowing duration, constructing a first comprehensive feature vector that can more accurately capture signal features. The type recognition module 130 uses a pre-trained respiratory recognition model to identify the respiratory cycle type of the main signal window, effectively distinguishing between pure breathing and swallowing situations. Finally, the efficiency evaluation module 140 determines the target mouth breathing efficiency based on the respiratory cycle type of each main signal window. Thus, through the synergistic effect of these modules, swallowing interference and real mouth breathing events in the nasal and oral airflow signals can be accurately distinguished, thereby accurately assessing children's mouth breathing function and respiratory efficiency.

[0040] In some embodiments of the present invention described above, a system 100 for monitoring oral breathing function and assessing respiratory efficiency in children is proposed, which requires feature extraction of airflow signals from the mouth and nose based on a main signal window. However, the size of the main signal window directly affects the accuracy of feature extraction and the subsequent identification of respiratory cycle types. If the size of the main signal window is fixed or unreasonable, it may be impossible to accurately capture the respiratory characteristics of the child being monitored, leading to deviations in the assessment results.

[0041] To address this, the present invention further proposes that, before constructing the first comprehensive feature vector of the main signal window based on the nasal and oral airflow signals and the various signal sub-windows included within the main signal window, the system further includes: The sequence construction module is used to perform adaptive peak detection on the respiratory cycle signal to construct the respiratory cycle sequence of the child to be monitored; the respiratory cycle sequence includes the duration of each respiratory cycle of the child to be monitored; The sequence analysis module is used to determine the median, mean, and standard deviation of the respiratory cycle duration of the child under monitoring based on the respiratory cycle sequence of the child under monitoring. The sequence analysis module is also used to determine the coefficient of variation of the respiratory cycle based on the average respiratory cycle duration and the standard deviation of the respiratory cycle duration. The window determination module is used to determine the size of the main window of the signal based on the respiratory cycle variation coefficient and the median respiratory cycle duration.

[0042] In this embodiment, the sequence construction module accurately identifies the start and end points of each respiratory cycle from continuous respiratory cycle signals, calculates the duration of each respiratory cycle, and constructs a respiratory cycle sequence for the child to be monitored. Adaptive peak detection can adjust detection parameters according to the real-time characteristics of the signal. For example, a threshold-based peak detection algorithm can be used, where the threshold is dynamically adjusted based on the local mean or standard deviation of the signal; or signal processing methods such as wavelet transform can be used to identify periodic components and peaks in the signal through wavelet coefficients at different scales. In this way, the system can obtain a series of discrete respiratory cycle duration data, laying the foundation for subsequent statistical analysis. The respiratory cycle sequence is a time series data, where each element represents the duration of a complete respiratory cycle, and its accuracy directly affects the assessment of respiratory pattern stability.

[0043] The sequence analysis module performs statistical analysis on respiratory cycle sequences to obtain the median, mean, and standard deviation of respiratory cycle duration. The median respiratory cycle duration reflects the typical duration of a respiratory cycle and is less susceptible to extreme values. The mean respiratory cycle duration reflects the overall trend of the respiratory cycle. The standard deviation measures the dispersion of respiratory cycle duration data; a larger standard deviation indicates greater fluctuation in respiratory cycle duration and a more unstable respiratory pattern. These statistics can be obtained using conventional statistical calculation methods. The coefficient of variation (CVO) of the respiratory cycle is calculated by the sequence analysis module based on the mean and standard deviation of respiratory cycle duration. It is the ratio of the standard deviation to the mean, a dimensionless indicator used to measure the relative volatility or instability of respiratory cycle duration. A higher CVO usually indicates higher respiratory pattern instability.

[0044] The window determination module determines the size of the main signal window based on the respiratory cycle variation coefficient and the median respiratory cycle duration. Its function is to dynamically adjust the length of the main signal window according to the current respiratory pattern characteristics of the child being monitored, ensuring that the main signal window better adapts to the actual respiratory cycle. The size of the main signal window can be determined using the following formula: Formula 1 In formula 1, The main window size is used to characterize the main signal window; M is used to characterize the median respiratory cycle duration; and CV is used to characterize the respiratory cycle coefficient of variation. This is used to characterize the floor function. The median respiratory cycle duration is used as the baseline window size, and the main window size is adaptively adjusted according to the child's fluctuations using the respiratory cycle variation coefficient.

[0045] The present invention employs a sequence construction module to adaptively detect peaks in the respiratory cycle signal, thereby accurately identifying each respiratory cycle of the child under monitoring and constructing a respiratory cycle sequence containing the duration of each respiratory cycle. This adaptive detection method effectively addresses individual differences and variations in respiratory patterns, ensuring high accuracy of the acquired respiratory cycle duration data. Subsequently, the sequence analysis module performs in-depth statistical analysis on the respiratory cycle sequence, calculating the median, mean, and standard deviation of the respiratory cycle duration. These statistics characterize the typical features and stability of the child's respiratory pattern from different dimensions. Based on this, the sequence analysis module further calculates the respiratory cycle coefficient of variation, which, as a dimensionless indicator, objectively reflects the relative volatility of the respiratory cycle. Finally, the window determination module dynamically determines the size of the main signal window by comprehensively utilizing the respiratory cycle coefficient of variation and the median respiratory cycle duration. In this way, the size of the main signal window is no longer fixed but can be adaptively adjusted according to the real-time respiratory pattern of the child under monitoring. This adaptive adjustment mechanism ensures that the main signal window can better match the actual respiratory cycle of the child being monitored, thereby providing more accurate and representative signal segments for the subsequent feature extraction module, thus improving the construction quality of the first comprehensive feature vector and ultimately improving the accuracy of respiratory cycle type identification.

[0046] The following is a concrete example to illustrate this. The sequence construction module can use a wavelet transform-based peak detection algorithm to identify peaks in the respiratory cycle signal. For example, continuous wavelet transforms can be performed on the respiratory cycle signal, and appropriate wavelet basis functions (such as Morlet wavelets) and scales can be selected. The location of the respiratory peaks can be determined by analyzing the modulus maxima of the wavelet coefficients, and then the time interval between adjacent peaks can be calculated as the respiratory cycle duration to construct the respiratory cycle sequence. The sequence analysis module can use standard statistical library functions to calculate the statistics of the respiratory cycle sequence. For example, correlation functions can be used to calculate the median, mean, and standard deviation of the respiratory cycle duration. The respiratory cycle coefficient of variation is obtained directly by dividing the standard deviation by the mean. The window determination module finally determines the size of the main window of the signal according to Formula 1 above.

[0047] The above technical solution allows for the dynamic adjustment of the main signal window size based on the actual breathing pattern of the child being monitored. This adaptive window determination method avoids the inaccurate feature extraction problems that may arise from a fixed window size, enabling the main signal window to better match the respiratory cycle characteristics of the child being monitored. This not only improves the accuracy and representativeness of feature extraction from respiratory cycle and swallowing residual signals but also provides higher-quality input data for subsequent respiratory recognition models, thereby significantly improving the accuracy of respiratory cycle type identification. Consequently, the results of monitoring children's oral breathing function and assessing respiratory efficiency are more reliable and accurate.

[0048] In some embodiments of the present invention described above, the pediatric oral breathing function monitoring and respiratory efficiency assessment system 100 extracts features from the respiratory cycle signal and the swallowing residual signal based on a main signal window and sub-signal windows. The sub-signal windows need to be adapted to the swallowing duration of the child being monitored to accurately capture swallowing pulses. However, if the size of the sub-signal windows fails to accurately reflect individual swallowing characteristics, it may lead to inaccurate identification of swallowing pulses, thereby affecting the subsequent determination of respiratory cycle type and assessment of oral breathing efficiency.

[0049] To address this, the present invention further proposes that, before constructing the first comprehensive feature vector of the main signal window based on the nasal and oral airflow signals and the various signal sub-windows included within the main signal window, the system further includes: The signal analysis module is used to determine the median and standard deviation of swallowing duration for the child under monitoring based on the swallowing pulse signals of the swallowing residual term within a preset time period. The window determination module is also used to determine the sub-window size of the signal sub-window by summing twice the standard deviation of swallowing duration and the median of swallowing duration.

[0050] In this embodiment, the signal analysis module is primarily responsible for processing and analyzing the signal to extract key information related to swallowing. Its function is to identify discrete swallowing pulse signals from the complex swallowing residual signal and quantify their temporal characteristics. This module can be implemented using a dedicated digital signal processor or by running specific signal processing algorithms on a general-purpose processor (such as a microcontroller or embedded system). The swallowing residual signal is obtained by decomposing the acquired nasal and oral airflow signals and mainly contains signal components related to swallowing events. It is typically obtained by removing periodic respiratory components using signal decomposition techniques (e.g., wavelet transform, empirical mode decomposition, or adaptive filtering). This signal is characterized by its waveform containing instantaneous, high-energy pulses corresponding to swallowing actions. The preset time period refers to a continuous period used to analyze the swallowing residual signal and extract swallowing features. The length of this time period can be set according to the actual application scenario and data acquisition requirements; for example, it can be set to several minutes to tens of minutes to ensure that a sufficient number of swallowing pulse samples can be collected for statistical analysis. Each swallowing pulse signal refers to an independent, instantaneous signal event representing a single swallowing action, identified from the swallowing residual signal using a specific algorithm (such as threshold detection, peak detection, or pattern recognition). These signals typically possess specific waveform characteristics, such as rapid rising and falling edges, and relatively high amplitude. The median swallowing duration is the intermediate value obtained after statistically analyzing the durations of all swallowing pulse signals detected within a preset time period. As a robust statistic, the median effectively avoids the influence of extreme swallowing durations on the average, more accurately reflecting the typical swallowing duration of the child being monitored. The standard deviation of swallowing duration is a statistic that measures the dispersion of the durations of all swallowing pulse signals detected within a preset time period. A larger standard deviation indicates greater fluctuation in swallowing duration; conversely, a smaller standard deviation indicates relatively stable swallowing duration. It provides quantitative information on the variability of swallowing duration. The window determination module calculates and determines the precise size of the signal sub-window based on the statistical results provided by the signal analysis module. This module can be a software functional unit that receives the median swallowing duration and standard deviation of swallowing duration as input and outputs the calculated sub-window size. The sub-window size of the signal sub-window determines the duration of the signal sub-window used for feature extraction. This size is dynamically calculated based on the median swallowing duration and standard deviation of the child being monitored, aiming to ensure that each signal sub-window can completely and accurately cover a swallowing pulse event, thereby optimizing the extraction effect of swallowing-related features.

[0051] The system of this invention introduces a signal analysis module and a window determination module before feature extraction. Specifically, the signal analysis module first performs in-depth analysis on the swallowing residual signal extracted from the oral and nasal airflow signal. Within a preset time period, the signal analysis module can accurately identify each swallowing pulse signal and, based on these swallowing pulse signals, calculate the median and standard deviation of swallowing duration for the child being monitored. The median swallowing duration reflects the typical swallowing duration of a child, while the standard deviation quantifies the individual differences and variability in swallowing duration. Subsequently, the window determination module uses these statistical parameters to accumulate twice the standard deviation of swallowing duration with the median swallowing duration, thereby dynamically determining the size of the sub-window of the signal sub-window. This calculation method ensures that the size of the signal sub-window not only considers the average swallowing duration of children but also takes into account the variability of swallowing duration, enabling the sub-window to more robustly cover the vast majority of swallowing events. In this way, the feature extraction module can extract features of the swallowing residual signal based on a signal sub-window that is highly adapted to the swallowing characteristics of the child being monitored in subsequent processing, thereby significantly improving the accuracy of swallowing pulse capture and the effectiveness of feature extraction, laying the foundation for accurate identification of subsequent respiratory cycle types.

[0052] The size of the sub-window of the signal sub-window can be determined by the following formula 2: Formula 2 In formula 2, The size of the sub-window used to characterize the signal sub-window. Used to characterize the median swallowing time Used to characterize the standard deviation of swallowing duration Used to characterize the floor operation.

[0053] The standard deviation of swallowing duration being twice the standard deviation aligns with the statistical principle of 2σ. Choosing two standard deviations balances window coverage and window redundancy: ensuring that most swallowing pulse signals are contained within the signal sub-window while preventing the sub-window from becoming too large and causing irrelevant signals to be mixed in. If the standard deviation of swallowing duration is three or four times, the signal sub-window will be excessively enlarged, reducing the specificity of signal feature extraction.

[0054] In one specific implementation, the signal analysis module can be implemented by a software program executed by an embedded processor (e.g., a microcontroller based on the ARM Cortex-M4 architecture). This program continuously receives swallowing residual signals and, within a preset time period of, for example, 10 minutes, identifies swallowing pulse signals by applying an adaptive threshold detection algorithm. Once a pulse is detected, the program measures and stores its duration. After the preset time period ends, the program calculates the median and standard deviation of all recorded swallowing durations. For example, if the calculated median swallowing duration is 0.4 seconds and the standard deviation is 0.08 seconds, the window determination module calculates the sub-window size of the signal sub-window based on these values. Specifically, the sub-window size can be determined to be 0.4 seconds plus twice 0.08 seconds, i.e., 0.56 seconds. This calculated 0.56-second sub-window size is then passed to the feature extraction module to guide it in dividing the signal sub-windows in 0.56-second increments when performing feature extraction on the swallowing residual signals.

[0055] By introducing a signal analysis module and a window determination module, this invention can dynamically determine the median and standard deviation of swallowing duration based on the swallowing residual signal of the child being monitored, and adaptively set the sub-window size of the signal sub-window accordingly. This method avoids the problems of swallowing pulse information truncation or the introduction of irrelevant signals that may occur when using a fixed or empirically set sub-window size, ensuring that the signal sub-window can accurately cover and capture individualized swallowing pulse events. Therefore, in the feature extraction stage, the system can more accurately extract swallowing-related features, significantly improving the accuracy and robustness of swallowing pulse recognition, thus providing a more reliable data foundation for the accurate classification of subsequent respiratory cycle types, and ultimately improving the overall accuracy and reliability of children's oral breathing function monitoring and respiratory efficiency assessment.

[0056] In some embodiments of the present invention described above, feature extraction of nasal and oral airflow signals is proposed to construct a first comprehensive feature vector. However, in its implementation, if the feature extraction method is not comprehensive or accurate enough, it may fail to fully capture the complex dynamics of respiratory periodicity and swallowing events in the nasal and oral airflow signals, thereby affecting the accuracy of subsequent respiratory period type identification.

[0057] In response, the present invention further proposes a feature extraction module 120, comprising: The first feature extraction unit is used to divide the sum of the energy of the respiratory cycle items in the main signal window by the sum of the energy of the respiratory cycle items in the main signal window and the sum of the energy of the swallowing residual items in the main signal window to obtain the energy ratio of the cycle items in the main signal window. The second feature extraction unit is used to perform adaptive peak detection on the respiratory cycle item signal in the main signal window to obtain the completeness of the cycle item in the main signal window. The third feature extraction unit is used to determine whether each signal sub-window included in the main signal window belongs to a strong pulse sub-window, and to obtain the strong pulse continuity of the main signal window; the strong pulse sub-window is a signal sub-window in the swallowing residual signal whose swallowing pulse peak value is greater than N times the stable signal threshold, where N is a positive integer; The fourth feature extraction unit is used to subtract the energy percentage of the periodic items in the previous signal main window from the energy percentage of the periodic items in the main signal window to obtain the energy fluctuation value in the main signal window. The vector construction unit is used to construct the first comprehensive feature vector of the main signal window based on the energy ratio of the periodic term, the completeness of the periodic term, the continuity of the strong pulse, and the energy fluctuation value.

[0058] In this embodiment, the first feature extraction unit is used to calculate the energy proportion of the periodic term, which quantifies the relative contribution of the respiratory periodic term signal to the overall oral and nasal airflow signal energy. Specifically, the sum of the energy of the respiratory periodic term signals and the sum of the energy of the swallowing residual term signals within the main signal window can be calculated, and then the sum of the energy of the respiratory periodic term signals can be divided by the cumulative value of the sum of the two energy values.

[0059] The second feature extraction unit performs adaptive peak detection on the respiratory cycle signal within the main signal window to obtain the cycle completeness. Cycle completeness reflects the regularity and continuity of the respiratory cycle signal. For example, a wavelet transform-based peak detection algorithm can be used to identify valid peaks in the respiratory cycle signal, and the completeness is calculated based on the number of detected peaks. Specifically, if the number of detected peaks is 1, the cycle is complete, and the cycle completeness is recorded as 1; if the number of detected peaks is not 1, the cycle is interrupted by a sudden pulse, and the cycle completeness is recorded as 0.

[0060] The third feature extraction unit is used to determine whether each signal sub-window included in the main signal window belongs to a strong pulse sub-window and to obtain the strong pulse continuity. A strong pulse sub-window refers to a signal sub-window in the swallowing residual signal whose swallowing pulse peak value is greater than N times the stationary signal threshold, where N is a positive integer. For example, N can be 3. The third feature extraction unit can be implemented as follows: First, determine the stationary signal threshold, for example, by analyzing the average energy or standard deviation of the swallowing residual signal. Then, for each signal sub-window in the main signal window, detect the pulse peak value of the swallowing residual signal and compare it with N times the stationary signal threshold. Finally, count the number of strong pulse sub-windows. If two or more consecutive signal sub-windows belong to strong pulse sub-windows, the strong pulse continuity of the main signal window is determined to be 1; if no two or more consecutive signal sub-windows belong to strong pulse sub-windows, the strong pulse continuity of the main signal window is determined to be 0.

[0061] The fourth feature extraction unit is used to calculate the energy fluctuation value, which is the difference between the energy percentage of the periodic items in the current signal main window and the energy percentage of the periodic items in the previous signal main window. This helps to capture dynamic changes in breathing patterns. The fourth feature extraction unit can be implemented by storing the energy percentage of the periodic items in the previous signal main window and subtracting it from the energy percentage of the periodic items in the current signal main window.

[0062] The vector construction unit combines the calculated periodic term energy ratio, periodic term completeness, strong pulse continuity, and energy fluctuation value into a first comprehensive feature vector for the main signal window. This unit can arrange these feature values ​​in a preset order to form a multi-dimensional vector. To eliminate the influence of different dimensions between features and ensure a relatively balanced weight for each feature in the vector, the feature values ​​can be further normalized.

[0063] In practical implementation, the feature extraction module receives the respiratory cycle signal and swallowing residual signal output by the signal splitting module. The first feature extraction unit calculates the energy proportion of the cycle item within the main signal window in parallel, quantifying the proportion of respiratory activity in the total energy. Simultaneously, the second feature extraction unit performs adaptive peak detection on the respiratory cycle signal to assess the integrity and regularity of the respiratory cycle. The third feature extraction unit focuses on the swallowing residual signal, quantifying the intensity and continuity of swallowing events by identifying strong pulse sub-windows. Furthermore, the fourth feature extraction unit captures the dynamic changes in the breathing pattern by comparing the energy proportion of the cycle item between the current and previous main signal windows. Finally, the vector construction unit integrates these multi-dimensional and complementary features into a unified first comprehensive feature vector. This vector comprehensively and precisely characterizes the respiratory and swallowing characteristics of the nasal and oral airflow signals within the current main signal window, providing a solid data foundation for the subsequent type identification module to accurately distinguish between pure respiratory types and types involving swallowing.

[0064] Through the above technical solution, the feature extraction module can extract multi-dimensional and discriminative features from the nasal and oral airflow signals, including the energy proportion of respiratory activity, the integrity of the respiratory cycle, the intensity and continuity of swallowing events, and the dynamic changes in breathing patterns. These features comprehensively reflect the complex characteristics of the nasal and oral airflow signals, enabling subsequent respiratory recognition models to more accurately distinguish between pure breathing types and types involving swallowing, thereby significantly improving the accuracy and reliability of the pediatric oral breathing function monitoring and respiratory efficiency assessment system.

[0065] In some embodiments of the present invention described above, it is proposed to analyze the swallowing residual signal within the main signal window to determine the strong pulse sub-window. However, in its implementation, the determination of the strong pulse sub-window depends on a stable signal threshold. If this stable signal threshold is not accurately determined, it may lead to misjudgment or omission of swallowing pulses, thereby affecting the accuracy of the strong pulse continuity and ultimately affecting the accuracy of respiratory cycle type identification.

[0066] In response, the present invention further proposes that, before determining whether each signal sub-window included in the main signal window belongs to a strong pulse sub-window and obtaining the strong pulse continuity of the main signal window, the feature extraction module 120 further includes: The respiratory segment screening unit is used to screen out stable respiratory segments from the main signal window based on the energy percentage of the periodic items in the main signal window and the completeness of the periodic items in the main signal window. The threshold determination unit is used to determine the average signal value of the swallowing residual signal within the steady breathing segment as the steady signal threshold.

[0067] In this embodiment, the respiratory segment screening unit identifies and extracts signal segments representing the stable breathing state of the child being monitored from the complex oral and nasal airflow signals. This is because the swallowing residual signal used to determine the stable signal threshold should be the baseline signal without swallowing interference. The unit can screen based on the energy percentage and completeness of the periodic items within the main signal window. For example, when the energy percentage of the periodic items is higher than a preset threshold (indicating that the respiratory signal is dominant) and the completeness of the periodic items is also 1 (indicating that the respiratory cycle is complete and regular), the signal segment can be identified as a stable breathing segment. Alternatively, the respiratory segment screening unit can employ a machine learning model, trained on a large amount of labeled data, to learn and recognize the characteristic patterns of stable breathing segments, thereby achieving more intelligent screening.

[0068] The threshold determination unit calculates a stable signal threshold that accurately reflects the characteristics of the swallowing residual signal in the non-swallowing state, based on the stable breathing segments identified by the breathing segment screening unit. This threshold determination unit determines the average value of the swallowing residual signal within the stable breathing segment as the stable signal threshold. For example, after acquiring all stable breathing segments, the swallowing residual signal values ​​within these segments are summed and divided by the total number of samples to obtain the average value, which is then used as the stable signal threshold.

[0069] The present invention optimizes the determination process of the stable signal threshold by introducing a respiratory segment screening unit and a threshold determination unit. Specifically, the feature extraction module first obtains the energy percentage and completeness of the periodic terms within the main signal window through a first feature extraction unit and a second feature extraction unit. Subsequently, the respiratory segment screening unit uses these indicators to accurately identify the signal segments in the main signal window that represent the stable breathing state of the child to be monitored, thereby avoiding interference from swallowing activity on the threshold calculation. Based on this, the threshold determination unit calculates the average signal value based solely on the swallowing residual signal within these purely stable breathing segments, thus obtaining a highly accurate stable signal threshold that is unaffected by swallowing pulses. This precisely determined stable signal threshold is then used by the third feature extraction unit to determine whether each signal sub-window within the main signal window belongs to a strong pulse sub-window, thereby ensuring that the calculation of strong pulse continuity is more reliable. This mechanism ensures that the subsequent respiratory recognition model can receive more accurate feature vectors, thereby improving the accuracy of the entire system in monitoring children's mouth breathing function and assessing respiratory efficiency.

[0070] The following is a concrete example. Assume that within a main signal window, the energy percentage and integrity of the periodic term have been calculated by the first and second feature extraction units. The respiratory segment screening unit can set a threshold for the energy percentage of the periodic term, for example, 0.8, and a threshold for the integrity of the periodic term, for example, 1. When the energy percentage of the periodic term is consistently higher than 0.8 and the integrity of the periodic term is consistently not lower than 1 for a continuous time period within the main signal window, this time period is screened as a stable respiratory segment. For example, the period from the 5th to the 15th second of the main signal window is identified as a stable respiratory segment. Subsequently, the threshold determination unit extracts all values ​​of the swallowing residual signal corresponding to this stable respiratory segment (i.e., from the 5th to the 15th second) and calculates the arithmetic mean of these values. Assuming the calculated average is 0.05mV, this 0.05mV is determined as the stable signal threshold. This threshold is then used in the third feature extraction unit to determine whether the peak value of the swallowing pulse in the swallowing residual signal is greater than N times 0.05mV, in order to identify strong pulse sub-windows.

[0071] The above technical solution effectively avoids interference from swallowing pulses on the calculation of the steady-state signal threshold during swallowing activities, thus ensuring that the determined steady-state signal threshold accurately reflects the baseline level in the non-swallowing state. This makes the identification of the strong pulse sub-window more accurate, thereby improving the accuracy of strong pulse continuity and ultimately enhancing the reliability of the pediatric oral breathing function monitoring and respiratory efficiency assessment system in identifying respiratory cycle types.

[0072] In some embodiments of the present invention described above, a first comprehensive feature vector is input into a pre-trained respiratory recognition model to obtain the respiratory cycle type of the nasal and oral airflow signals within the main signal window. However, in practical applications, how to specifically implement this respiratory recognition model and ensure its accurate classification of complex nasal and oral airflow signals are technical issues that require further clarification and optimization.

[0073] In response, the present invention further proposes a type identification module 130, comprising: The distance determination unit is used to determine the sample Euclidean distance between the signal main window and the labeled training samples based on the first comprehensive feature vector of the signal main window and the second comprehensive feature vector of the labeled training samples. The type identification unit is used to determine the respiratory cycle type of the nasal and oral airflow signal in the main signal window based on the sample Euclidean distance between the main signal window and each labeled training sample, as well as the respiratory cycle type label of each labeled training sample.

[0074] In this embodiment, the core function of the distance determination unit is to quantify the similarity between the first comprehensive feature vector of the nasal airflow signal to be identified and the second comprehensive feature vector of the nasal airflow signal of a known type. This similarity quantification is usually achieved by calculating their distance in a multi-dimensional feature space. Sample Euclidean distance is a commonly used distance metric, which calculates the square root of the sum of the squares of the differences between two vectors in each feature dimension, and can intuitively reflect the geometric distance between two samples. The second comprehensive feature vector of the labeled training samples is a feature representation of samples with known respiratory cycle type labels that have been pre-annotated manually or by experts. These second comprehensive feature vectors of labeled training samples are constructed in the same way as the first comprehensive feature vector, but they are stored in the system as known reference points to guide the classification of unknown samples. The type identification unit is responsible for making the final classification decision based on the sample Euclidean distance calculated by the distance determination unit and the respiratory cycle type labels of the labeled training samples. Its role is to classify the main window of the signal to be identified into the most similar respiratory cycle type based on the distance relationship between samples. Type identification units can employ various classification algorithms, such as the K-Nearest Neighbors (KNN) algorithm, which finds the K nearest labeled training samples to the sample to be identified and determines the type of the sample by voting or weighted averaging based on the type labels of these nearest samples. Alternatively, distance-based clustering algorithms or Support Vector Machines (SVMs) can also be used for classification.

[0075] The present invention provides a concrete and operable implementation mechanism for the type recognition module by introducing a distance determination unit and a type recognition unit. After the feature extraction module generates the first comprehensive feature vector of the main signal window, this first comprehensive feature vector is sent to the distance determination unit. The distance determination unit compares it with the second comprehensive feature vector of pre-stored labeled training samples with known respiratory cycle type labels, and calculates the sample Euclidean distance between the current main signal window and each labeled training sample. These sample Euclidean distances are then passed to the type recognition unit. Based on this distance information and the respiratory cycle type labels of the labeled training samples, the type recognition unit uses specific classification logic (e.g., the K-nearest neighbor algorithm) to determine whether the nasal and oral airflow signal corresponding to the current main signal window belongs to the pure breathing type or the swallowing type. This distance-based recognition method allows the type recognition module to explicitly classify based on the similarity between feature vectors, thereby overcoming the black box problem that may exist in traditional "pre-trained respiratory recognition models" and improving the transparency and interpretability of the recognition process. In this way, the system can more accurately identify whether a swallowing event is included in the respiratory cycle, providing reliable basic data for subsequent oral breathing efficiency evaluation.

[0076] Through the above technical solution, this invention clarifies the specific implementation method of the type recognition module, namely, calculating the Euclidean distance of samples through a distance determination unit, and then classifying the samples based on these distances and the labels of the labeled training samples by the type recognition unit. This distance-based recognition method provides a concrete and quantifiable implementation path for the breathing recognition model, effectively solving the problem of unclear recognition mechanisms that may exist in traditional models. This enables the system to more accurately and reliably identify pure breathing types and swallowing types in the oral and nasal airflow signals, thereby providing more accurate and reliable input data for subsequent oral breathing efficiency evaluation, significantly improving the performance of the entire monitoring and evaluation system.

[0077] In some embodiments of the present invention described above, the type recognition module determines the sample Euclidean distance between the main signal window and the labeled training samples based on the first comprehensive feature vector of the main signal window and the second comprehensive feature vector of the labeled training samples by the distance determination unit, in order to identify the respiratory cycle type. However, when directly calculating the Euclidean distance, different feature dimensions may have different dimensions and numerical ranges, causing some features to dominate the distance calculation, while other features with important discriminative power may be weakened. Furthermore, not all feature dimensions contribute equally to distinguishing different respiratory cycle types; failure to differentiate them may affect the accuracy of the identification.

[0078] In response, the present invention further proposes a distance determination unit, used for: The first comprehensive feature vector and the second comprehensive feature vector are standardized respectively to obtain the first standardized feature vector and the second standardized feature vector. The squared feature differences between the first and second standardized feature vectors in each feature dimension are multiplied by the discriminant index of the corresponding feature dimension to obtain the feature difference between the main signal window and the labeled training samples in each feature dimension. Based on the feature differences between the main signal window and the labeled training samples in each feature dimension, the Euclidean distance between the main signal window and the labeled training samples is determined.

[0079] In this embodiment, standardization is a data preprocessing technique designed to eliminate differences in dimensions and numerical ranges between different feature dimensions. Standardization transforms feature data to a uniform scale; for example, through Z-score standardization or Min-Max standardization, all features are given equal importance in distance calculation, avoiding unfair influences from features with large numerical ranges on the distance calculation results. The first and second standardized feature vectors are the standardized feature vectors, representing the feature representations of the main signal window and labeled training samples at a uniform scale, respectively. The squared feature difference is calculated by squared the difference between the first and second standardized feature vectors for each corresponding feature dimension. This operation quantifies the degree of difference between the two vectors in each dimension and amplifies larger differences while ensuring that the difference value is non-negative. The discriminant index is a weight assigned to each feature dimension to measure its importance or effectiveness in distinguishing different respiratory cycle types. For example, a feature dimension that effectively distinguishes between pure breathing and swallowing-related types will have a high discriminant index. Discrimination indices can be obtained through statistical analysis or machine learning methods, or determined based on the statistical distribution differences of different types of samples along that feature dimension. Feature discrimination is a measure combining the squared feature differences and the discrimination index; it represents the weighted difference between the main signal window and the labeled training samples along a specific feature dimension. By multiplying the squared feature differences by the corresponding discrimination index, the feature dimensions that contribute more to classification are given greater weight in the overall distance calculation. Sample Euclidean distance is calculated based on the feature discrimination of each feature dimension, comprehensively reflecting the overall similarity or difference between the main signal window and the labeled training samples. This weighted Euclidean distance more accurately measures the "distance" between two samples, thereby improving classification accuracy.

[0080] The present invention effectively eliminates the influence of differences in dimensions and numerical ranges among different feature dimensions by standardizing the first and second comprehensive feature vectors, ensuring fairness in distance calculation. Furthermore, a discrimination index is introduced to weight the squared differences of each feature dimension, allowing feature dimensions with stronger discriminative power for respiratory cycle type identification to play a greater role in distance calculation. In this way, the calculated Euclidean distance between samples more accurately reflects the true similarity or difference between the main signal window and the labeled training samples. This standardized and weighted distance calculation method enables the type identification module to more accurately distinguish between pure breathing types and types involving swallowing, thereby significantly improving the classification performance of the breathing recognition model and providing more reliable basic data for subsequent oral breathing efficiency evaluation.

[0081] The Euclidean distance between the main signal window and the labeled training samples can be determined using the following formula 3: Formula 3 In formula 3, Used to characterize the Euclidean distance between the main signal window and the labeled training samples. The discriminant index used to characterize the j-th feature dimension The eigenvalues ​​of the first normalized eigenvector of the main window representing the signal, in the j-th feature dimension. This is used to characterize the eigenvalues ​​of the second standardized eigenvector in the j-th eigendimensional. There are four eigendimensional dimensions: the energy proportion of the periodic term, the completeness of the periodic term, the continuity of the strong pulse, and the energy fluctuation value.

[0082] Through the above technical solution, this invention effectively solves the problem of inaccurate distance calculation caused by differences in feature dimensions and discriminative abilities. By standardizing the feature vectors, the inconsistency in the numerical range of different feature dimensions is eliminated, ensuring the fairness of distance calculation. Simultaneously, the introduction of a discriminative index to weight each feature dimension ensures that feature dimensions contributing more to respiratory cycle type identification play a more significant role in distance calculation, thereby improving the accuracy and robustness of distance calculation. This enables the type recognition module to more accurately identify the respiratory cycle type of the nasal and oral airflow signals, thus providing more reliable and accurate recognition results for pediatric mouth breathing function monitoring and respiratory efficiency assessment systems.

[0083] In some embodiments of the present invention described above, the type recognition module standardizes the first and second comprehensive feature vectors and calculates the squared differences in each feature dimension to determine the Euclidean distance between the main signal window and the labeled training samples. However, in practical applications, the importance of different feature dimensions in distinguishing between pure breathing types and types involving swallowing may differ. Simply summing the squared differences of all feature dimensions may not fully reflect the true distinguishing ability of each feature dimension, thus affecting the accuracy of respiratory cycle type recognition.

[0084] To address this, the present invention further proposes that, before multiplying the squared feature differences between the first and second standardized feature vectors in each feature dimension by the corresponding discriminant index of the feature dimension to obtain the feature difference between the main signal window and the labeled training samples in each feature dimension, the type recognition module 130 also includes: The information acquisition unit is used to acquire the first mean and first standard deviation of each labeled training sample with a respiratory cycle type label of pure respiratory type label in the target feature dimension, and to acquire the second mean and second standard deviation of each labeled training sample with a respiratory cycle type label containing swallowing type label in the target feature dimension; the target feature dimension can be any feature dimension; The discrimination determination unit is used to determine the discrimination index of the target feature dimension based on the first mean, the first standard deviation, the second mean, and the second standard deviation.

[0085] In this embodiment, the information acquisition unit performs statistical analysis on the labeled training samples used by the pre-trained respiratory recognition model to quantify the distribution characteristics of different respiratory cycle types across various feature dimensions. The information acquisition unit can traverse the stored labeled training sample dataset and, for each feature dimension, calculate the numerical distribution (such as the first mean and the first standard deviation) of pure respiratory type samples in that dimension, as well as the numerical distribution (such as the second mean and the second standard deviation) of samples containing swallowing type samples in that dimension.

[0086] The role of the discrimination determination unit is to assess the importance or contribution of each feature dimension in distinguishing between pure breathing types and types involving swallowing, based on the statistical information provided by the information acquisition unit. This unit can employ various statistical or machine learning methods to calculate the discrimination index. For example, based on the idea of ​​Fisher discriminant analysis, the discrimination index can be determined by calculating the ratio of the inter-class dispersion to the intra-class dispersion of two classes of samples on a certain feature dimension. The higher the ratio, the stronger the discrimination ability of that feature dimension in distinguishing between the two classes of samples. Alternatively, information theory metrics such as information gain can be used to evaluate the amount of information in a feature dimension when distinguishing different respiratory cycle types, thereby determining the discrimination index.

[0087] Furthermore, the discriminative index of the target feature dimension can be determined using the following formula 4: Formula 4 In formula 4, The discriminant index used to characterize the j-th feature dimension Used to characterize the second average value of the swallowing type label in the j-th feature dimension. Used to characterize the first average value of the pure respiratory type label in the j-th feature dimension. Used to characterize the second standard deviation of the swallowing type label in the j-th feature dimension. The first standard deviation of the pure breathing type label in the j-th feature dimension is used to characterize the first standard deviation, and the sigmoid is used to characterize the nonlinear activation function operation.

[0088] It should be noted that, in order to ensure that the calculation results are meaningful, when performing fractional operations in this embodiment of the invention, if the denominator is 0, a parameter adjustment factor greater than 0 needs to be added to the denominator before summing to prevent the denominator from being 0. The value of the parameter adjustment factor can be 0.001. In specific applications, it can also be set by the implementer according to the actual situation.

[0089] The present invention introduces an information acquisition unit and a discrimination determination unit, enabling the calculation of the Euclidean distance between the main signal window and the labeled training samples to fully consider the differences in the contribution of different feature dimensions to the identification of respiratory cycle types. Specifically, the information acquisition unit first performs statistical analysis on a large number of labeled training samples to obtain the first mean, first standard deviation, second mean, and second standard deviation for each feature dimension for both pure respiratory and swallowing types. These statistics reflect the distribution characteristics of different respiratory types on specific feature dimensions. Subsequently, the discrimination determination unit uses these statistics to calculate the discrimination index for each feature dimension. This discrimination index quantifies the effectiveness of each feature dimension in distinguishing between the two respiratory types. In the subsequent distance calculation process, the squared differences between the first and second standardized feature vectors in each feature dimension are multiplied by the corresponding discrimination index. This means that feature dimensions that are more important for distinguishing respiratory types will have higher weights in the Euclidean distance calculation, thus having a greater impact on the final distance result. Conversely, feature dimensions with lower discrimination will have a relatively smaller impact on their differences. In this way, the calculation of the Euclidean distance between samples is no longer a simple summation of the differences in each dimension, but a weighted average based on the actual distinguishing power of each dimension. This makes the calculated distance more accurately reflect the true similarity or difference between the two samples in terms of respiratory cycle type, thereby significantly improving the accuracy and robustness of respiratory cycle type identification.

[0090] Through the above technical solution, this invention can dynamically adjust the weight of different feature dimensions in distance calculation based on their actual distinguishing ability for respiratory cycle type identification. This avoids the problem of decreased identification accuracy caused by treating all feature dimensions equally, allowing the calculated Euclidean distance of the samples to more accurately reflect the true similarity between the respiratory cycle type of the child being monitored and known samples. Therefore, this invention significantly improves the accuracy and robustness of respiratory cycle type identification, providing more reliable basic data for subsequent mouth breathing efficiency assessment, thereby enhancing the overall system's assessment accuracy and clinical application value.

[0091] In some embodiments of the present invention described above, it is proposed to assess children's mouth breathing function by identifying the respiratory cycle type of the nasal and oral airflow signals in each signal main window. However, relying solely on the identification of respiratory cycle type (such as pure breathing or swallowing type) makes it difficult to directly quantify the actual degree of mouth breathing participation of the child under monitoring within a specific time period and its impact on overall respiratory efficiency, which may result in an insufficiently refined and accurate assessment of mouth breathing efficiency.

[0092] In response, the present invention further proposes an efficiency evaluation module 140, comprising: The participation determination unit is used to determine the oral breathing participation of the child to be monitored in the respiratory cycle corresponding to each signal main window based on the respiratory cycle type of the oral and nasal airflow signals in each signal main window. The window filtering unit is used to filter out target signal main windows from each signal main window based on the oral breathing participation of the child under monitoring in the respiratory cycle corresponding to each signal main window. The efficiency assessment unit is used to determine the target mouth breathing efficiency of the child to be monitored based on the window duration percentage of the target signal main window.

[0093] In this embodiment, the participation determination unit aims to transform the qualitative respiratory cycle type identification results into a quantitative indicator of mouth breathing participation. This can be achieved, but is not limited to: analyzing the relative energy, waveform characteristics, or frequency components of the respiratory cycle signal and the swallowing residual signal within the main signal window, combined with the respiratory cycle type, to calculate the contribution ratio of mouth breathing; or, using a pre-trained machine learning model, inputting the first comprehensive feature vector of the main signal window and the respiratory cycle type, and outputting a numerical value representing the degree of mouth breathing participation. The window filtering unit focuses on respiratory cycles with significant mouth breathing behavior, excluding cycles with low or no mouth breathing participation, thereby improving the targeting and accuracy of subsequent efficiency assessments. The preset participation threshold can be a fixed value or dynamically adjusted based on the individual characteristics, clinical needs, or historical data of the child being monitored. The efficiency assessment unit summarizes the information of the selected target signal main windows and calculates their proportion in the total monitoring time, thus providing an intuitive and quantitative mouth breathing efficiency indicator. This indicator reflects the proportion of time during which mouth breathing behavior is significant for the child being monitored throughout the entire monitoring period.

[0094] The present invention first employs a participation determination unit to conduct in-depth analysis of the respiratory cycle type of the nasal and oropharyngeal airflow signals within each signal master window, and combines this with other features within the signal master window to quantify the mouth breathing participation of the child under monitoring in each respiratory cycle. This quantification process elevates the assessment of mouth breathing behavior from simple type identification to a more refined measurement. Subsequently, a window screening unit uses a preset participation threshold to precisely filter out target signal master windows with high mouth breathing participation from all signal master windows, thereby focusing the assessment on the periods when mouth breathing problems actually exist. Finally, an efficiency assessment unit calculates the target mouth breathing efficiency of the child under monitoring based on the window duration percentage of these target signal master windows. The entire process closely integrates the type identification results of the nasal and oropharyngeal airflow signals with the actual level of mouth breathing participation, forming an assessment chain from qualitative to quantitative and from local to overall, making the monitoring and efficiency assessment of children's mouth breathing function more comprehensive, accurate, and clinically instructive.

[0095] By introducing a quantitative assessment of mouth breathing participation and using this to screen for signal windows with significant mouth breathing behavior, this invention overcomes the limitation of accurately quantifying the degree of mouth breathing based solely on respiratory cycle type identification. This approach enables the system to more precisely capture and evaluate the actual duration and intensity of mouth breathing in the monitored children, thus providing a more clinically significant target mouth breathing efficiency index. This provides more accurate and reliable data support for the early detection, intervention, and treatment effectiveness evaluation of mouth breathing problems in children.

[0096] In some embodiments of the present invention described above, the participation determination unit needs to determine the oral breathing participation of the child under monitoring within the corresponding respiratory cycle based on the respiratory cycle type of the oral and nasal airflow signals in the main signal window. However, this may not accurately reflect the energy distribution characteristics of the oral and nasal airflow signals under different respiratory cycle types (e.g., pure breathing type and swallowing type), resulting in insufficient accuracy and robustness of oral breathing participation assessment.

[0097] In response, the present invention further proposes a participation degree determination unit, used for: Based on the respiratory cycle type of the oral and nasal airflow signal in the main signal window, obtain the upper and lower energy threshold values ​​corresponding to the respiratory cycle type. Based on the energy percentage of the periodic items, the upper energy threshold, and the lower energy threshold within the main signal window, the oral breathing participation of the child under monitoring within the corresponding respiratory cycle of the main signal window is determined. The energy percentage of the periodic items is used to characterize the sum of the energy of the respiratory cycle items within the main signal window, divided by the sum of the energy of the respiratory cycle items within the main signal window and the sum of the energy of the swallowing residual items.

[0098] In this embodiment, when determining the oral breathing participation of the child to be monitored, a set of matching upper and lower energy thresholds is first obtained based on the respiratory cycle type of the nasal and oral airflow signals in the current main signal window. These thresholds are pre-set or trained to distinguish the energy distribution characteristics under different breathing types. For example, a lookup table or database can be pre-established, storing the upper and lower energy thresholds corresponding to different respiratory cycle types (such as pure breathing and swallowing-involved types). Once the type identification module identifies the respiratory cycle type in the current main signal window, the participation determination unit can retrieve the corresponding threshold value from the lookup table or database based on that type. Alternatively, this can be achieved by configuring a rule engine to trigger different rule sets to calculate or select the corresponding energy threshold value based on the identified respiratory cycle type.

[0099] Subsequently, the participation determination unit will utilize the energy percentage of the periodic items within the main signal window, combined with the aforementioned upper and lower energy thresholds, to accurately calculate the child's oral breathing participation within the respiratory cycle corresponding to the current main signal window. The energy percentage of the periodic items is the ratio of the respiratory cycle item signal energy to the total energy (the sum of the respiratory cycle item signal energy and the swallowing residual item signal energy), quantifying the relative intensity of respiratory components within the overall nasal and oral airflow signal. By comparing this energy percentage of the periodic items with an adaptive threshold set for a specific breathing type, the degree of oral breathing can be determined more accurately.

[0100] Specifically, the degree of mouth breathing participation of the child to be monitored during the respiratory cycle corresponding to the main signal window can be determined using the following formula 5: Formula 5 In formula 5, Used to characterize mouth breathing involvement The upper energy threshold value used to characterize the type of respiratory cycle. The lower energy threshold value used to characterize the type of respiratory cycle. Used to characterize the energy percentage of periodic terms within the main window of the signal.

[0101] The upper and lower energy thresholds can be calculated as follows: sort the energy percentages of the cycle terms in the labeled training samples of the two breathing cycle types in ascending order, then calculate the energy percentage of the cycle terms corresponding to the 25th percentile as the lower energy threshold, and calculate the energy percentage of the cycle terms corresponding to the 75th percentile as the upper energy threshold.

[0102] The present invention identifies the respiratory cycle type of the main signal window through a type recognition module. The participation determination unit no longer uses a single, fixed energy threshold value, but dynamically acquires and applies matching upper and lower energy threshold values ​​based on the identified respiratory cycle type. Simultaneously, the energy percentage of the cycle item provided by the feature extraction module accurately reflects the relative intensity of respiratory components in the oral and nasal airflow signal. The participation determination unit compares this energy percentage of the cycle item with an adaptive energy threshold value, thereby enabling a more refined and accurate quantification of the child's oral breathing participation in the current respiratory cycle. This mechanism allows the assessment of oral breathing participation to fully consider the presence of swallowing events during respiration, avoiding misjudgments caused by the influence of swallowing activities on energy distribution, thus improving the accuracy and reliability of oral breathing monitoring.

[0103] Through the above technical solution, the system can dynamically adjust the energy threshold for assessing mouth breathing participation based on the respiratory cycle type of the nasal and oral airflow signals, thereby significantly improving the accuracy and robustness of mouth breathing participation assessment. This adaptive assessment mechanism avoids misjudgments that may occur with fixed thresholds under different respiratory situations, making the monitoring of children's mouth breathing function and the assessment of respiratory efficiency more accurate and reliable, and providing more valuable data for clinical diagnosis and intervention.

[0104] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0105] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A system for monitoring oral breathing function and evaluating respiratory efficiency in children, characterized in that, The system includes: The signal splitting module is used to split the oral and nasal airflow signal of the child to be monitored into a respiratory cycle term signal reflecting the respiratory cycle and a swallowing residual term signal capturing swallowing pulses in response to the acquisition of the oral and nasal airflow signal. The feature extraction module is used to extract features from the respiratory cycle signal and the swallowing residual signal based on the main signal window of the oral and nasal airflow signal and the signal sub-windows included in the main signal window, and to construct a first comprehensive feature vector of the main signal window; the main signal window is adapted to the respiratory cycle of the child to be monitored, and the signal sub-windows are adapted to the swallowing duration of the child to be monitored; The type recognition module is used to input the first comprehensive feature vector of the main signal window into a pre-trained respiratory recognition model to obtain the respiratory cycle type of the oral and nasal airflow signal in the main signal window; the respiratory cycle type includes pure respiratory type and swallowing type. The efficiency evaluation module is used to determine the target mouth breathing efficiency of the child to be monitored based on the respiratory cycle type of the oral and nasal airflow signals in each of the main signal windows.

2. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 1, characterized in that, Before constructing the first comprehensive feature vector of the main signal window based on the nasal and oral airflow signals and the various signal sub-windows included in the main signal window, the system further includes: The sequence construction module is used to perform adaptive peak detection on the respiratory cycle item signal to construct the respiratory cycle sequence of the child to be monitored; the respiratory cycle sequence includes the duration of each respiratory cycle of the child to be monitored; The sequence analysis module is used to determine the median respiratory cycle duration, the mean respiratory cycle duration, and the standard deviation of the respiratory cycle duration of the child under monitoring based on the respiratory cycle sequence of the child under monitoring. The sequence analysis module is also used to determine the coefficient of variation of the respiratory cycle based on the average duration of the respiratory cycle and the standard deviation of the duration of the respiratory cycle; The window determination module is used to determine the main window size of the signal main window based on the respiratory cycle variation coefficient and the median respiratory cycle duration.

3. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 1, characterized in that, Before constructing the first comprehensive feature vector of the main signal window based on the nasal and oral airflow signals and the various signal sub-windows included in the main signal window, the system further includes: The signal analysis module is used to determine the median and standard deviation of the swallowing duration of the child to be monitored based on the swallowing pulse signals of the swallowing residual term signal within a preset time period. The window determination module is also used to determine the sub-window size of the signal sub-window by summing twice the standard deviation of the swallowing duration and the median of the swallowing duration.

4. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 1, characterized in that, The feature extraction module includes: The first feature extraction unit is used to divide the sum of the energy of the respiratory cycle item signals in the main signal window by the sum of the energy of the respiratory cycle item signals and the sum of the energy of the swallowing residual item signals in the main signal window to obtain the energy ratio of the cycle item in the main signal window. The second feature extraction unit is used to perform adaptive peak detection on the respiratory cycle item signal in the main signal window to obtain the completeness of the cycle item in the main signal window. The third feature extraction unit is used to determine whether each of the signal sub-windows included in the main signal window belongs to a strong pulse sub-window, and to obtain the strong pulse continuity of the main signal window; the strong pulse sub-window is the signal sub-window in the swallowing residual signal where the swallowing pulse peak value is greater than N times the stable signal threshold, where N is a positive integer; The fourth feature extraction unit is used to subtract the energy percentage of the periodic item in the previous main signal window from the energy percentage of the periodic item in the main signal window to obtain the energy fluctuation value in the main signal window. The vector construction unit is used to construct the first comprehensive feature vector of the signal main window based on the energy ratio of the periodic term, the completeness of the periodic term, the continuity of the strong pulse, and the energy fluctuation value.

5. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 4, characterized in that, Before determining whether each of the signal sub-windows included in the main signal window belongs to a strong pulse sub-window and obtaining the strong pulse continuity of the main signal window, the feature extraction module further includes: A respiratory segment screening unit is used to screen out stable respiratory segments from the main signal window based on the energy percentage of periodic items and the completeness of periodic items in the main signal window. The threshold determination unit is used to determine the average value of the swallowing residual signal within the stable breathing segment as the stable signal threshold.

6. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 1, characterized in that, The type recognition module includes: The distance determination unit is used to determine the sample Euclidean distance between the signal main window and the labeled training samples based on the first comprehensive feature vector of the signal main window and the second comprehensive feature vector of the labeled training samples. A type identification unit is used to determine the respiratory cycle type of the nasal airflow signal in the main signal window based on the sample Euclidean distance between the main signal window and each of the labeled training samples, and the respiratory cycle type label of each of the labeled training samples.

7. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 6, characterized in that, The distance determination unit is used for: The first comprehensive feature vector and the second comprehensive feature vector are standardized respectively to obtain the first standardized feature vector and the second standardized feature vector. The squared feature differences between the first standardized feature vector and the second standardized feature vector in each feature dimension are multiplied by the discriminant index of the corresponding feature dimension to obtain the feature difference between the main signal window and the labeled training sample in each feature dimension. Based on the feature difference between the main signal window and the labeled training samples in each feature dimension, the sample Euclidean distance between the main signal window and the labeled training samples is determined.

8. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 7, characterized in that, Before multiplying the squared feature differences between the first standardized feature vector and the second standardized feature vector in each feature dimension by the discriminant index of the corresponding feature dimension to obtain the feature difference between the main signal window and the labeled training samples in each feature dimension, the type recognition module further includes: The information acquisition unit is used to acquire the first average value and first standard deviation of each labeled training sample whose respiratory cycle type label is a pure respiratory type label in the target feature dimension, and to acquire the second average value and second standard deviation of each labeled training sample whose respiratory cycle type label includes a swallowing type label in the target feature dimension; the target feature dimension can be any one of the feature dimensions; The discrimination determination unit is used to determine the discrimination index of the target feature dimension based on the first average value, the first standard deviation, the second average value, and the second standard deviation.

9. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 1, characterized in that, The efficiency evaluation module includes: The participation determination unit is used to determine the oral breathing participation of the child to be monitored in the respiratory cycle corresponding to each of the main signal windows based on the respiratory cycle type of the oral and nasal airflow signals in each of the main signal windows; The window filtering unit is used to filter out target signal main windows from each of the signal main windows based on the mouth breathing participation of the child to be monitored in the respiratory cycle corresponding to each of the signal main windows. The mouth breathing participation is greater than a preset participation threshold. An efficiency evaluation unit is used to determine the target mouth breathing efficiency of the child to be monitored based on the window duration percentage of the target signal main window.

10. The system for monitoring children's mouth breathing function and evaluating respiratory efficiency according to claim 9, characterized in that, The participation determination unit is used for: Based on the respiratory cycle type of the oral and nasal airflow signal in the main signal window, obtain the upper energy threshold and lower energy threshold corresponding to the respiratory cycle type; Based on the energy percentage of the periodic items within the main signal window, the upper energy threshold, and the lower energy threshold, the oral breathing participation of the child to be monitored within the respiratory cycle corresponding to the main signal window is determined; the energy percentage of the periodic items is used to characterize the ratio obtained by dividing the sum of the energy of the respiratory cycle items within the main signal window by the sum of the energy of the respiratory cycle items within the main signal window and the sum of the energy of the swallowing residual items.

Images