Swallowing dysfunction measurement and evaluation system and training method

Abstract

The invention relates to the technical field of swallowing dysfunction measurement evaluation and training, and discloses a swallowing dysfunction measurement evaluation system and a training method. A closed-loop process from reference data acquisition, signal processing, swallowing event detection, energy ratio analysis, function evaluation to training feedback is constructed. According to the system, a skin microphone is attached to the side of the thyroid cartilage of the neck, swallowing signals of a healthy control and a patient are collected in a non-invasive mode, low-frequency noise is removed through polynomial drifting, a swallowing event is accurately positioned in combination with STFT and envelope extraction, low / high-frequency energy is extracted according to a frequency boundary, and the energy ratio is calculated; and carrying out standard normal score calculation and function grading. A personalized training target is set through a reference template, and a patient completes periodic training according to rhythm and performs self-test evaluation.

Description

Technical Field

[0001] This invention relates to the field of swallowing dysfunction measurement, assessment and training technology, specifically a swallowing dysfunction measurement and assessment system and training method. Background Technology

[0002] Dysphagia refers to a pathological condition in which the bolus or liquid from the mouth to the esophagus is partially or completely obstructed as it passes through the pharynx-esophageal junction, resulting in abnormal swallowing movements. Its main characteristics include pauses in swallowing, coughing, aspiration, a foreign body sensation, or difficulty eating. In severe cases, it can lead to complications such as malnutrition, dehydration, and aspiration pneumonia.

[0003] On the one hand, imaging assessments primarily rely on video fluoroscopy or fiberoptic bronchoscopy for swallowing evaluation. While VFSS allows direct observation of the dynamic process of the bolus passing through the oropharynx, it requires repeated exposure to X-ray radiation, and the equipment is expensive and demands a high level of expertise. FEES, although radiation-free, cannot simultaneously observe the entire flow of the bolus from the mouth to the pharynx and carries a risk of minor damage to the patient's pharynx. Both methods require scripted operation and subjective interpretation by professional physicians, making large-scale, long-term, and self-monitoring difficult. On the other hand, physiological signal detection mainly includes techniques such as swallowing pressure measurement, electromyography (EMG), esophageal manometry, and acoustic analysis. While esophageal manometry can quantify changes in internal pressure, it requires intubation, leading to poor patient compliance and high invasiveness. Surface electromyography can evaluate the activity of the glossopharyngeal muscles, but the signal is easily interfered with by skin impedance, electrode placement deviations, etc., and lacks unified feature extraction and discrimination standards. Acoustic analysis methods, using a neck microphone to collect acoustic vibration signals generated during swallowing, offer advantages such as being non-invasive, inexpensive, and easy to operate. However, current acoustic assessments often rely on simplified energy or temporal characteristics, such as average energy and peak amplitude, lacking fine-grained spectral distribution. They frequently use fixed thresholds or manual experience to set criteria, making them susceptible to background noise and baseline drift, and difficult to guarantee consistency across populations and environments. Furthermore, previous studies mostly employed black-box classification models or empirical formulas, lacking self-calibrating and adaptive parameter optimization mechanisms, resulting in poor reproducibility of assessment results under different individual or device conditions. In terms of swallowing function rehabilitation training, traditional methods rely heavily on physician commands, mirror feedback, or weighted pharyngeal training devices. Training typically requires face-to-face guidance, making it difficult to conduct independently at home. Existing home training tools primarily focus on masticatory muscle strength assessment and tongue thrust exercises, often employing fixed movement patterns. They lack personalized training goal setting and rhythm guidance based on real-time signal analysis, making it difficult to fully motivate patients to actively participate.

[0004] Therefore, this case aims to propose a measurement and assessment system and training method for swallowing dysfunction. It utilizes a neck-mounted microphone to acquire sound signals during swallowing, removes low-frequency drift through multinomial modeling, and then employs short-time Fourier transform (STFT) and energy distribution analysis to perform temporal localization and frequency band deconstruction of swallowing events. This extracts energy ratio indicators reflecting pharyngeal muscle activity characteristics, and combines statistical reference templates and a standard normal score evaluation system to achieve automatic grading of swallowing function. Simultaneously, the system generates customized training objectives and templates based on the degree of difference, and promotes continuous recovery of swallowing function through rhythmic imitation training and self-test feedback. Summary of the Invention

[0005] This invention provides a measurement and assessment system and training method for swallowing dysfunction, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a swallowing dysfunction measurement and assessment system and training method, comprising: A skin microphone was placed next to the thyroid cartilage in the neck of the subject to collect the sound signal of the healthy control subject as baseline data. The reference signal is fitted with a quadratic polynomial to remove low-frequency drift components; The complex spectrum is calculated based on a preset window function, and swallowing event intervals are identified by combining envelope extraction and threshold setting. Based on the frequency boundary parameters, the low-frequency and high-frequency energies in swallowing events are statistically analyzed, and the energy baseline values ​​are determined. Calculate the energy ratio for each swallowing event and generate an average reference energy envelope; The system collects raw sound signals during the patient's swallowing process, performs drift removal and envelope extraction, and identifies swallowing events. Based on the boundary index, low-frequency and high-frequency energies are calculated separately to obtain the energy ratio and perform standard normal fraction calculation. Patient swallowing function categories are then classified according to the scores. The training level is determined based on the relative deviation of the patient's energy ratio from the offline baseline, and a target training ratio is set. Based on the set training goals, patients undergo periodic training and conduct self-testing and training achievement assessments after a specific period.

[0007] Optionally, the step of placing a skin-surface microphone next to the thyroid cartilage in the subject's neck to collect sound signals from healthy control subjects as baseline data specifically includes: Configure a skin-mounted microphone attached to the thyroid cartilage in the neck, and set the sampling frequency to [frequency value missing]. The sampling interval is ; right Healthy control subjects were selected and treated with the same... and duration The sound signal is collected in seconds, and the first sound signal is obtained. The first subject The original sound pressure value at each sampling time is , ;in, The total number of healthy control subjects; This represents the total number of sampling points; Index for subjects; Synchronization in a silent environment, for the first Set of baseline segment indexes for identified subjects: .

[0008] Optionally, the step of performing quadratic polynomial fitting on the reference signal to remove low-frequency drift components specifically includes: For the 10 subjects, at baseline The quadratic polynomial was fitted using the least squares method: , ;in, For the first Baseline drift estimates for 10 subjects; , , The fitting coefficients are obtained by solving the following normal equation: ; Index of fitting coefficients; The first after removing low-frequency drift was obtained The signal of the subjects was .

[0009] Optionally, the step of calculating the complex spectrum based on a preset window function and combining envelope extraction and threshold setting to identify swallowing event intervals specifically includes: set up Window length is Window moved Set the parabolic window function as follows: , ;in, The coefficient of the parabolic window; Index for parabolic window coefficients; calculate Frame count ; Calculate the first The number of subjects in the first Frame, First Complex spectrum values ​​under frequency index : ; in, For frame indexing; Frequency index; Extract the first Temporal envelope of drift signal of subjects ; Set the global swallowing detection threshold to ;in, To find the median function, we take the median of a set of numbers arranged from smallest to largest. Find all the longest connected intervals: satisfy ;in, For the first The total number of swallowing events in the subjects; For the first Index of swallowing events of 10 subjects; For the first The start and end sampling point index of the event in the time sequence.

[0010] Optionally, the step of statistically analyzing the low-frequency and high-frequency energies in the swallowing event based on the frequency boundary parameter and determining the energy baseline value specifically includes: For the The subject, the first For this event, set the frequency boundary parameters: ; ; in, For the first The number of subjects in the first The next incident Low-frequency accumulated energy at the boundary; For the first The number of subjects in the first The next incident High-frequency accumulated energy at the boundary; Calculate the first The number of subjects in the first Local boundary index of the sub-event: ; For all The global bounding index is obtained by taking the median: , ;in, This corresponds to the global intermediate frequency; fixed Calculate the low-frequency energy reference and the high-frequency energy reference respectively: , .

[0011] Optionally, the calculation of the energy ratio of each swallowing event and the generation of an average reference energy envelope specifically includes: Calculate the first The number of subjects in the first Acoustic energy ratio of swallowing ; Calculate the overall mean of all acoustic energy ratios. with standard deviation : , ; Constructing the average reference envelope for training: ;in, For the first The average low-frequency energy envelope of the frame.

[0012] Optionally, the process of acquiring the raw sound signals during the patient's swallowing process, performing drift removal and envelope extraction, and identifying swallowing events specifically includes: Collect patient raw signals Baseline set Fitting polynomial And obtain the drift signal ; Calculate patients With temporal envelope ; Use global threshold Location Swallowing Event Index Set .

[0013] Optionally, the step of calculating low-frequency and high-frequency energies based on the boundary index, obtaining the energy ratio, and performing a standard normal fraction calculation, classifying the patient's swallowing function category according to the score, specifically includes: Fixed Global Boundary Index Calculate the first one respectively Low-frequency energy and high-frequency energy of the second swallow: , ; Calculate the patient's first Energy ratio per swallow ; Calculating the patient's standard normality Fraction ; according to The score categorizes each event into swallowing function categories: ; in, For the first Functional categories of the first swallow; This is normal. Mild disability. It is a severe disability.

[0014] Optionally, determining the training level based on the relative deviation between the patient's energy ratio and the offline baseline, and setting a target training ratio, specifically includes: Calculate the overall average ratio of patients ;in, The total number of swallowing events for the patient; in accordance with Set training level labels relative to offline benchmarks : ;in, No training required; For light training; This is moderate to heavy training. Set training target ratio : ; The training template is the reference envelope sequence. .

[0015] Optionally, based on the set training goals, patients undergo periodic training, and after a specific period, they conduct self-testing and training achievement assessments, specifically including: Set the number of swallows per group. Interval of swallowing movements Number of training sets per group Training cycle , ; Set the length of the reference envelope audio for each group of playback. The patient swallows rhythmically. Second-rate; Obtain the patient's first [test result] after the end of each weekly training day or after the end of 4 weeks of training. Self-test swallowing energy ratio And calculate the self-tested average ratio. ;in, This represents the total number of swallowing events detected through self-testing. Computational training evaluation ; like If the training is successful, then the training is up to standard; if Continue repeating the training until .

[0016] A system for measuring and assessing the swallowing dysfunction, comprising: The swallowing dysfunction measurement and assessment system includes: Offline data processing module: used for hardware setup and baseline data acquisition, signal de-drifting, time-frequency analysis and swallowing event localization, frequency band demarcation and energy calculation, ratio statistics and reference template construction; Online assessment module: used for patient-side signal acquisition and processing, and swallowing function assessment.

[0017] The present invention has the following beneficial effects: 1. This paper proposes a method that uses a skin-mounted microphone attached to the thyroid cartilage in the neck to capture neck audio signals generated during swallowing. Compared with traditional methods using multi-channel physiological sensors (such as electromyography or pressure sensors), this method offers greater ease of signal acquisition, wearability, and user compliance. The system, through parameter settings for sampling frequency and duration, performs standardized benchmark acquisition on healthy subjects in a silent environment, effectively ensuring the consistency and high quality of the reference signal. Traditional systems rely on hospital-equipped facilities to collect electromyography signals, which is not only costly and inaccessible but also requires a high level of technical expertise from users. This solution achieves swallowing recognition through low-cost audio acquisition, significantly reducing the deployment and application barriers.

[0018] 2. A quadratic polynomial model based on a static baseline segment is introduced, and the least squares method of the normal equation is used to remove low-frequency background drift in swallowing audio caused by hardware or environmental changes. Compared with direct filtering methods, polynomial fitting has the advantages of adjustable model, high signal integrity preservation, and more detailed error control, ensuring the accuracy of subsequent energy analysis. Traditional methods often use bandpass filtering to deal with low-frequency drift, but this may result in the loss of important signal components. This method extracts the background change trend in the form of mathematical modeling, which is more interpretable and less likely to lose swallowing transient features.

[0019] 3. This method integrates STFT spectrum calculation and temporal envelope extraction, while simultaneously utilizing threshold and connected component algorithms to identify swallowing segments. The use of a parabolic window function suppresses leakage effects at the window edges, improving the resolution and stability of spectrum estimation. This method not only boasts high accuracy but also enables independent temporal localization for each swallow, making it crucial for swallowing segmentation within the entire system. Existing technologies largely rely on manual labeling or simple threshold detection, resulting in poor accuracy and versatility. This method significantly enhances the robustness and universality of automatic detection through a multimodal feature fusion detection strategy.

[0020] 4. This paper proposes a method to find the optimal balance point based on event-level energy trends and constructs a standard frequency band boundary standard by unifying frequency division using the global median. Based on this, the cumulative energy of low and high frequencies is calculated, and the energy ratio is derived as a quantitative indicator of pharyngeal muscle coordination. This ratio is dimensionless, facilitating standardized comparisons between different individuals. Traditional methods often use swallowing duration or peak frequency as the basis for functional assessment, lacking clear physical or physiological significance. The energy ratio constructed by this method better reflects the changes in pharyngeal muscle vocal dynamics and is easy to standardize, quantifiable, and statistically significant.

[0021] 5. A swallowing training template is constructed using the average energy envelope of a large number of healthy controls. Through audio imitation training, patients are induced to imitate the ideal energy curve within specific time periods, achieving rhythmic recovery of functional muscle groups. This method combines a dual mechanism of "auditory feedback + rhythmic guidance," demonstrating significant advantages in rehabilitation training. Unlike traditional swallowing training that relies solely on physical stimulation or vague guidance, this approach provides a real-time rhythm template, offering clear, audible, and imitable goals for patients' self-training, greatly improving training efficiency and compliance.

[0022] 6. By converting energy ratios into standard normal scores and combining them with set threshold ranges, the functional level of each patient's swallow is classified. This approach is statistically robust, avoids interference from single outliers in the overall judgment, and has good interpretability and repeatability. Conventional methods, such as subjective scoring or classification based on static thresholds, have low accuracy and are subject to subjective bias. This scheme, based on a large-sample statistical model, establishes a unified scoring mechanism, which is conducive to cross-institutional and cross-population applications.

[0023] 7. A closed-loop mechanism is proposed, consisting of periodic group training, post-training self-testing, and result comparison and feedback. After each round of training, the decision to proceed to the next stage is based on whether the training goals have been achieved. This truly realizes a personalized, iteratively optimized, and goal-oriented rehabilitation pathway. Traditional rehabilitation often lacks a real-time feedback mechanism, resulting in low training efficiency and sustainability. This system enhances learning effectiveness and improves patient initiative and recovery outcomes through a three-step closed-loop control of "imitation – feedback – adjustment." Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0026] Example, refer to Figure 1 Swallowing dysfunction measurement and assessment systems and training methods, including: A skin microphone was placed next to the thyroid cartilage in the neck of the subject to collect the sound signal of the healthy control subject as baseline data. The reference signal is fitted with a quadratic polynomial to remove low-frequency drift components; The complex spectrum is calculated based on a preset window function, and swallowing event intervals are identified by combining envelope extraction and threshold setting. Based on the frequency boundary parameters, the low-frequency and high-frequency energies in swallowing events are statistically analyzed, and the energy baseline values ​​are determined. Calculate the energy ratio for each swallowing event and generate an average reference energy envelope; The system collects raw sound signals during the patient's swallowing process, performs drift removal and envelope extraction, and identifies swallowing events. Based on the boundary index, low-frequency and high-frequency energies are calculated separately to obtain the energy ratio and perform standard normal fraction calculation. Patient swallowing function categories are then classified according to the scores. The training level is determined based on the relative deviation of the patient's energy ratio from the offline baseline, and a target training ratio is set. Based on the set training goals, patients undergo periodic training and conduct self-testing and training achievement assessments after a specific period.

[0027] This paper proposes a complete system framework for the measurement, assessment, and training of swallowing dysfunction, constructing a full-process path from baseline data acquisition, signal processing, swallowing event detection, energy ratio analysis, functional level assessment to feedback training. By attaching a microphone to the subject's neck, a non-invasive swallowing signal acquisition method is achieved, replacing traditional invasive or costly electromyography or pressure sensors. Polynomial drift-free processing is then performed to eliminate the influence of low-frequency noise and ensure signal purity. STFT and envelope methods are used to locate swallowing periods, ensuring high-precision extraction of swallowing events. An analytical model of low-frequency to high-frequency energy ratio is introduced to construct quantitative indicators, solving the problems of strong subjectivity and lack of standardized assessment scales in traditional swallowing assessments. Combined with training level classification and periodic training, a closed-loop system is established, improving the controllable training effect and initiative of patients in a home environment, and promoting intelligent and personalized rehabilitation of swallowing disorders.

[0028] The procedure involves placing a skin-tipped microphone near the thyroid cartilage in the subject's neck to collect sound signals from healthy control subjects as baseline data. Specifically, this includes: Configure a skin-mounted microphone attached to the thyroid cartilage in the neck, and set the sampling frequency to [frequency value missing]. The sampling interval is ; right Healthy control subjects were selected and treated with the same... and duration The sound signal is collected in seconds, and the first sound signal is obtained. The first subject The original sound pressure value at each sampling time is , ;in, The total number of healthy control subjects; This represents the total number of sampling points; Indexing of subjects; obtaining high-quality raw swallowing sound signals for subsequent algorithm calibration; Synchronization in a silent environment, for the first Set of baseline segment indexes for identified subjects: Provides a baseline sample without swallowing components for polynomial drift removal.

[0029] This study defines how to collect swallowing sound signals from a healthy control group to build a baseline database, and collects a "baseline segment" in a silent environment where no swallowing behavior occurs. By using a standardized microphone attachment location (next to the thyroid cartilage), a fixed sampling rate, and sampling time, the consistency and comparability of data from different subjects are ensured. In particular, baseline segment annotation is introduced specifically for subsequent drift modeling. This mechanism addresses the problem of background interference and equipment differences affecting the acquired signal, while providing a clean dataset for subsequent polynomial fitting to remove low-frequency background fluctuations, thus enhancing the algorithm's adaptability and generalization ability. This approach improves the standardization of swallowing signal processing in practical applications and provides a reliable foundation for subsequent automatic calibration and model training.

[0030] The process of performing quadratic polynomial fitting on the reference signal to remove low-frequency drift components specifically includes: For the 10 subjects, at baseline The quadratic polynomial was fitted using the least squares method: , ;in, For the first Baseline drift estimates for 10 subjects; , , The fitting coefficients are obtained by solving the following normal equation: ; Index of fitting coefficients; Least squares fitting of normal equations, used to estimate and model low-frequency drift; The first after removing low-frequency drift was obtained The signal of the subjects was Eliminate low-frequency drift noise to obtain a clean signal for feature extraction.

[0031] This paper proposes a method using least squares to perform quadratic polynomial fitting on baseline data to estimate and model the low-frequency drift trend in swallowing signals. This trend term is then extracted from the original signal to obtain a clean swallowing sound signal. Background noise is fitted using mathematical modeling, preserving more original signal features compared to traditional filter methods, making it particularly suitable for analyzing weak signal components in swallowing events. This approach solves the problem of background noise masking the target event in traditional signal analysis, improving the accuracy of subsequent feature extraction and energy calculation. Its advantages lie not only in improving the sensitivity of swallowing event detection but also in enhancing the robustness of the signal processing system to environmental changes, ensuring the accuracy of swallowing function assessment.

[0032] The process of calculating the complex spectrum based on a preset window function, and combining envelope extraction and threshold setting to identify swallowing event intervals, specifically includes: set up Window length is Window moved Set the parabolic window function as follows: , ;in, The coefficient of the parabolic window; Indexing of parabolic window coefficients; reducing window edge leakage and improving the accuracy of spectrum estimation; calculate Frame count ; Calculate the first The number of subjects in the first Frame, First Complex spectrum values ​​under frequency index : ; in, For frame indexing; For frequency indexing; decompose the time-domain signal into a sequence frame spectrum for frequency band energy calculation; Extract the first Temporal envelope of drift signal of subjects Obtaining the temporal envelope facilitates event detection. Set the global swallowing detection threshold to ;in, To obtain the median function, a set of numbers is arranged from smallest to largest and the median is taken; a robust swallowing detection threshold is established to distinguish between events and background. Find all the longest connected intervals: satisfy ;in, For the first The total number of swallowing events in the subjects; For the first Index of swallowing events of 10 subjects; For the first The start and end sampling point index of each event in time sequence; accurately locate the time domain range of each swallow.

[0033] This paper proposes a parabolic window function-based STFT analysis, combined with envelope extraction and thresholding, followed by a connected component detection algorithm to determine the temporal interval of swallowing events. This combined strategy effectively suppresses spectral edge leakage, resulting in more accurate boundary localization of swallowing events in multi-frame analysis. Envelope extraction reflects the trend of signal energy changes, and together with thresholding and connected component screening, forms an automated event detection mechanism with a low false positive rate. This method solves the problem that traditional methods relying on manual observation or static thresholding cannot adapt to individual differences, effectively achieving highly universal and sensitive swallowing event recognition, and improving the intelligence and automation level of the swallowing assessment system.

[0034] The process involves statistically analyzing the low-frequency and high-frequency energies in a swallowing event based on frequency boundary parameters, and determining an energy baseline value. Specifically, this includes: For the The subject, the first For this event, set the frequency boundary parameters: ; ; in, For the first The number of subjects in the first The next incident Low-frequency accumulated energy at the boundary; For the first The number of subjects in the first The next incident High-frequency accumulated energy at the boundary index; quantize low / high-frequency energy within the event window according to the boundary index. Changes; Calculate the first The number of subjects in the first Local boundary index of the sub-event: Find the boundary index where the low / high frequency energy is most balanced for each event; For all The global bounding index is obtained by taking the median: , ;in, For the corresponding global intermediate frequency; unify the low / high frequency boundary standard for all events; fixed Calculate the low-frequency energy reference and the high-frequency energy reference respectively: , Get the low / high frequency energy value for each event at the global index.

[0035] By introducing a "local boundary index" and a "global mid-frequency boundary" mechanism, low-frequency and high-frequency energy segments are statistically analyzed for swallowing events, and energy ratios are further extracted. The most balanced energy boundary point (local boundary index) is calculated for each event, and the median of all samples is used as the global boundary point, achieving data standardization. This method solves the problem of inconsistent frequency band division caused by individual differences in pharyngeal anatomy or swallowing habits, ensuring that all samples can be compared within the same frequency range. Standardized energy benchmark calculations provide highly reliable input data for subsequent ratio calculations and template establishment, enhancing the system's evaluation consistency and analytical stability.

[0036] The calculation of the energy ratio of each swallowing event and the generation of an average reference energy envelope specifically includes: Calculate the first The number of subjects in the first Acoustic energy ratio of swallowing Construct a dimensionless index representing the coordination of pharyngeal muscles; Calculate the overall mean of all acoustic energy ratios. with standard deviation : , Determine the normal distribution parameters of the energy ratio; Constructing the average reference envelope for training: ;in, For the first The average low-frequency energy envelope of the frame; generating audio templates for training to guide patients in imitating typical low-frequency energy patterns.

[0037] This study proposes using the low / high frequency energy ratio as a dimensionless indicator to assess swallowing function, and statistically analyzes its overall mean and standard deviation to generate a training reference template. Compared to traditional pharyngeal muscle atlases or spectral visualization analysis, this method is more quantitative and structured. The reference template serves as an ideal target envelope during training, guiding patients to imitate it in conjunction with rhythmic audio, thereby achieving visual training of muscle activity patterns. This mechanism effectively addresses the problems of poor guidance and vague goals in traditional rehabilitation, improving the directionality and repeatability of training, and providing a data-driven intervention pathway for the rehabilitation of swallowing disorders.

[0038] The process of collecting raw sound signals during the patient's swallowing process, performing drift removal and envelope extraction, and identifying swallowing events specifically includes: Collect patient raw signals Baseline set Fitting polynomial And obtain the drift signal Obtain the patient's clean signal and prepare for time-frequency analysis; Calculate patients With temporal envelope ; Use global threshold Location Swallowing Event Index Set Extract time-domain and frequency-domain feature windows of the patient's swallowing events.

[0039] This paper proposes collecting raw signals from patients during their swallowing training phase and using the same algorithm to perform drift removal and swallowing event recognition. This "patient-side signal processing module" mirrors the baseline data processing structure, enabling standardized application of the model algorithm and facilitating cross-user comparison and model transfer. This approach avoids algorithm mismatch issues in clinical or home use, improving the reliability of training and evaluation. Through standardized processing steps, users can complete the collection and analysis of swallowing signals without the need for specialized equipment or personnel, enhancing the system's usability in scenarios such as telemedicine and home rehabilitation.

[0040] The process involves calculating low-frequency and high-frequency energies based on a boundary index, obtaining the energy ratio, and performing a standard normal fraction calculation. Based on the score, patients are categorized into swallowing function classes, specifically including: Fixed Global Boundary Index Calculate the first one respectively Low-frequency energy and high-frequency energy of the second swallow: , Quantify the low / high frequency energy of each swallow by the patient; Calculate the patient's first Energy ratio per swallow ; Calculating the patient's standard normality Fraction ; Standardize the patient ratio to facilitate comparison with the control distribution; according to The score categorizes each event into swallowing function categories: ; in, For the first Functional categories of the first swallow; This is normal. Mild disability. The severity of the impairment is classified for each swallowing function to guide training needs.

[0041] By using a fixed global frequency band boundary index, the low-frequency and high-frequency energies of each swallowing event are calculated separately, and then their energy ratio is calculated and converted into a standard normal score. Subsequently, the functional status of each swallowing event is classified (normal, mild impairment, severe impairment) according to a set score range. This process quantifies swallowing behavior at the signal level and addresses the problem of large absolute differences between individuals and difficulties in cross-sectional comparisons through statistical normalization. This mechanism makes the swallowing assessment results highly interpretable and valuable, and also provides clear functional labels for subsequent training program design, filling the gap in the lack of standardized scales in the field of swallowing rehabilitation training.

[0042] The process of determining the training level based on the relative deviation between the patient's energy ratio and the offline baseline, and setting a target training ratio, specifically includes: Calculate the overall average ratio of patients ;in, The total number of swallowing events for the patient; a summary of the patient's overall swallowing ability indicators; in accordance with Set training level labels relative to offline benchmarks : ;in, No training required; For light training; The training intensity is classified as moderate to severe; the training intensity level is determined based on the difference compared to the control group. Set training target ratio : Set achievable personalized training goals; The training template is the reference envelope sequence. Provide patients with typical envelope audio and guide them in imitation.

[0043] Based on the difference between the patient's energy ratio and offline references, training levels are categorized (no need, mild, moderate to severe) and target energy ratios are set. Further integration with a reference envelope template provides patients with intuitive training goals. By introducing a tiered mechanism, training resources are allocated rationally, reducing unnecessary training burdens. Setting target ratios helps quantify "achievement" and "progress" during training, addressing the lack of feedback in traditional "sensory training," and improving the verifiability of rehabilitation effects and the patient's sense of purpose in training.

[0044] Based on the established training goals, patients undergo periodic training, and after a specific period, they conduct self-testing and training achievement assessments, specifically including: Set the number of swallows per group. Interval of swallowing movements Number of training sets per group Training cycle , Clearly define the training frequency and total volume plan; Set the length of the reference envelope audio for each group of playback. The patient swallows rhythmically. Secondly, by guiding students through "listening and doing," the energy output of the target frequency band is enhanced. Obtain the patient's first [test result] after the end of each weekly training day or after the end of 4 weeks of training. Self-test swallowing energy ratio And calculate the self-tested average ratio. ;in, To measure the total number of swallowing events; to evaluate training effectiveness and obtain the latest functional metrics; Computational training evaluation Determine whether the training goals have been achieved, and decide whether to continue training or advance to the next level; like If the training is successful, then the training is up to standard; if Continue repeating the training until .

[0045] A complete closed-loop rehabilitation pathway is proposed: patients perform imitation training at a set frequency and rhythm, and swallowing signals and energy ratio self-assessments are conducted at the end of each training cycle. Based on whether the self-assessed ratio meets the target, the decision is made whether to continue training or upgrade the training level, thus realizing an intelligent adaptive training plan. This mechanism solves the problem of the separation between training behavior and effect assessment in the rehabilitation cycle, and also addresses the pain point of "training without feedback" for patients, significantly improving their training compliance and enthusiasm, and propelling swallowing function rehabilitation into the era of "data-driven + adaptive" approaches.

[0046] This embodiment also provides a system for measuring and assessing swallowing dysfunction, including: The swallowing dysfunction measurement and assessment system includes: Offline data processing module: used for hardware setup and baseline data acquisition, signal de-drifting, time-frequency analysis and swallowing event localization, frequency band demarcation and energy calculation, ratio statistics and reference template construction; Online assessment module: used for patient-side signal acquisition and processing, and swallowing function assessment.

[0047] From a system perspective, the overall structure is summarized as "offline data processing module" and "online assessment module." The former is for professional institutions to establish a standard benchmark database, while the latter serves the patient's signal processing and assessment feedback. This structure ensures that the system can simultaneously support large-scale modeling training and individualized assessment training, supports centralized and distributed collaborative operation in deployment, enhances the system's engineering feasibility, provides technical support for telemedicine, and aligns with the development direction of intelligent rehabilitation platforms.

[0048] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0049] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A measurement and assessment system and training method for swallowing dysfunction, characterized in that, include: A skin microphone was placed next to the thyroid cartilage in the neck of the subject to collect the sound signal of the healthy control subject as baseline data. The reference signal is fitted with a quadratic polynomial to remove low-frequency drift components; The complex spectrum is calculated based on a preset window function, and swallowing event intervals are identified by combining envelope extraction and threshold setting. Based on the frequency boundary parameters, the low-frequency and high-frequency energies in swallowing events are statistically analyzed, and the energy baseline values ​​are determined. Calculate the energy ratio for each swallowing event and generate an average reference energy envelope; The system collects raw sound signals during the patient's swallowing process, performs drift removal and envelope extraction, and identifies swallowing events. Based on the boundary index, low-frequency and high-frequency energies are calculated separately to obtain the energy ratio and perform standard normal fraction calculation. Patient swallowing function categories are then classified according to the scores. The training level is determined based on the relative deviation of the patient's energy ratio from the offline baseline, and a target training ratio is set. Based on the set training goals, patients undergo periodic training and conduct self-testing and training achievement assessments after a specific period.

2. The swallowing dysfunction measurement and assessment system and training method according to claim 1, characterized in that, The step of placing a skin-based microphone next to the thyroid cartilage in the subject's neck to collect sound signals from healthy control subjects as baseline data specifically includes: Configure a skin-mounted microphone attached to the thyroid cartilage in the neck, and set the sampling frequency to [frequency value missing]. The sampling interval is ; right 100 healthy control subjects, respectively, with the same and duration The sound signal is collected in seconds, and the first sound signal is obtained. The first subject The original sound pressure value at each sampling time is , ;in, The total number of healthy control subjects; This represents the total number of sampling points; Index for subjects; Synchronization in a silent environment, for the first Set of baseline segment indexes for identified subjects: 。 3. The swallowing dysfunction measurement and assessment system and training method according to claim 2, characterized in that, The process of performing quadratic polynomial fitting on the reference signal to remove low-frequency drift components specifically includes: For the 10 subjects, at baseline The quadratic polynomial was fitted using the least squares method: , ;in, For the first Baseline drift estimates for 10 subjects; , , The fitting coefficients are obtained by solving the following normal equation: ; Index of fitting coefficients; The first after removing low-frequency drift was obtained The signal of the subjects was .

4. The swallowing dysfunction measurement and assessment system and training method according to claim 3, characterized in that, The process of calculating the complex spectrum based on a preset window function, and combining envelope extraction and threshold setting to identify swallowing event intervals, specifically includes: set up Window length is Window moved Set the parabolic window function as follows: , ;in, The coefficient of the parabolic window; Index for parabolic window coefficients; calculate Frame count ; Calculate the first The number of subjects in the first Frame, First Complex spectrum values ​​under frequency index : ； in, For frame indexing; Frequency index; Extract the first Temporal envelope of drift signal of subjects ; Set the global swallowing detection threshold to ;in, To find the median function, we take the median of a set of numbers arranged from smallest to largest. Find all the longest connected intervals: satisfy ;in, For the first The total number of swallowing events in the subjects; For the first Index of swallowing events of 10 subjects; For the first The start and end sampling point index of the event in the time sequence.

5. The swallowing dysfunction measurement and assessment system and training method according to claim 4, characterized in that, The process involves statistically analyzing the low-frequency and high-frequency energies in a swallowing event based on frequency boundary parameters, and determining an energy baseline value. Specifically, this includes: For the The subject, the first For this event, set the frequency boundary parameters: ； ； in, For the first The number of subjects in the first The next incident Low-frequency accumulated energy at the boundary; For the first The number of subjects in the first The next incident High-frequency accumulated energy at the boundary; Calculate the first The number of subjects in the first Local boundary index of the sub-event: ； For all The global bounding index is obtained by taking the median: , ;in, This corresponds to the global intermediate frequency; fixed Calculate the low-frequency energy reference and the high-frequency energy reference respectively: ， 。 6. The swallowing dysfunction measurement and assessment system and training method according to claim 5, characterized in that, The calculation of the energy ratio of each swallowing event and the generation of an average reference energy envelope specifically includes: Calculate the first The number of subjects in the first Acoustic energy ratio of swallowing ; Calculate the overall mean of all acoustic energy ratios. with standard deviation : ， ； Constructing the average reference envelope for training: ;in, For the first The average low-frequency energy envelope of the frame.

7. The swallowing dysfunction measurement and assessment system and training method according to claim 6, characterized in that, The process of collecting raw sound signals during the patient's swallowing process, performing drift removal and envelope extraction, and identifying swallowing events specifically includes: Collect patient raw signals Baseline set Fitting polynomial And obtain the drift signal ; Calculate patients With temporal envelope ; Use global threshold Location Swallowing Event Index Set .

8. The swallowing dysfunction measurement and assessment system and training method according to claim 7, characterized in that, The process involves calculating low-frequency and high-frequency energies based on a boundary index, obtaining the energy ratio, and performing a standard normal fraction calculation. Based on the score, patients are categorized into swallowing function classes, specifically including: Fixed Global Boundary Index Calculate the first one respectively Low-frequency energy and high-frequency energy of the second swallow: ， ； Calculate the patient's first Energy ratio per swallow ; Calculating the patient's standard normality Fraction ; according to The score categorizes each event into swallowing function categories: ； in, For the first Functional categories of the first swallow; This is normal. Mild disability. It is a severe disability.

9. The swallowing dysfunction measurement and assessment system and training method according to claim 8, characterized in that, The process of determining the training level based on the relative deviation between the patient's energy ratio and the offline baseline, and setting a target training ratio, specifically includes: Calculate the overall average ratio of patients ;in, The total number of swallowing events for the patient; in accordance with Set training level labels relative to offline benchmarks : ;in, No training required; For light training; This is moderate to heavy training. Set training target ratio : ； The training template is the reference envelope sequence. .

10. The swallowing dysfunction measurement and assessment system and training method according to claim 9, characterized in that, Based on the established training goals, patients undergo periodic training, and after a specific period, they conduct self-testing and training achievement assessments, specifically including: Set the number of swallows per group. Interval of swallowing movements Number of training sets per group Training cycle , ; Set the length of the reference envelope audio for each group of playback. The patient swallows rhythmically. Second-rate; Obtain the patient's first [test result] after the end of each weekly training day or after the end of 4 weeks of training. Self-test swallowing energy ratio And calculate the self-tested average ratio. ;in, This represents the total number of swallowing events detected through self-testing. Computational training evaluation ; like If the training is successful, then the training is up to standard; if Continue repeating the training until .

11. The swallowing dysfunction measurement and assessment system and training method according to claim 10, characterized in that, include: The swallowing dysfunction measurement and assessment system includes: Offline data processing module: used for hardware setup and baseline data acquisition, signal de-drifting, time-frequency analysis and swallowing event localization, frequency band demarcation and energy calculation, ratio statistics and reference template construction; Online assessment module: used for patient-side signal acquisition and processing, and swallowing function assessment.

Images