An intelligent monitoring and auxiliary diagnosis system and method based on multi-modal cough acoustic analysis

Abstract

The application discloses an intelligent monitoring and auxiliary diagnosis system and method based on multi-modal cough acoustic analysis, relates to the technical field of intelligent monitoring and auxiliary diagnosis, and comprises a natural cough triggering component, a recovery window judging component, a coding perturbation applying component, a multi-path synchronous receiving component, a recovery hysteresis analyzing component, a consistency verifying component and an output control component; natural cough main events are confirmed through a near-field acoustic channel and a contact vibration channel, a recovery window is determined at a main peak ending moment, two-stage coding perturbations are applied in the recovery window, and active responses are synchronously collected through an air path receiving channel and a structure path receiving channel; passive tail trace responses and the active responses are combined to generate a recovery reopening step graph and a path inversion hysteresis loop, and based on challenge consistency marks, cause-related risk prompts, disease activity reference information and dynamic trend reference information or to-be-reviewed marks are output. The application is beneficial to improving the stability of cough recovery state monitoring.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring and assisted diagnosis technology, specifically an intelligent monitoring and assisted diagnosis system and method based on multimodal cough acoustic analysis. Background Technology

[0002] Cough sounds, chest wall vibrations, airflow disturbances, and the recovery process after coughing all contain technical information related to the respiratory state. With the development of wearable sensors, near-field acoustic acquisition, edge computing, and multimodal signal processing technologies, cough monitoring has gradually shifted from manual recording and single auscultations to long-term, continuous, and automated data acquisition. Such systems typically aim to capture naturally occurring cough events in home, ward, follow-up, or screening environments and convert cough fragments into acoustic characteristics, event frequency, activity trends, or risk indicators for subsequent analysis.

[0003] Existing technologies include continuous cough monitoring methods that utilize human vibration signals to trigger microphone acquisition. For example, the scheme disclosed in publication number WO2024066962A1 continuously listens for suspected cough events using human vibration signals, activates the microphone after a trigger condition is met, and combines acoustic and physiological data to determine health status. This approach can reduce power consumption and invalid recording issues caused by a constantly active microphone, and also helps reduce interference caused by other people's voices or coughs. However, the focus of this type of technology is on how to stably capture the main natural cough event, and it usually still uses the audio or physiological data after triggering as the object of subsequent analysis, lacking a protocol-based design for actively measuring the short recovery phase after the cough peak ends.

[0004] In the field of respiratory mechanics measurement, forced oscillation techniques and related active oscillation measurement techniques have been publicly disclosed for measuring respiratory system impedance, resistance, responsiveness, or airway distensibility through small-amplitude pressure, airflow, or acoustic disturbances. For example, WO2012014024A1 discloses applying small-amplitude pressure changes during spontaneous breathing and recording pressure and airflow responses; the literature "Methodology and Implementation of Forced Oscillation Technique for Respiratory Mechanics Measurement" and "Assessment of Airway Distensibility by the Forced Oscillation Technique" also illustrate the technical basis for using externally applied controlled oscillation signals for respiratory system mechanics measurement. These techniques can transform mechanical properties that are difficult to separate from passive breathing observations into active responses, but they are usually geared towards spontaneous breathing or pulmonary function measurement scenarios, do not use the recovery window after the main event of natural cough as the measurement entry point, and do not establish a consistency verification relationship between the natural cough trail and the active perturbation response.

[0005] In the field of active acoustic echo or acoustic reflection measurement, techniques exist for inferring changes in airways, tubing, or local structures by detecting emitted sound waves, sound pulses, or echoes. For example, US12337110B2 relates to using acoustic reflection information in ventilation equipment to determine the status of tubing or airways; the SonarMed AirwayMonitoring System discloses the identification of endotracheal tube location, movement, and obstruction using a transmitter and microphone receiver; and Acoustic reflectometry for airway measurement also introduces the basic principles of reconstructing changes in airway cross-sectional area using acoustic pulses and reflected signals. These approaches demonstrate that active acoustic measurement can convert structural changes into continuously interpretable signals, but their focus is typically on artificial airways, tubing, obstruction locations, or geometric reflection information, without directly addressing the issue of how the recovery state after a natural cough event is synchronously correlated with a multipath active response.

[0006] Another approach involves cough acoustic classification and intelligent auxiliary judgment, which extracts spectrum, energy, temporal, and depth features or event frequency from cough audio and outputs classification results, risk warnings, or trend information. This type of method has engineering value in terms of low-cost acquisition and large-scale screening. However, when environmental noise, changes in body position, sensor attachment status, continuous and dense coughing, or cross-interference from multiple patients occur, relying solely on natural cough peaks or passive wake features can easily lead to event boundary drift and feature instability. If the system directly releases output information based on a single passive acoustic anomaly, sporadic noise, coughs from non-wearers, or channel artifacts may also enter the subsequent judgment chain.

[0007] In summary, existing technologies disclose segmented capabilities such as natural cough event triggering, active small-amplitude respiratory mechanics measurement, and active acoustic reflection measurement. However, significant gaps remain between these capabilities. Natural cough triggering technology primarily addresses the event entry problem, active oscillation technology mainly addresses the mechanical response measurement problem, and active echo technology mainly addresses the structural or pipeline status measurement problem. Existing disclosures rarely use the end time of the main peak after the natural cough event as a common time reference, apply two-stage coded perturbations with different detection functions within a limited recovery window, and simultaneously acquire the active responses of the air path and structural path.

[0008] Therefore, there is an urgent need for a multimodal monitoring technology for the natural cough recovery phase, enabling the system not only to capture the main natural cough event but also to generate a controlled active challenge response within an appropriate recovery window after coughing, and to transform both the passive trail response and the active response into verifiable intermediate technical objects. The core of this technical challenge lies in improving the reliability of cough recovery status data acquisition, response alignment, anomaly filtering, consistency verification, and dynamic trend output. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art and propose an intelligent monitoring and auxiliary diagnosis system and method based on multimodal cough acoustic analysis to solve the above-mentioned problems.

[0010] The objective of this invention is achieved through the following technical solution: an intelligent monitoring and auxiliary diagnostic system based on multimodal cough acoustic analysis, comprising a natural cough triggering component, a recovery window determination component, an coded perturbation application component, a multipath synchronous reception component, a recovery hysteresis analysis component, a consistency verification component, and an output control component. The natural cough triggering component is configured to confirm the main natural cough event based on the near-field acoustic channel and the contact vibration channel, determine the end time of the main peak of the main natural cough event, and use the end time of the main peak as a common time reference for recovery window determination and response alignment. The recovery window determination component is configured to determine the recovery window after the end time of the main peak based on the inflection point of the main peak energy decay, glottal reopening characteristics, contact vibration fallback state, and the risk of the next event, and output the recovery window to the coded perturbation application component. The coded perturbation application component is configured to apply a two-stage coded perturbation within the recovery window. The two-stage coded perturbation includes a first excitation segment undertaking overall damping detection and a second excitation segment undertaking local reopening detection, forming a... The system includes: a recovery window-bound active challenge signal; a multi-path synchronous receiving component configured to synchronously acquire the active response generated by the active challenge signal through the air path receiving channel and the structure path receiving channel, and align the active response with the end time of the main peak; a recovery hysteresis analysis component configured to receive the passive wake response and active response after the main event of natural cough, merge the passive wake response and active response, generate a recovery reopening ladder diagram based on the frequency band recurrence order and frequency band hysteresis time, and generate a path reversal hysteresis loop based on the change in response strength in the air path receiving channel and the structure path receiving channel of the same frequency band; a consistency verification component configured to generate a challenge consistency flag based on the recovery criterion corresponding to the passive wake response and the perturbation criterion corresponding to the active response; and an output control component configured to output etiology-related risk warnings, disease activity reference information, and dynamic trend reference information based on the recovery reopening ladder diagram and the path reversal hysteresis loop when the challenge consistency flag is in a passed state, and generate a check mark when the challenge consistency flag is in a failed state.

[0011] The natural cough trigger component is configured to read the acoustic peak of the near-field acoustic channel and the vibration peak of the contact vibration channel. When the arrival time difference between the acoustic peak and the vibration peak falls within the set synchronization range, the natural cough main event is confirmed, and the end time of the main peak is determined by the energy drop segment of the acoustic peak and the fall segment of the vibration peak.

[0012] The recovery window determination component is configured to generate a candidate recovery period after the end of the main peak, read the main peak energy decay inflection point, glottis reopening characteristics, contact vibration fallback state and the risk of the next event within the candidate recovery period, and determine the candidate recovery period as the recovery window when the above four items meet the perturbation entry condition.

[0013] The coding perturbation application component is configured such that the first excitation segment and the second excitation segment have different frequency band ranges, coding lengths, output timings, and energy boundaries. The output time of the first excitation segment is earlier than the output time of the second excitation segment, and both the first excitation segment and the second excitation segment are located within the recovery window.

[0014] The air path receiving channel includes a pre-larynx near-field receiving channel, and the structural path receiving channel includes a chest wall vibration receiving channel and a bed structure receiving channel. The multi-path synchronous receiving component is configured to record the response start point, energy attenuation, missing frequency band, and stabilization time of the active response under the same clock reference.

[0015] The recovery hysteresis analysis component is configured to generate a recovery and reopening ladder diagram based on the frequency band recurrence order, frequency band hysteresis time, and path recovery order in the passive wake response and active response. It also generates a path reversal hysteresis loop based on the response strength changes, damping reversal point, and hysteresis loop area in the air path receiving channel and the structure path receiving channel of the same frequency band.

[0016] The consistency verification component is configured to compare the recovery criterion corresponding to the passive trail response with the perturbation criterion corresponding to the active response. When both meet the set consistency conditions, a challenge consistency flag indicating a pass state is generated. When neither meets the set consistency conditions, a challenge consistency flag indicating a fail state is generated.

[0017] The recovery hysteresis analysis component is also configured to establish a patient recovery challenge map, which includes postural conditions, natural wake response, two-stage coded perturbation response, high-frequency reopening order, and hysteresis loop area. The output control component is also configured to perform a drift comparison between the patient recovery challenge map corresponding to the current natural cough master event and the patient recovery challenge map of the same patient in the stable period, and correct the dynamic trend reference information based on the drift comparison results.

[0018] A smart monitoring and assisted diagnosis method based on multimodal cough acoustic analysis includes: confirming the main event of natural cough through a natural cough triggering component based on near-field acoustic channels and contact vibration channels, determining the end time of the main peak, and using the end time of the main peak as a common time reference; determining the recovery window after the end time of the main peak through a recovery window determination component; applying a two-stage coded perturbation within the recovery window through a coded perturbation application component; synchronously acquiring the active response through a multipath synchronous receiving component and aligning the active response with the end time of the main peak; merging the passive wake response and the active response through a recovery hysteresis analysis component to generate a recovery reopening ladder diagram and a path reversal hysteresis loop; generating a challenge consistency flag through a consistency verification component; and outputting etiology-related risk warnings, disease activity reference information, and dynamic trend reference information through an output control component when the challenge consistency flag is in a passed state, and generating a check mark when the challenge consistency flag is in a failed state.

[0019] The first excitation segment is extended when the patient is in a supine position at night, the second excitation segment is shortened and the invalidation threshold is increased when the patient is coughing continuously and intensively, and the recovery window is refused when the event start points of the air path receiving channel and the structure path receiving channel are inconsistent.

[0020] The beneficial effects of this invention are: By using both the near-field acoustic channel and the contact vibration channel to confirm the primary event of natural cough, and using the end time of the primary peak as a common time reference for recovery window determination and active response alignment, this invention helps reduce event boundary offsets caused by environmental noise, changes in body position, or localized sensor loosening in single-path acquisition. Existing cough analysis methods that rely solely on acoustic segments often struggle to distinguish between natural decay, glottal reopening, and interference from the next event after the cough primary peak ends. This invention, through the synchronous constraint of the acoustic and vibration primary peaks, ensures that the subsequent recovery window is not merely a preset fixed time period, but is jointly determined by the state variables after the primary event. This time reference provides a consistent alignment entry point for passive wake responses, active challenge responses, and long-term patient profiles, thereby improving the stability and verifiability of the recovery state calculation.

[0021] By determining the recovery window after the peak energy decay inflection point, glottal reopening characteristics, contact vibration fallback state, and the risk of the next event, this invention helps avoid active perturbations being applied before the cough has recovered or before the next cough is about to occur. Compared to fixed-delay triggering or triggering solely based on sound intensity decay, the recovery window determination component incorporates acoustics, structural vibration, and event risk into the gating, ensuring that the two-stage encoded perturbation is executed only during periods more suitable for extracting the recovery response. This processing path reduces the probability of invalid challenge responses entering subsequent fusion calculations and provides a technical basis for abnormal exits in scenarios such as nighttime lying position, continuous and intense coughing, and cross-interference among multiple patients.

[0022] By applying a two-stage coded perturbation within the recovery window, and having the first excitation segment handle overall damping detection while the second excitation segment handles local reopening detection, this invention facilitates the decomputation of the post-cough recovery state into computable technical objects with different physiological response sources. Traditional passive monitoring methods primarily observe the natural cough trail, making it difficult to actively distinguish between changes in overall airway damping and local reopening hysteresis. This invention utilizes different frequency bands, coding lengths, output timings, and energy boundaries to form an active challenge signal, enabling the active response and passive trail response to be compared within the same recovery window. The two-stage challenge coupling index in the embodiment... The technical role of constraining the damping, bandwidth, and energy boundary of the two perturbations is to reduce the interpretation bias caused by the overlap of the excitation responsibilities of the two segments, rather than pursuing a larger excitation intensity alone.

[0023] By synchronously acquiring active responses using both air path and structural path receiving channels under the same clock reference, this invention facilitates the simultaneous acquisition of air path information related to airflow propagation and structural path information related to the chest wall, bed, or contact structures. Existing single-path acquisition methods easily misinterpret channel attenuation, changes in contact state, or external sound source interference as abnormal cough recovery. This invention records the response start point, energy attenuation, missing frequency bands, and stabilization time as alignable fields, and further generates a recovery-reopening step diagram and a path reversal hysteresis loop. This structure allows passive wake responses and active responses to no longer remain at the level of isolated features, but rather to form computational objects with sequential, hysteretic, and closed relationships, thereby improving the technical interpretability of subsequent consistency verification.

[0024] By comparing the recovery criteria corresponding to the passive wake response with the perturbation criteria corresponding to the active response using a consistency challenge flag, this invention facilitates the addition of technical gating before outputting etiology-related risk alerts, disease activity reference information, and dynamic trend reference information. When relying solely on passive acoustic anomalies to directly output results, environmental noise, occasional coughing postures, or abnormal equipment contact may lead to false triggering. This invention generates a verification flag when the consistency condition is not met, rather than directly releasing the complete output. This mechanism enables the output control component to differentiate between valid events, invalid events, and events requiring verification, which helps improve the stability of continuous monitoring results and reduces the risk of misjudgment caused by output directly driven by a single abnormal segment.

[0025] By establishing a patient recovery challenge atlas and comparing the recovery challenge atlas corresponding to the current natural cough master event with the stable phase atlas of the same patient, this invention facilitates the expansion of single-event analysis into continuously updated dynamic trend monitoring. Existing cough acoustic analysis typically focuses on single-segment identification, making it difficult to handle changes in the same patient under different body positions, time periods, or device contact conditions. This invention organizes body position conditions, natural wake response, two-stage coded perturbation response, high-frequency reopening order, and hysteresis loop area as atlas fields. The resulting long-term comparison objects can support the technical correction of disease activity reference information and dynamic trend reference information, while retaining a verification path in cases of insufficient samples, cross-device calibration, and inconsistent event initiation points.

[0026] By setting up processing routes for the complete protocol, passive listening only, and single-stage perturbation in a dedicated comparative experimental embodiment, this invention helps to illustrate the technical division of labor among the various stages of natural cough triggering, recovery window, two-stage coded perturbation, multi-path synchronous reception, recovery hysteresis resolution, and consistency verification. (Comparison of credibility gain) Incorporating recovery window efficiency, ladder graph stability, hysteresis loop closure, consistency flag, and invalid event ratio into the same evaluation object facilitates observation of the improvement direction of the complete protocol relative to the control route. By removing the consistency flag, structural path receive channel, and second excitation segment in the dedicated ablation experimental embodiment, this invention can also use the link necessity index... This indicates the impact of core components on the closure of the technical measurement chain. The aforementioned beneficial effects are based on verifiable technical measurement fields and technical evaluation criteria, and are not extrapolated to medical conclusive evaluations.

[0027] By adjusting the recovery window and invalidation criteria based on scenarios such as nighttime lying position, continuous and frequent coughing, and cross-interference among multiple patients, this invention enhances the robustness of the system in real-world monitoring environments. (Scenario-adaptive recovery window) By incorporating the baseline recovery window, recumbent position markers, intensive cough volume, and cross-interference volume into the same regulatory framework, the system can promptly adjust the challenge timing or refuse entry into the recovery window when the recovery process is prolonged, events are too frequent, or the source of the channels is inconsistent. This approach can improve output fluctuations caused by scene changes in long-term monitoring and make the output of etiology-related risk indicators, disease activity reference information, and dynamic trend reference information more consistent with the closed state of the multimodal data chain. Attached Figure Description

[0028] Figure 1 The process of this invention Figure 1 ; Figure 2 The process of this invention Figure 2 ; Figure 3 The process of this invention Figure 3 . Detailed Implementation

[0029] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Example 1 like Figure 1As shown, this embodiment presents an active challenge implementation method for the recovery window after a single natural cough event. The system includes a natural cough triggering component, a recovery window determination component, a coded perturbation application component, a multi-path synchronous reception component, a recovery hysteresis resolution component, a consistency verification component, and an output control component. The near-field acoustic channel is positioned in front of the mouth and nose to the pre-larynx region, with the pickup port facing the direction of airflow propagation in the pre-larynx region; the contact vibration channel is attached to the suprasternal notch or subclavian region, and the attachment position is recorded as the channel calibration position during initialization; the air path reception channel adopts the pre-larynx near-field reception channel, and the structural path reception channel adopts the chest wall vibration reception channel and the bed structure reception channel. All channels are triggered by the same sampling clock, and the default sampling rate in the project is [missing information]. The envelope update cycle of the contact vibration channel is The channel timestamps during the active challenge period are uniformly written into the event log.

[0031] Work process After system startup, a cold start calibration is performed first. The natural cough trigger component collects no less than [number missing] data. In the quiet background segment, the mean background energy of the near-field acoustic channel and the mean background envelope of the contact vibration channel are obtained; the multipath synchronous receiving component records the baseline noise of the air path receiving channel and the structure path receiving channel when there is no active challenge signal. If the background energy of the near-field acoustic channel is continuous If the calibration limit is exceeded, the system prompts for repositioning of the pickup unit; if the envelope of the contact vibration channel is lower than the attachment stability limit, the system prompts for reattaching the contact vibration unit. After calibration is completed, the system establishes an event cache, the cache fields of which include event time, channel calibration status, natural cough master event flag, peak end time, recovery window, two-stage coded perturbation record, active response, passive wake response, recovery and reopening ladder diagram, path reversal hysteresis loop, challenge consistency flag, and output control status.

[0032] The natural cough trigger component continuously reads the acoustic energy from the near-field acoustic channel. Vibration envelope of contact vibration channel When the arrival time difference between the acoustic peak and the vibrational peak falls within... Both channels are higher than their respective background mean plus When the background standard deviation is doubled, the system confirms the primary event of a natural cough. The system does not use a single acoustic peak as the entry point for the primary event of a natural cough, avoiding triggering the recovery window by far-field speech, bed collisions, or coughs from non-wearers.

[0033] End of main peak It is determined by the acoustic energy descent segment and the vibration return segment: In the formula, The vibration pullback weight is set to a default value in the engineering settings. ; For the Acoustic energy after short-time energy smoothing For contact vibration envelope. Calculation. At that time, the system only after the acoustic peak to Within the search interval, find candidate points, and require that after the candidate points... The internal acoustic energy no longer exceeds the main peak. The peak's end time is simultaneously written to both the recovery window determination component and the multipath synchronization receiving component, serving as a common time reference for recovery window determination and active response alignment.

[0034] The recovery window determination component is in Candidate recovery periods are then generated. The default project range for candidate recovery periods is [missing information]. to Restore window and door control settings Calculate using the following formula: in, It is the indicator of the inflection point of the energy decay of the main peak, and its value is given by the normalized peak value of the second-order change of the energy curve after the main peak. The glottal reopening characteristic is given by the proportion of mid-to-high frequency airflow reproduction; This represents the contact vibration fallback state, with values ​​ranging from the vibration envelope falling back to above the background. The speed within the range is given; To assess the risk of the next event, the value is determined by the most recent event. The number of candidate new peaks, the amplitude of acoustic energy recovery, and the amplitude of vibration recovery are used to determine the peak. , , and The weights are non-negative, and the default values ​​for the projects are as follows: , , and The weights sum to .when and At that time, the candidate recovery period is determined as the recovery window; when If the value is below the threshold, the system will not initiate an active challenge and will instead add a pending review flag.

[0035] The coded perturbation application component outputs a two-stage coded perturbation within the recovery window. The first excitation segment covers... to ,continued to Used to detect overall damping; the second excitation section covers to ,continued to This is used to detect partial reopening. A minimum distance must be maintained between the two segments. The silence interval and the single-segment output energy do not exceed the engineering safety limit. of The coded perturbation application component rereads the contact vibration channel before initiating the first excitation stage. If the vibration envelope suddenly rises and exceeds the level at which the recovery window begins... The system will stop actively challenging and record invalid recovery windows.

[0036] Two-stage challenge coupling index Functional division of responsibilities used to constrain the two-stage incentive mechanism: In the formula, This represents the overall damping value corresponding to the first excitation segment. The duration of the first incentive phase, and These represent the frequency band spans of the first and second excitation segments, respectively. Let be the energy of this perturbation. This formula is used to determine whether a two-stage coded perturbation maintains the active challenge structure of "first overall damping, then local reopening". When Below When this occurs, the encoding perturbation application component stops the second excitation stage, and the output control component marks the event as an incomplete perturbation structure.

[0037] The multipath synchronous receiver component synchronously acquires the air path response during the active challenge signal output. and structural path response And transform both into a unified form relative to Alignment response matrix: matrix The fields include response start point, energy attenuation, missing frequency band, frequency band recurrence time, inter-path response ratio, and stabilization time. Air path response. After bandpass filtering, short-time energy normalization, and background noise subtraction, the signal enters the matrix; structural path response. After envelope extraction, impact artifact removal, and clock alignment, the signal enters the matrix. If the difference in the response start points of the air path receiving channel and the structure path receiving channel exceeds... The multipath synchronous receiving component does not output a valid matrix to the recovery hysteresis parsing component, but instead writes a channel start inconsistency flag.

[0038] The recovery hysteresis parsing component will combine the passive trail response following the natural cough main event with... Merge. Passive trail response taken from The natural recovery segment between the start of the recovery window and the active response segment is taken from the response segments after the first excitation segment, the silence interval, and the second excitation segment. The recovery reopening ladder diagram is arranged from early to late according to the frequency band recurrence time, and the node fields include frequency band number, recurrence time, hysteresis time, path source, and whether it was verified by the active response. The path reversal hysteresis loop is plotted according to the change in response strength in the air path receiving channel and the structure path receiving channel of the same frequency band, and the fields include the inter-path response ratio, damping reversal point, hysteresis loop area, and closure direction. The recovery hysteresis analysis component synchronously outputs the above two intermediate technical objects to the consistency verification component and the output control component.

[0039] Technical Principles The consistency verification component compares the recovery criterion corresponding to the passive wake response with the perturbation criterion corresponding to the active response. When both criteria point to the same frequency band reopening order and the path reversal direction is consistent, the consistency challenge flag is in a passed state; if any of the following conditions exist—frequency band order conflict, path reversal direction conflict, or active response energy lower than the minimum determinable energy—the consistency challenge flag is in a failed state. When the consistency challenge flag is in a failed state, the output control component freezes the output of etiology-related risk warnings, disease activity reference information, and dynamic trend reference information, retaining only the pending review mark, channel status, and event record number. In the frozen state, the recovery window determination component and the coded perturbation application component do not use this event to update subsequent thresholds.

[0040] When the challenge consistency flag is in a passed state, the output control component generates etiology-related risk indicators, disease activity reference information, and dynamic trend reference information based on the recovery reopening ladder diagram and path reversal hysteresis loop. The etiology-related risk indicators are not considered treatment plans, and their source fields include at least the set of recurrence bands, band hysteresis time, inter-path response ratio, hysteresis loop area, and challenge consistency flag. Disease activity reference information is obtained from the ladder diagram offset and hysteresis loop area changes of adjacent events for the same patient; dynamic trend reference information is obtained from the directional changes of consecutive event records. The output control component writes the above outputs, pending review flags, and invalid recovery window flags into the event log.

[0041] In one operation log, the system confirmed that the arrival time difference between the acoustic peak and the vibration peak was [missing information]. Calculations yielded The candidate recovery period is to , , Therefore, it enters the recovery window. The duration of the first excitation phase is... The duration of the second incentive phase is , Active response via matrix After alignment, the recovery and reopening ladder diagram generated by the recovery hysteresis resolution component shows that the low-frequency band recovers first, while the mid-frequency band hysteresis is delayed. The area of ​​the path reversal hysteresis loop is Once the active and passive recovery criteria are consistent, the output control component generates etiology-related risk alerts, disease activity reference information, dynamic trend reference information, and event log numbers, and writes the current recovery window, two-stage coded perturbation, active response matrix, and output control status into the event log table.

[0042] Example 2 like Figure 1 and Figure 2 As shown, this embodiment presents an enhanced implementation of a two-stage coded perturbation parameter adaptive approach. The system retains the natural cough triggering component, recovery window determination component, coded perturbation application component, multipath synchronization reception component, recovery hysteresis resolution component, consistency verification component, and output control component, and sets a parameter constraint table in the coded perturbation application component. The parameter constraint table includes at least the first excitation segment frequency band range, the second excitation segment frequency band range, single segment duration, silence interval, single segment output energy boundary, channel synchronization tolerance, and recovery window effective threshold. This parameter constraint table is generated after cold start calibration and remains frozen until the patient's single natural cough master event enters the recovery window.

[0043] The coded perturbation application component determines the parameter set for this active challenge based on the recovery window length, background noise, contact vibration attachment stability, and the risk of the next event. The center frequency of the first excitation segment... The center frequency of the second excitation segment satisfy ,and lie in to Within the range, lie in to Within the range. If the recovery window length is shorter than... The system prioritizes shortening the second excitation segment; if the background noise increases but remains below the calibrated upper limit, the system prioritizes reducing the output energy of a single segment without changing the order of the first and second excitation segments.

[0044] Parameter adaptation factor Calculate using the following formula: In the formula, This is the length of the window to be restored. The base window length is restored; the default value in the project is... ; To apply stability to the contact vibration channel, To normalize the background noise, Risk of the next event; , , and The weights are non-negative, and the default values ​​for the projects are as follows: , , and .when At that time, the coded perturbation application component employs a complete two-stage active challenge; when When the coding perturbation application component shortens the second excitation segment; when At that time, the system does not impose an active challenge and generates a flag to be reviewed.

[0045] Duration of the first excitation phase Duration of the second excitation segment Determined by the following formula: In the formula, and Based on the duration, the default values ​​for the project are respectively and ; To apply the stability compensation coefficient, the default value for the project is [value missing]. ; This is the risk contraction coefficient for the next event; the default value for the project is [value missing]. Calculated Restricted to to Within the range, Restricted to to Within the range.

[0046] Output energy constraint value Calculate using the following formula: In the formula, To meet the upper limit of engineering safety, The output energy is allowed for single-stage perturbation. The coded perturbation application component reads the energy before the first and second excitation stages are started. If the calculated value is lower than The system then stops the current active challenge and writes an energy constraint insufficiency status.

[0047] The adaptive challenge completeness of this embodiment Determine by the following formula: In the formula, The fundamental center frequency of the second excitation segment is [value], with a default value of [value]. Its output is directly fed to the consistency verification component to determine whether the structure of the two-stage coded perturbation is still maintained during this active challenge. At this time, the consistency verification component will not use this proactive response to generate a challenge consistency passed status.

[0048] In a representative operation record, the recovery window length is The contact vibration channel attachment stability is Normalized background noise is The risk of the next event is The system calculates The coded perturbation application component sets the first excitation segment to... The second excitation segment is set to The single-segment output energy constraint is The center frequency of the second excitation segment remains at [value missing]. hour, The consistency verification component allows the proactive response to enter the challenge consistency judgment chain.

[0049] Example 3 like Figures 1 to 3 As shown, this embodiment provides an implementation method for anomalies, boundaries, and degradation recovery. The system sets up an anomaly state register between natural cough triggering, recovery window determination, coded perturbation application, multipath synchronous reception, recovery hysteresis resolution, consistency verification, and output control. The anomaly state register includes at least six states: inconsistent channel start points, recovery window truncation, decreased attachment stability, insufficient active response energy, unclosed path reversal hysteresis loop, and failure to pass challenge consistency. Each state is bound to the event time, channel calibration status, peak end time, recovery window candidate range, and output control status.

[0050] After confirming the natural cough master event, the natural cough triggering component first performs a channel consistency boundary check. If the arrival time difference between the main peaks of the near-field acoustic channel and the contact vibration channel is still within the specified range... However, the difference in the active response starting point between the air path receiving channel and the structure path receiving channel exceeds [a certain value]. The multipath synchronization receiving component does not output a valid alignment response matrix; instead, it marks this event as a path inconsistency. In this state, the output control component only generates a flag for review and does not output any risk warnings related to the cause of the problem.

[0051] Anomaly Comprehensive Gating Value Calculate using the following formula: In the formula, This indicates a discrepancy in the starting point of the channel. To restore the window cutoff mark, To attach the insufficient stability label, This is a flag indicating that the path reversal hysteresis loop is not closed; the above flag is... or . , , and For abnormal weights, the default values ​​for the project are as follows: , , and .when At that time, the output control component freezes the output of risk warnings related to the cause of the disease and reference information on the activity of the disease, and retains the event log number, channel status and review entry.

[0052] Restoration window cutoff risk Determine by the following formula: In the formula, To restore the number of new peak candidates within the candidate range of the window, To allow an upper limit on the number of new peak candidates, the project default value is [value missing]. ; The magnitude of acoustic energy recovery after the main peak ends. This is the peak energy for natural coughing. It is a stability constant. The recovery window determination component marks the candidate recovery period as a truncation risk state and prohibits the coding perturbation application component from starting the second excitation segment.

[0053] Degraded response integrity Calculate using the following formula: In the formula, For air path response integrity, For structural path response integrity, The integrity of the path reversal hysteresis loop. and The parameters are given by the response start point, missing frequency band, stabilization time, and energy decay fields, respectively. It is given by the area of ​​the open segment, the area of ​​the closed segment, and the number of damping reversal points. When At this time, the recovery hysteresis parsing component does not output the recovery reopening ladder diagram conclusion, but only outputs the verifiable intermediate records.

[0054] The anomaly recovery priority in this embodiment Determine by the following formula: In the formula, To apply stability to the contact vibration channel, The hysteresis loop closure degree of the path reversal; , , , and To restore priority weights, the default values ​​for the project are as follows: , , , and Its output goes directly to the output control component, which determines whether the system allows the resumption of active challenges after the next natural cough master event. Furthermore, when two consecutive events do not show a discrepancy in the channel starting point, the coding perturbation application component restores the complete two-stage coding perturbation.

[0055] In a representative anomaly record, the difference in the response start point between the air path receiving channel and the structure path receiving channel is: The recovery window candidate range appears A new peak candidate, with an attachment stability of [value missing]. The path reversal hysteresis loop closure degree is The system calculates... , , The output control component freezes the output of etiology-related risk warnings and disease activity reference information, and writes the event to the pending review record. The difference in the response start points for the subsequent two events are respectively... and ,and The system restores the complete two-stage coded perturbation.

[0056] Example 4 This embodiment describes the implementation method of the comparative experiment. The comparative experiment used the same near-field acoustic channel, contact vibration channel, air path receiving channel, and structural path receiving channel, maintaining consistent sampling rate, channel calibration method, recovery window candidate range, event filtering rules, and recording format. The present invention group performed natural cough triggering, recovery window determination, two-stage coded perturbation, multi-path synchronous reception, passive wake response and active response merging processing, and challenge consistency flag control output. Control group A only passively listened to natural coughs without applying active challenges. Control group B applied a single-segment perturbation within the recovery window, without distinguishing between overall damped detection and local reopening detection.

[0057] Each batch of records was formed using the same event screening criteria. The main event of natural coughing satisfies the time difference between the arrival of the acoustic peak and the vibrational peak. Furthermore, the background noise was below the channel calibration upper limit. Both the invention group and control group B were in [the specified range]. to The system searches for a recovery window; control group A does not apply an active challenge, but still records passive trail responses corresponding to the candidate range of the recovery window. Each event record includes the peak end time, recovery window state, active challenge state, step plot field, hysteresis loop field, consistency state, and output state.

[0058] Recovery window effectiveness Calculate using the following formula: In the formula, The number of natural cough primary events that were identified as valid recovery windows. The number of primary natural cough events that meet the event screening criteria. Control group A, because no active challenge was applied, The number of events that can be recovered can be recorded based solely on the passive trail response.

[0059] Recovery Hysteresis Analytical Component Calculation of Ladder Plot Stability : In the formula, The number of frequency bands participating in the sorting. For the current event The recovery order of each frequency band This is the baseline recovery order for the same configuration. The baseline recovery order is determined by the order in which the consensus challenge passes within the group. A valid event is generated and remains frozen in this batch of records.

[0060] The recovery hysteresis parsing component also calculates the hysteresis loop closure. : In the formula, Let the area of ​​the opening segment of the path reversal hysteresis loop be denoted by . Let the area be the closed segment. To avoid stability constants with a denominator of zero, the engineering default value is [value missing]. .when At that time, the path reversal hysteresis loop of this event does not enter the candidate set of the output control component. Control group A does not have an active response; if a closed loop cannot be formed solely by the passive wake response, then... according to Included in the evaluation form.

[0061] This embodiment compares the credibility gain. Calculate using the following formula: In the formula, To challenge the consistency indicator score, the state is taken. The status was not passed. ; The proportion of invalid events; , , , and The weights are non-negative, and the default values ​​for the projects are as follows: , , , and And the weights sum to Its output is used to evaluate the technical measurement reliability of the complete active challenge protocol relative to the control protocol.

[0062] The comparison record table is as follows: The recording process of this invention group is as follows: First, the main event of natural cough and the end time of the main peak are identified; then, the recovery window is determined; subsequently, a two-stage coded perturbation is applied and the air path response and structural path response are collected; the recovery hysteresis analysis component generates a recovery-reopening ladder diagram and a path reversal hysteresis loop; and the consistency verification component generates a challenge consistency flag. Control group A lacks an active response, and the recovery hysteresis analysis component cannot form a complete active / passive verification chain; control group B lacks a two-stage functional division, with overall damping and local reopening mixed within the same perturbation segment. When calculated using the default engineering weights, this invention group... Higher values ​​than the two control groups indicate that a complete active challenge protocol is conducive to forming a more stable recovery measurement chain. The above evaluation objects are limited to recovery window effectiveness, ladder plot stability, hysteresis loop closure, consistency pass rate, and proportion of samples to be reviewed.

[0063] Example 5 This embodiment describes the implementation method of the ablation experiment. The ablation experiment uses the same event screening criteria, channel calibration method, recovery window candidate range, and recording fields as in Embodiment 4, and maintains consistent sampling conditions for the near-field acoustic channel, contact vibration channel, air path receiving channel, and structural path receiving channel. This invention group retains the complete link of natural cough triggering, recovery window determination, two-stage coded perturbation, multi-path synchronous reception, recovery hysteresis analysis, consistency verification, and output control. Ablation group C cancels the challenge consistency flag, and the output control component directly generates output information based on the recovery reopening ladder diagram and path reversal hysteresis loop. Ablation group D retains the challenge consistency flag, but cancels the structural path receiving channel, and only uses the air path receiving channel to form an active response. Ablation group E retains multi-path synchronous reception, but cancels the second excitation segment, and only applies the first excitation segment.

[0064] In the ablation experiments, except for the core components that were canceled or disabled, the determination of the peak end time, the candidate range of the recovery window, the basic frequency band of the first excitation segment, the upper limit of channel noise, and the event recording format were kept consistent for each group. Each group recorded the recovery window effectiveness, ladder plot stability, hysteresis loop closure, consistency pass rate, proportion to be reviewed, and output freeze proportion.

[0065] Overall quality value Calculate using the following formula: In the formula, To restore window efficiency, For step plot stability, The hysteresis loop closure degree To challenge the consistency indicator score, The proportion of invalid events; , , , and The weights are non-negative, and the default values ​​for the projects are as follows: , , , and And the weights sum to .

[0066] ablation contribution difference Calculate using the following formula: In the formula, This represents the overall quality value of the invention group. This represents the overall quality value for the corresponding ablation group. This is used to record changes in comprehensive indicators after the removal of a core component, and is not intended as a clinical conclusion.

[0067] Ablation stability retention rate Calculate using the following formula: In the formula, This is the stability constant; the default value in engineering is [value]. .when When this happens, the output control component will mark the canceled link as a high-impact link and retain the enabled status of the link in the configuration record.

[0068] The link necessity index in this embodiment Calculate using the following formula: In the formula, To output the change in the freeze ratio. To freeze the impact weights, the default value for the project is... Its output is used to determine the impact of the ablated element on the closure of the technical measurement chain. When In subsequent configurations, the system will ensure that this core component remains enabled and will not be allowed to be set to be disabled by default.

[0069] The ablation record is as follows: Ablation group C lacks a consistency gate; although the proportion awaiting review is reduced, inconsistency events will directly enter the output path. Ablation group D lacks a structural path receiving channel, resulting in reduced path reversal hysteresis loop closure. Ablation group E lacks a local reopening probe segment, leading to decreased step plot stability and consistency pass rate. The above records indicate that the challenge consistency flag, structural path receiving channel, and second excitation segment all participate in restoring the measurement chain closure.

[0070] Example 6 This embodiment describes the implementation method for application scenarios. The system employs the same natural cough triggering, recovery window determination, two-stage coded perturbation, multi-path synchronous reception, recovery hysteresis parsing, and consistency verification link as single-event processing, and writes events that pass the challenge consistency into the patient's recovery challenge atlas. Application scenarios include continuous nighttime lying position monitoring, continuous intensive cough monitoring, and cross-interference monitoring in multiple patient wards. In all the above scenarios, monitoring information, auxiliary judgment information, or pending verification markers are output, but no treatment plan or treatment plan is output.

[0071] Patient recovery challenge graph generates vectors by event : In the formula, For the first Positional conditions for this event. This is a characteristic of natural wake response. It is a two-stage encoded perturbation response. For high-frequency reopening sequence, The area of ​​the path reversal hysteresis loop is given by the low-frequency attitude components of the bed attitude sensor or contact vibration channel; the natural wake response comes from the passive recovery segment after the end of the main peak; the two-stage coded perturbation response comes from the active response after the first and second excitation segments; the high-frequency reopening sequence and hysteresis loop area are given by the recovery hysteresis analytical component. Vector Write the data into the graph database and bind it to the event time, acquisition channel status, challenge consistency flag, pending review flag, and scene flag.

[0072] The system establishes a stable-period atlas during relatively stable monitoring periods for patients. Instead of using individual events as substitutes, the stable-period atlas generates a mean vector for events that meet the criteria of challenge consistency passing, no cross-interference markers, and no invalid recovery window markers. The stable phase map consists of at least... A valid event is generated; when there are insufficient valid events, the output control component only outputs the monitoring information of a single event and does not update the dynamic trend.

[0073] The deviation between the current event and the stable period pattern is determined by the stable period drift. express: In the formula, As a set of weights for the map fields, the project defaults to setting the weights for positional conditions, natural wake response, two-stage coded perturbation response, reopening order, and hysteresis loop area to [value missing]. , , , and .when Exceeding three times in a row When the output control component marks the dynamic trend as upward; when Falling back to The dynamic trend is marked as stable if two valid events are maintained. If any key field in the graph is missing, the system will not calculate. And write the event to the pending review log.

[0074] Scene Adaptive Recovery Window Determine by the following formula: In the formula, Based on the window length restoration, This serves as a marker for the nighttime lying position. This refers to the amount of continuous, intense coughing. This refers to cross-interference. , and These are scene adjustment coefficients; the default values ​​for the engineering settings are as follows: , and Its output is directly fed to the recovery window decision component to determine whether to allow active challenges and whether to shorten, lengthen, or reject the recovery window. And the difference in the starting point of the event exceeds At that time, the system will Set as And they refused to take the initiative to challenge.

[0075] The output control component calculates the output convergence score before generating etiology-related risk indicators, disease activity reference information, and dynamic trend reference information. : In the formula, To challenge the consistency indicator score, This represents the proportion of invalid events within the most recent monitoring window. When... At this time, the output control component allows the output of etiology-related risk alerts, disease activity reference information, and dynamic trend reference information; when When this happens, the system generates a flag to be reviewed and retains the event in the spectral database but does not use it to update the stable period spectral.

[0076] In a nighttime lying-down scenario, the system reads The first incentive segment is from Extended to The second incentive phase remains If the consistency challenge flag is passed, the current event is written into the patient's recovery challenge map. In a scenario of continuous, intense coughing, if... If more than two primary event candidates of spontaneous coughing appear within the system, the system will... Set as The second excitation stage is shortened to And set the invalidation threshold from Adjust to In a multi-patient ward cross-interference scenario, if the air path receiving channel shows an event peak first while the structural path receiving channel does not show a synchronous vibration peak, the system does not apply a two-stage coded perturbation, writes a cross-interference flag, and freezes the update of the steady-state spectrum for that event.

[0077] The continuous monitoring records are shown in the table below: In the nighttime supine scenario, the system extends the first stimulus segment and maintains challenge consistency, writing the current event into the patient's recovery challenge map and participating in dynamic trend correction. In the continuous, intense coughing scenario, the system shortens the second stimulus segment and increases the invalidation threshold, still allowing [further action / action]. When conditions are met, the system outputs dynamic trends. In scenarios involving multiple patients with overlapping interference, the system refuses to enter the recovery window due to inconsistent event start points; the event is only written to the pending review record and does not enter the stable period map update chain. All the above outputs are monitoring information generated based on multimodal technology data chains and are not intended as treatment plans.

[0078] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be modified within the scope of the concept described herein by means of the above teachings or the technology or knowledge in related fields.

Claims

1. An intelligent monitoring and auxiliary diagnostic system based on multimodal cough acoustic analysis, characterized in that, The system includes a natural cough triggering component, a recovery window determination component, a coded perturbation application component, a multi-path synchronous reception component, a recovery hysteresis analysis component, a consistency verification component, and an output control component. The natural cough triggering component is configured to confirm the main natural cough event based on a near-field acoustic channel and a contact vibration channel, determine the peak end time of the main natural cough event, and use the peak end time as a common time reference for recovery window determination and response alignment. The recovery window determination component is configured to determine a recovery window after the peak end time based on the peak energy attenuation inflection point, glottal reopening characteristics, contact vibration fallback state, and the risk of the next event, and output the recovery window to the coded perturbation application component. The coded perturbation application component is configured to apply a two-stage coded perturbation within the recovery window. The two-stage coded perturbation includes a first excitation segment for overall damping detection and a second excitation segment for local reopening detection, forming an active challenge signal bound to the recovery window. The multi-path synchronous reception component is configured to synchronously acquire the active response generated by the active challenge signal through an air path reception channel and a structural path reception channel, and align the active response with the peak end time. The recovery hysteresis parsing component is configured to receive the passive wake response and the active response after the natural cough master event, merge the passive wake response and the active response, generate a recovery reopening ladder diagram based on the frequency band recurrence order and frequency band hysteresis time, and generate a path reversal hysteresis loop based on the response strength changes of the same frequency band in the air path receiving channel and the structure path receiving channel; the consistency verification component is configured to generate a challenge consistency flag based on the recovery criterion corresponding to the passive wake response and the perturbation criterion corresponding to the active response; the output control component is configured to output etiology-related risk warnings, disease activity reference information and dynamic trend reference information based on the recovery reopening ladder diagram and the path reversal hysteresis loop when the challenge consistency flag is in a passed state, and generate a check mark when the challenge consistency flag is in a failed state.

2. The system according to claim 1, characterized in that, The natural cough triggering component is configured to read the acoustic peak of the near-field acoustic channel and the vibration peak of the contact vibration channel, confirm the natural cough main event when the arrival time difference between the acoustic peak and the vibration peak falls within a set synchronization range, and determine the end time of the main peak based on the energy drop segment of the acoustic peak and the fall segment of the vibration peak.

3. The system according to claim 1, characterized in that, The recovery window determination component is configured to generate a candidate recovery period after the end of the main peak, read the main peak energy attenuation inflection point, glottis reopening characteristics, contact vibration fallback state and the risk of the next event within the candidate recovery period, and determine the candidate recovery period as the recovery window when the above four items meet the perturbation entry condition.

4. The system according to claim 1, characterized in that, The coding perturbation application component is configured such that the first excitation segment and the second excitation segment have different frequency band ranges, coding lengths, output timings, and energy boundaries, the output time of the first excitation segment is earlier than the output time of the second excitation segment, and both the first excitation segment and the second excitation segment are located within the recovery window.

5. The system according to claim 1, characterized in that, The air path receiving channel includes a pre-larynx near-field receiving channel, the structural path receiving channel includes a chest wall vibration receiving channel and a bed structure receiving channel, and the multi-path synchronous receiving component is configured to record the response start point, energy attenuation, missing frequency band and stabilization time of the active response under the same clock reference.

6. The system according to claim 1, characterized in that, The recovery hysteresis analysis component is configured to generate the recovery reopening ladder diagram based on the frequency band recurrence order, frequency band hysteresis time, and path recovery order in the passive wake response and the active response, and to generate the path reversal hysteresis loop based on the response strength changes, damping reversal points, and hysteresis loop areas in the air path receiving channel and the structure path receiving channel of the same frequency band.

7. The system according to claim 1, characterized in that, The consistency verification component is configured to compare the recovery criterion corresponding to the passive wake response with the perturbation criterion corresponding to the active response, generate the challenge consistency flag in a passed state when the two meet the set consistency conditions, and generate the challenge consistency flag in a failed state when the two do not meet the set consistency conditions.

8. The system according to claim 1, characterized in that, The recovery hysteresis analysis component is further configured to establish a patient recovery challenge map, which includes postural conditions, natural wake response, two-stage coded perturbation response, high-frequency reopening order, and hysteresis loop area. The output control component is further configured to perform a drift comparison between the patient recovery challenge map corresponding to the current natural cough master event and the patient recovery challenge map of the same patient in the stable period, and correct the dynamic trend reference information based on the drift comparison results.

9. A smart monitoring and auxiliary diagnosis method based on multimodal cough acoustic analysis, applied to the system described in claim 1, characterized in that, include: The natural cough triggering component confirms the main natural cough event based on the near-field acoustic channel and the contact vibration channel, determines the end time of the main peak, and uses the end time of the main peak as a common time reference; the recovery window determination component determines the recovery window after the end time of the main peak; the coded perturbation application component applies the two-stage coded perturbation within the recovery window; The active response is collected synchronously by the multi-path synchronous receiving component, and the active response is aligned with the end time of the main peak. The passive trail response and the active response are merged and processed by the recovery hysteresis parsing component to generate the recovery reopening ladder diagram and the path reversal hysteresis loop; the challenge consistency flag is generated by the consistency verification component. The output control component outputs the etiology-related risk warning, the disease activity reference information, and the dynamic trend reference information when the challenge consistency flag is in a passed state, and generates the pending review flag when the challenge consistency flag is in a failed state.

10. The method according to claim 9, characterized in that, The first excitation segment is extended when the patient is in a supine position at night, the second excitation segment is shortened and the invalidation threshold is increased when the patient is in a state of continuous and intense coughing, and the recovery window is rejected when the event start point of the air path receiving channel and the structure path receiving channel are inconsistent.

Images