Search

CN-121987183-A - Snore relieving system and method based on breathing sound blood oxygen and nitric oxide asthma identification

CN121987183ACN 121987183 ACN121987183 ACN 121987183ACN-121987183-A

Abstract

The invention relates to the technical field of respiratory function monitoring, in particular to a snore stopping system and a snore stopping method based on breathing blood oxygen and nitric oxide asthma recognition, which are characterized in that abnormal time sequence association constraint formed by a time sequence convolution network is introduced to ensure that abnormal judgment is not stopped in independent recognition within a single time window, but abnormal types of adjacent time positions are used as continuous sequences to carry out overall constraint and consistency check, the risk assessment is changed from simple time sequence judgment to propagation and consistency analysis based on the connection relation between nodes by introducing node propagation weight formed by a graph neural network, abnormal grade change and time position information of adjacent turning nodes are summarized through multi-round adjacent node information transmission, a stable link mark is formed by a continuous risk section under the consistency check of the propagation weight, the matching degree of intervention time and ventilation night risk evolution rhythm is improved, and quantifiable description of risk accumulation and turning is realized under a sleeping scene.

Inventors

  • HAN GUOJING
  • CHEN YIBING
  • LIANG ZHIXIN

Assignees

  • 中国人民解放军总医院第一医学中心

Dates

Publication Date
20260508
Application Date
20260127

Claims (10)

  1. 1. Snore relieving system based on breath sound blood oxygen and nitric oxide asthma recognition, characterized in that the system comprises: the breath sound driving mapping module is used for positioning the starting and ending nodes of the mouth-nose ventilation and marking a single-period ventilation period based on the triggering state of the breath period, comparing the duration change directions of adjacent periods, endowing a stable period ventilation driving grade and sequentially solidifying to generate a ventilation driving distribution mark; Based on the ventilation driving distribution marks, reading the blood oxygen value change amplitude of the corresponding time window and marking the continuous descending section, reading the nitric oxide value change direction, combining the states and screening out conflict combinations to obtain a breathing abnormality type mark; the sleep monitoring period state segmentation module is used for counting the occurrence times of each abnormal period after dividing the continuous time period of the sleep monitoring period by combining the breathing abnormal type identification and the abnormal time sequence association constraint formed by the time sequence convolution network, selecting the dominant abnormal type and comparing the abnormal consistency of the adjacent time period, and establishing a night ventilation state segmentation; The risk evolution path module screens a risk increment turning section based on the night ventilation state segmentation, sequences the time positions of the state sections and combines node propagation weights formed by the graph neural network, synchronously checks the corresponding ventilation driving distribution marks and the blood oxygen nitric oxide change sections, and outputs a risk change track; And the snore stopping control triggering module is used for aligning the high-risk state segment with the risk turning node based on the night ventilation state segment and the risk change track, counting the number of continuous abnormal segments, calculating the inter-segment time interval, and matching the triggering segment with the observing segment to obtain a snore stopping intervention instruction.
  2. 2. The snore relieving system based on breath sound blood oxygen and nitric oxide asthma recognition of claim 1, wherein the ventilation driving distribution markers comprise ventilation driving grades, grading sequences and corresponding time window identifiers, the breathing abnormality type identifiers comprise blood oxygen change association abnormality types, nitric oxide change association abnormality types, blood oxygen and nitric oxide combination abnormality types, the night ventilation state segments comprise time period numbers, time period starting and ending ranges and time period leading abnormality types, the risk change track comprises risk turning node positions, adjacent turning node connection relations and corresponding propagation weight results, and the snore relieving intervention instructions comprise triggering section identifiers, observation section identifiers, starting identifiers and ending identifiers.
  3. 3. The snore relieving system based on breath sound blood oxygen and nitric oxide asthma recognition of claim 1, the breath sound driving mapping module is characterized by comprising: The ventilation period positioning sub-module is used for identifying continuous triggering time points and pairing adjacent triggering point intervals based on the respiratory period triggering state, calibrating a single period boundary formed by an oral-nasal ventilation starting node and an end node, recording the duration of each period, eliminating abnormal offset periods and generating a ventilation period effective identifier; and the driving level solidifying sub-module is used for reading the duration of each effective period based on the effective identification of the ventilation period, arranging the effective periods according to a time sequence, grouping and classifying the continuous change directions of adjacent periods, endowing the corresponding ventilation driving levels, locking the sequence of the continuous levels and generating a ventilation driving distribution mark.
  4. 4. The snore relieving system based on breath sound blood oxygen and nitric oxide asthma recognition of claim 1, wherein the breathing abnormality joint discrimination module comprises: based on the ventilation driving distribution marks, reading blood oxygen values in a corresponding time window and arranging the blood oxygen values in time sequence, performing point-by-point difference calculation on adjacent values, marking descending directions, continuously connecting sections with consistent directions, checking continuous spans, excluding intermittent sections, and generating blood oxygen change section marks; And the nitrogen-oxygen combined judging sub-module is used for reading the nitric oxide values in the corresponding time window based on the blood oxygen change section identification, marking the change direction, connecting the time periods with consistent directions, checking the duration length, aligning the time of each blood oxygen section and the nitric oxide time period, removing the inconsistent sections, classifying the consistent sections and obtaining the breathing abnormality type identification.
  5. 5. The snore relieving system of claim 1, wherein the sleep monitoring cycle status segmentation module comprises: Based on the breathing abnormality type identification, combining a time sequence association constraint result of a time sequence convolution network on abnormal distribution of a sleep monitoring period, sequentially dividing long-time sections along a time axis of the sleep monitoring period, writing abnormality types of all time points into the sections and recording abnormality occurrence positions in the sections one by one, and after the occurrence times of all the abnormalities are accumulated, arranging section contents according to time sequence to generate abnormal distribution of the time period; The dominant anomaly judging sub-module is used for reading anomaly accumulation results in each time period based on the anomaly distribution of the time period, comparing the occurrence times of anomalies in the same section item by item, selecting the peak value of the times as a dominant item, and recording the change condition of dominant anomalies between adjacent sections to obtain dominant anomaly arrangement; And the state boundary solidification submodule is used for comparing the consistency of the dominant anomalies of the adjacent sections section by section based on the dominant anomaly arrangement, continuously merging the consistent sections, prolonging the continuous length, reserving the boundary position of the inconsistent sections, writing the section numbers and the lengths into the sections after merging, and establishing the night ventilation state section.
  6. 6. The snore relieving system based on breathing sound blood oxygen and nitric oxide asthma recognition according to claim 1, wherein the time sequence convolution network sequentially receives a time sequence formed by the breathing abnormality type identifiers according to a sleep monitoring period time axis, sequentially selects adjacent continuous positions on the time axis to form a fixed length window, performs convolution kernel coverage and position displacement processing on the abnormality type arrangement in the window, performs cumulative recording on response values generated under different coverage windows at the same time position, outputs corresponding time sequence association constraint results at each position of the time axis, and writes the time sequence association constraint results into an abnormality distribution data structure in a time index alignment mode.
  7. 7. The snore relieving system based on breath sound blood oxygen and nitric oxide asthma recognition of claim 1, the risk evolution path module is characterized by comprising: a state sequence construction submodule, based on the night ventilation state segmentation, reading the segment numbers of each state segment and sequencing according to the occurrence time, writing connection relations after connecting adjacent state segments one by one, recording the corresponding continuous length of each connection, checking the sequence continuity, eliminating missing connection and generating a state segment sequence structure; The turning node screening submodule is used for comparing abnormal grade change directions of adjacent state segments one by one based on the state segment sequence structure, marking the change positions from low grade to high grade, synchronously reading corresponding ventilation driving distribution marks and blood oxygen nitric oxide change sections, and aligning time to obtain a risk turning node set; And the evolution chain solidification submodule is used for introducing a judgment result of propagation weight and link consistency among nodes formed by the graph neural network based on the risk turning node set, connecting adjacent turning nodes in time sequence, checking state segment continuity among the nodes, removing a broken section, writing the continuous nodes into a unified link identifier, recording the starting and ending positions and the coverage segment numbers of the links, and outputting a risk change track.
  8. 8. The snore relieving system based on breathing sound blood oxygen and nitric oxide asthma recognition according to claim 1, wherein the graph neural network constructs a node set through a risk turning node set, writes a time adjacent relation and a state segment continuous relation between adjacent turning nodes into a node connection relation, carries out multi-round adjacent node information transfer according to the connection relation, gathers abnormal grade change information and time position identification of the adjacent nodes in each round of transfer, records weight values after propagation at node level, carries out consistency judgment on continuous links formed by a plurality of nodes, writes node connection meeting continuity conditions into a unified link identification structure and outputs corresponding propagation weight results.
  9. 9. The snore relieving system based on breath sound blood oxygen and nitric oxide asthma recognition of claim 1, the snore relieving control trigger module is characterized by comprising: the risk alignment screening submodule is used for comparing the time range of the state segment with the time position of the turning node segment by segment based on the night ventilation state segment and the risk change track, finishing alignment through overlapping interval verification, counting the number of continuous high risk state segments, calculating the time interval of adjacent segments, screening out interval overrun combinations and generating an intervention triggering segment mark; And the intervention instruction generation submodule is used for matching the time ranges of the corresponding observation sections one by one and positioning the positions of the risk falling nodes based on the intervention triggering section identifiers, checking the sequence relation between the start and the stop of the triggering section and the falling nodes, writing in the start identifier and the end identifier, and summarizing according to the time sequence to obtain the snore stopping intervention instruction.
  10. 10. A method for snoring prevention based on breath sound blood oxygen and nitric oxide asthma recognition, characterized in that the system for snoring prevention based on breath sound blood oxygen and nitric oxide asthma recognition according to any one of claims 1-9 is performed, comprising the steps of: S1, determining an initial node and an end node of the ventilation of the mouth and the nose based on a respiratory cycle triggering state, forming a single-cycle ventilation period, comparing the duration change directions of adjacent cycles, endowing stable ventilation driving grades, and solidifying in time sequence to generate a ventilation driving distribution mark; S2, based on the ventilation driving distribution marks, reading the blood oxygen value change in the corresponding time window and marking the continuous descending section, simultaneously reading the nitric oxide value change direction, combining the two change states and removing the conflict situation to obtain a breathing abnormality type mark; S3, dividing a sleep monitoring period time axis into continuous time periods based on the breathing abnormality type identification and combining abnormality time sequence association constraint formed by a time sequence convolution network, counting the occurrence times of abnormality of each period, selecting a dominant abnormality type, comparing the consistency of adjacent periods, and establishing night ventilation state segmentation; S4, based on the night ventilation state segmentation, the state segments are arranged in time sequence, the turning segments with the abnormal level increasing are screened by combining the node propagation weight formed by the graph neural network, the ventilation driving distribution marks and the blood oxygen nitric oxide change segments are synchronously checked, and a risk change track is output; And S5, aligning the high-risk state segment and the risk turning node based on the night ventilation state segment and the risk change track, counting the number of continuous abnormal segments, calculating the time interval between segments, and matching the triggering segment with the observing segment to obtain the snore stopping intervention instruction.

Description

Snore relieving system and method based on breathing sound blood oxygen and nitric oxide asthma identification Technical Field The invention relates to the technical field of respiratory function monitoring, in particular to a snore stopping system and a snore stopping method for recognizing blood oxygen and nitric oxide asthma based on breathing sound. Background The technical field of respiratory function monitoring aims at measurable physiological signals generated in the respiratory process of a human body, and through continuous and periodic monitoring of respiratory activity related parameters, the assessment, abnormal identification and risk judgment of respiratory function states are realized, the respiratory function monitoring is usually focused on air flow smoothness, air exchange capacity and respiratory rhythm stability, and covers application scenes of sleep respiratory disorder, chronic airway diseases and respiratory attenuation states, the field emphasizes the direct relevance between a monitored object and the respiratory physiological process, and the monitoring result is required to reflect the airway obstruction degree, respiratory efficiency change and disease induction characteristics, so that a quantitative basis is provided for subsequent intervention and treatment. The snore stopping system based on breathing sound blood oxygen and nitric oxide asthma recognition is a respiratory state recognition and intervention system for users in sleeping and resting states, the system uses sound signals generated in the breathing process and combines blood oxygen saturation and nitric oxide related parameters as a discrimination basis to distinguish normal breathing, snore states and asthma induced respiratory abnormal states, and the system aims to recognize respiratory abnormalities possibly accompanied with airway contraction and airway inflammation reaction on the premise of not depending on complex operation, and trigger snore stopping and intervention processes when the recognition result meets preset conditions, thereby reducing fluctuation amplitude of respiratory resistance, improving night ventilation continuity, reducing occurrence probability of repeated snore and respiratory interruption and improving respiratory stability during sleeping. In the prior art, normal respiration, snore state and asthma-induced respiratory abnormal state are distinguished by taking respiratory sound and combining blood oxygen saturation and nitric oxide related parameters as distinguishing bases, operation mode bias is mainly distinguished by immediate state in a resting scene and triggering meeting preset conditions, a periodic structure difference which appears under the conditions of posture change, short-time breath-holding and snore strong and weak fluctuation at the same night is difficult to form unified reference due to lack of a sequential solidification mechanism for the duration change direction and stable driving level of a respiratory cycle, the distinguishing bases are more dependent on transient segments, the situation that the same state is repeatedly classified into different types in adjacent time windows is easy to appear, and state evaluation at night is discontinuous, although the prior art emphasizes anomaly identification and risk judgment, clear constraint is not available for abnormal type distribution and adjacent time period consistency on a sleep monitoring period time axis, risk judgment is difficult to provide increasing turning time and stage link support, the triggering process is easier to stay at a single threshold meeting triggering level, short-time occasional anomalies and continuous accumulated anomalies are difficult to distinguish, and frequent triggering intervention and missed triggering concurrent risks appear. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a snore stopping system and a snore stopping method based on breathing sound blood oxygen and nitric oxide asthma identification. In order to achieve the above purpose, the invention adopts the following technical scheme that the snore relieving system based on breathing sound blood oxygen and nitric oxide asthma identification comprises: the breath sound driving mapping module is used for positioning the starting and ending nodes of the mouth-nose ventilation and marking a single-period ventilation period based on the triggering state of the breath period, comparing the duration change directions of adjacent periods, endowing a stable period ventilation driving grade and sequentially solidifying to generate a ventilation driving distribution mark; Based on the ventilation driving distribution marks, reading the blood oxygen value change amplitude of the corresponding time window and marking the continuous descending section, reading the nitric oxide value change direction, combining the states and screening out conflict combinations to obtain a breathing abn