A child rhinitis nebulization treatment control system and method based on behavior guidance and intelligent monitoring

Abstract

The present application relates to the technical field of intelligent monitoring, in particular to a child rhinitis nebulization treatment control system and method based on behavior guidance and intelligent monitoring, the system comprising: a voice emotion analysis module, a respiratory airflow monitoring module, a rhythm feature quantification module, a respiratory behavior guidance module, and a nebulization drug release control module.In the present application, the internal emotional tendency is quantified by extracting the audio spectrum features to establish the classification boundary and determine the mapping relationship between the acoustic parameters and the state code, then the multi-dimensional respiratory airflow extreme value data is collected and the adjacent time stamp difference points are matched to calculate the duration of a specific stage, the deep breathing features are constructed, the rhythm deviation degree is fused with the reference amplitude to generate a compensation factor to update the visual animation track to realize dynamic intervention guidance, the guided parameters and the state code are spliced and fused, the rate adjustment instruction is generated according to the comprehensive judgment result, and the on-demand drug delivery for different deep and shallow breathing rhythms is realized to avoid drug liquid waste and greatly improve the absorption and conversion effect.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, and in particular to a control system and method for nebulized treatment of rhinitis in children based on behavior guidance and intelligent monitoring. Background Technology

[0002] The field of intelligent monitoring technology mainly involves collecting physiological and behavioral information of target objects through sensing devices, and continuously recording and judging the data in conjunction with embedded control devices, thereby realizing dynamic monitoring and process intervention for specific groups of people or scenarios. Its core aspects include the deployment and data acquisition of multi-source sensors, the identification and judgment rule setting of state characteristics, the signal transmission method between monitoring terminals and execution devices, and the response mechanism to changes in the behavior of monitored objects. This technology is widely used in scenarios such as medical assistance, rehabilitation management, and child care, and achieves process management and intervention control through synchronous observation of individual states and behaviors.

[0003] Traditional pediatric rhinitis nebulization therapy control systems based on behavior guidance and intelligent monitoring involve setting up a nebulizer to output a fixed flow rate of liquid aerosol medication during nebulization therapy. This is done in conjunction with a mask or nasal / oral interface for the child to inhale. A respiratory rate sensor is installed on the device to record the inhalation and exhalation rhythms. A camera captures changes in the child's facial orientation and movements. The nebulization output is controlled to start and stop based on preset time parameters and respiratory rhythm thresholds. A voice prompt plays a loop to guide the child to maintain the inhalation posture. The treatment duration and output intensity can be manually adjusted by the caregiver via buttons or a touch interface to complete the basic control and behavior guidance during the treatment process.

[0004] Existing technologies, in actual operation, set a fixed flow rate for outputting liquid aerosol and rely on manual parameter setting for start-stop control. At the same time, they guide the audience to cooperate through a single voice loop. This operating mode makes it difficult to accurately capture the individual's emotional fluctuations and changes in breathing rhythm during the treatment process. As a result, the fixed flow rate of liquid aerosol cannot be effectively matched with the variable breathing depth, leading to waste of liquid aerosol and low absorption efficiency. In addition, mechanical prompts are difficult to maintain focus and can easily cause anxiety. The static mechanism is difficult to dynamically adjust according to the real-time status, which seriously affects the overall intervention effect. Summary of the Invention

[0005] To address the technical problems of existing technologies that rely on manually setting parameters for starting and stopping medication aerosols at a fixed flow rate and using a single, repetitive voice prompt to guide the patient's response, this invention provides a nebulized treatment control system and method for children's rhinitis based on behavioral guidance and intelligent monitoring. These systems typically involve setting a fixed flow rate for medication aerosol output and relying on manual parameter settings for start-stop control. Furthermore, the mechanical prompts are difficult to maintain focus and can easily induce anxiety. The static mechanism also struggles to dynamically adjust based on real-time conditions, severely impacting the overall intervention effect.

[0006] On the one hand, a nebulized treatment control system for pediatric rhinitis based on behavior guidance and intelligent monitoring is provided, the system including: The voice emotion analysis module collects children's voice audio sequences, extracts the frequency cepstral coefficients of the children's voice audio sequences to establish an acoustic spectral feature set, and uses a support vector machine to calculate the distance between the acoustic spectral feature set and the state classification boundary to generate an emotion state code. The respiratory airflow monitoring module collects respiratory airflow sensing signals through a flow sensor, extracts the peak amplitude and duration of the respiratory airflow sensing signals to establish a deep and long breathing feature sequence, calculates the time interval between adjacent peaks within the deep and long breathing feature sequence, and obtains the duration of the inspiratory phase. The rhythm feature quantification module calculates the absolute value of the difference between the duration of the inhalation phase and the preset baseline duration threshold to obtain the duration deviation. It sets the mean amplitude of the deep and long breathing feature sequence as the baseline amplitude value. Based on the timestamp, it performs alignment and combination calculations on the duration deviation, the baseline amplitude value, and the emotional state code to construct a rhythm deviation feature matrix. The breathing behavior guidance module performs a weighted summation calculation on the distribution elements of the rhythm deviation feature matrix to obtain the behavior guidance compensation amount, and performs trajectory offset conversion on the visual animation coordinate nodes of the terminal display interface based on the behavior guidance compensation amount to generate dynamic guidance feedback parameters. The nebulized drug delivery control module concatenates and fuses the dynamic guidance feedback parameters with the emotional state code to construct a comprehensive state feature vector. It then calls the proportional-integral-differential algorithm to perform duty cycle mapping transformation calculation based on the comprehensive state feature vector, generating a nebulization rate adjustment command.

[0007] As a further aspect of the present invention, the emotional state encoding includes anger encoding, anxiety encoding, and calm encoding; the deep and long breathing feature sequence includes peak amplitude, duration, and time interval; the rhythm deviation feature matrix includes duration deviation, baseline amplitude value, and emotional state encoding; the dynamic guidance feedback parameters include coordinate offset, refresh frame rate, and guidance trajectory direction; and the nebulization rate adjustment command includes duty cycle value, nebulization frequency, and drug release sequence.

[0008] As a further aspect of the present invention, the voice emotion analysis module: The feature extraction submodule collects children's speech audio sequences, extracts instantaneous frequency components within the children's speech audio sequences, extracts frequency cepstral coefficients by performing logarithmic and cosine transforms on the instantaneous frequency components, merges the frequency cepstral coefficients, and establishes an acoustic spectrum feature set. The interval calculation submodule calls the acoustic spectrum feature set, maps the acoustic spectrum feature set into multi-dimensional feature points, obtains the state classification boundary hyperplane parameter, calculates the vertical normal distance from the multi-dimensional feature points to the plane corresponding to the state classification boundary hyperplane parameter, and generates the interval distance value. The state coding submodule calls the interval distance value, calculates the absolute difference between the interval distance value and the preset benchmark classification threshold, obtains the difference data volume, inputs the difference data volume into a discrete integer matrix to retrieve the corresponding identifier bit, and generates an emotion state code.

[0009] As a further aspect of the present invention, the respiratory airflow monitoring module includes: The signal acquisition submodule acquires respiratory airflow sensing signals through a flow sensor, identifies the extreme position coordinates and start and end time points of the respiratory airflow sensing signals, extracts the corresponding baseline peak amplitude and single ventilation time, and merges the baseline peak amplitude and single ventilation time in order to establish a deep and long breathing characteristic sequence. The interval analysis submodule calls the deep and long breathing feature sequence, iterates through the adjacent peak timestamps of the recorded elements in the deep and long breathing feature sequence one by one, calculates the absolute time difference component of the two timestamps, accumulates all the absolute time difference components in the group to define the span parameter, and generates the peak interval value. The duration calculation submodule compares the peak interval value with the inhalation cycle determination benchmark value, selects specific peak interval data segments that are less than the inhalation cycle determination benchmark value, and performs a summation operation on the span parameters corresponding to all specific peak interval data segments to obtain the duration of the inhalation phase.

[0010] As a further aspect of the present invention, the rhythm feature quantification module includes: The difference calculation submodule calls the duration of the inhalation phase, compares the duration of the inhalation phase with the preset reference duration threshold, calculates the duration difference between the duration of the inhalation phase and the preset reference duration threshold, extracts the absolute value of the duration difference, and generates the duration deviation. The benchmark extraction submodule, for the deep and long breathing feature sequence, sequentially traverses the peak amplitude elements inside the deep and long breathing feature sequence, accumulates the peak amplitude elements and divides them by the total number of peak amplitude elements to obtain the waveform average parameter, assigns a constant value to the waveform average parameter, and obtains the benchmark amplitude value. The matrix building submodule retrieves the duration deviation, the baseline amplitude value, and the emotion state code based on the recorded timestamp, aligns the time nodes of the duration deviation, the baseline amplitude value, and the emotion state code by column, and concatenates the various parameters to build a feature rhythm deviation matrix.

[0011] As a further aspect of the present invention, the preset reference duration threshold is a reference duration parameter obtained by statistically processing the historical data set of the duration of the inhalation phase corresponding to the deep and long breathing characteristic sequence according to the recorded timestamp order, and the reference duration parameter maintains a fixed value within the same sampling period.

[0012] As a further aspect of the present invention, the breathing behavior guidance module includes: The weighted operation submodule extracts the record distribution elements of each row and column inside the rhythm deviation feature matrix, sequentially assigns the distribution elements to the corresponding preset feature weight coefficients, and performs numerical multiplication calculation between the distribution elements and the preset feature weight coefficients to generate weighted distribution element values. The compensation extraction submodule calls the weighted distribution element value, extracts and filters all weighted distribution element values ​​located in the same row vector interval, performs an accumulation operation on all weighted distribution element values ​​after classification, and obtains the behavior guidance compensation amount. The offset conversion submodule, based on the behavior guidance compensation amount, collects the built-in visual animation coordinate nodes of the terminal display interface, performs coordinate conversion processing of the internal node trajectory offset of the visual animation coordinate nodes according to the deflection direction of the behavior guidance compensation amount, and generates dynamic guidance feedback parameters.

[0013] As a further aspect of the present invention, the atomized drug release control module includes: The feature fusion submodule extracts discrete numerical sequences of the dynamic guidance feedback parameters and the emotional state codes, aligns the dynamic guidance feedback parameters and the emotional state codes by columns, performs matrix row and column concatenation operations, and generates a comprehensive state feature vector. The duty cycle conversion submodule calls the comprehensive state feature vector, calculates the deviation ratio term, error integral term, and error differential term between the comprehensive state feature vector and the preset reference state vector, accumulates the deviation ratio term, error integral term, and error differential term to perform a linear duty cycle mapping transformation, and generates duty cycle mapping parameters. The adjustment control submodule, based on the duty cycle mapping parameter, sequentially compares the duty cycle mapping parameter with the preset operating limit threshold, selects the duty cycle mapping parameter that does not exceed the preset operating limit threshold, performs underlying format encoding conversion processing, and generates atomization rate adjustment command.

[0014] As a further aspect of the present invention, the preset benchmark state vector is based on a stable state data set generated by performing statistical mean calculation and standard deviation constraint filtering on the discrete numerical sequence of dynamic guidance feedback parameters and emotional state codes within the historical operating cycle, and the preset benchmark state vector is constructed after performing normalization processing on the stable state data set. The preset operating limit threshold is formed by dividing the range of the maximum and minimum values ​​of the duty cycle mapping parameter within the historical operating cycle, and then performing numerical convergence processing on the upper and lower boundaries of the range in combination with the preset safety margin.

[0015] On the other hand, a method for controlling nebulized rhinitis treatment in children based on behavior guidance and intelligent monitoring, wherein the method is executed based on the aforementioned nebulized rhinitis treatment control system based on behavior guidance and intelligent monitoring, includes the following steps: S1: Collect children's speech audio sequences, extract the frequency cepstral coefficients of the children's speech audio sequences to establish an acoustic spectrum feature set, use support vector machine to calculate the interval distance between the acoustic spectrum feature set and the state classification boundary, and generate emotional state codes; S2: Collect respiratory airflow sensing signals through a flow sensor, extract the peak amplitude and duration of the respiratory airflow sensing signals to establish a deep and long breathing feature sequence, calculate the time interval between adjacent peaks in the deep and long breathing feature sequence, and obtain the duration of the inspiratory phase. S3: Calculate the absolute value of the difference between the duration of the inhalation phase and the preset baseline duration threshold to obtain the duration deviation amount. Set the mean amplitude of the deep breathing feature sequence as the baseline amplitude value. Based on the timestamp, perform alignment and combination calculation of the duration deviation amount, the baseline amplitude value and the emotional state code to construct a rhythm deviation feature matrix. S4: Perform a weighted summation calculation on the distribution elements of the rhythm deviation feature matrix to obtain the behavior guidance compensation amount, and perform trajectory offset conversion on the visual animation coordinate nodes of the terminal display interface based on the behavior guidance compensation amount to generate dynamic guidance feedback parameters. S5: The dynamic guidance feedback parameters and the emotional state code are spliced ​​and fused to construct a comprehensive state feature vector. The proportional-integral-differential algorithm is called to perform duty cycle mapping transformation calculation based on the comprehensive state feature vector to generate atomization rate adjustment instructions.

[0016] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By extracting audio spectral features to establish classification boundaries and determine the mapping relationship between acoustic parameters and state codes, the intrinsic emotional tendencies can be accurately quantified. Then, multidimensional respiratory airflow extreme data are collected and adjacent timestamp differences are matched to calculate the duration of specific stages to construct deep and long breathing characteristics. The degree of rhythm deviation is fused with the baseline amplitude to generate compensation factors, thereby accurately updating the visual animation trajectory to achieve dynamic intervention guidance. At the same time, the guidance parameters and state codes are spliced ​​and fused to convert the mapping coefficients, thereby generating rate adjustment instructions based on the comprehensive judgment results. This enables precise on-demand drug delivery for different deep and shallow breathing rhythms, thereby avoiding drug waste and significantly improving absorption and conversion efficiency. Attached Figure Description

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

[0018] Figure 1 This is a system schematic diagram of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the speech emotion analysis module in this invention; Figure 4 This is a flowchart of the respiratory airflow monitoring module in this invention; Figure 5 This is a flowchart of the rhythm feature quantization module in this invention; Figure 6 This is a flowchart of the breathing behavior guidance module in this invention; Figure 7 This is a flowchart of the atomized drug release control module in this invention; Figure 8 This is a flowchart of the method of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0020] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0021] This invention provides a nebulized treatment control system for pediatric rhinitis based on behavior guidance and intelligent monitoring, such as... Figure 1-2 The diagram shown illustrates a nebulized treatment control system for pediatric rhinitis based on behavior guidance and intelligent monitoring. The system includes: The voice emotion analysis module collects children's voice audio sequences, extracts the frequency cepstral coefficients of the children's voice audio sequences to establish an acoustic spectral feature set, and uses a support vector machine to calculate the distance between the acoustic spectral feature set and the state classification boundary to generate an emotion state code. The respiratory airflow monitoring module collects respiratory airflow sensing signals through a flow sensor, extracts the peak amplitude and duration of the respiratory airflow sensing signals to establish a deep and long breathing feature sequence, calculates the time interval between adjacent peaks within the deep and long breathing feature sequence, and obtains the duration of the inspiratory phase. The rhythm feature quantification module calculates the absolute value of the difference between the duration of the inhalation phase and the preset baseline duration threshold to obtain the duration deviation. It sets the mean amplitude of the deep breathing feature sequence as the baseline amplitude value and performs alignment and combination calculations based on the timestamp, the duration deviation value, the baseline amplitude value, and the emotional state code to construct a rhythm deviation feature matrix. The breathing behavior guidance module performs a weighted summation calculation on the distribution elements of the rhythm deviation feature matrix to obtain the behavior guidance compensation amount. Based on the behavior guidance compensation amount, it performs trajectory offset conversion on the visual animation coordinate nodes of the terminal display interface to generate dynamic guidance feedback parameters. The nebulized drug delivery control module splices and fuses the dynamic guidance feedback parameters and emotional state codes to construct a comprehensive state feature vector. It then calls the proportional-integral-differential algorithm to perform duty cycle mapping transformation calculation based on the comprehensive state feature vector, generating nebulization rate adjustment instructions.

[0022] The emotional state encoding includes anger encoding, anxiety encoding, and calm encoding; the deep breathing feature sequence includes peak amplitude, duration, and time interval; the rhythm deviation feature matrix includes duration deviation, baseline amplitude value, and emotional state encoding; the dynamic guidance feedback parameters include coordinate offset, refresh rate, and guidance trajectory direction; and the nebulization rate adjustment command includes duty cycle value, nebulization frequency, and drug release sequence.

[0023] Specifically, such as Figure 2 , 3 As shown, the voice emotion analysis module: The feature extraction submodule collects children's speech audio sequences, extracts instantaneous frequency components within the children's speech audio sequences, extracts frequency cepstral coefficients by performing logarithmic and cosine transforms on the instantaneous frequency components, merges the frequency cepstral coefficients, and establishes an acoustic spectrum feature set. The feature extraction submodule calls the microphone array sound acquisition driver interface to continuously record the child's vocalizations before the atomization process at a fixed sampling frequency of 16000 Hz, generating an initial 30-second child speech audio sequence. After receiving this initial child speech audio sequence, the submodule performs pre-emphasis processing, subtracting the product of the previous sampling point value and the pre-emphasis coefficient (set to 0.97) from the current sampling point value to obtain a preprocessed speech sequence. A Hamming window with a length of 400 sampling points and a step size of 160 sampling points is used to perform frame-by-frame windowing processing on the preprocessed speech sequence, forming a continuous set of audio frames. The discrete time-domain data of each audio frame is extracted, and a Fast Fourier Transform is performed to convert the discrete time-domain data into complex frequency-domain data. The amplitude and phase components of this complex frequency-domain data are then extracted. Based on the quotient of the phase difference and time interval between adjacent frequency-domain sampling points, the instantaneous frequency components within the child's speech audio sequence are calculated and extracted. The instantaneous frequency component is input into a Mel filter bank consisting of 26 triangular bandpass filters. The energy components within each frequency band are multiplied and accumulated with the corresponding filter weights to obtain the energy distribution values ​​for each band. A base-10 logarithmic operation is performed on each band's energy distribution value to generate logarithmic energy values. A discrete cosine transform (DCT) is then performed on these logarithmic energy values. The logarithmic energy values ​​for each band are multiplied by the DCT transformation matrix and summed. The values ​​of the first 13 low-frequency dimensions are extracted as frequency cepstral coefficients. This transformation and extraction process is repeated for all audio frames, and the frequency cepstral coefficients of each audio frame are merged in chronological order. Taking the first audio frame as an example, the extracted instantaneous frequency energy distribution value for the first frequency band is 400 microwatts. With a preset logarithmic baseline of 10, the logarithm is 2.60. Multiplying the DCT coefficients by 2.60 (0.8) yields the first frequency cepstral coefficient of 2.08. Summarize all frequency cepstral coefficients corresponding to a 30-second audio clip and output the final set of acoustic spectral features.

[0024] The interval calculation submodule calls the acoustic spectrum feature set, maps the acoustic spectrum feature set into multi-dimensional feature points, obtains the state classification boundary hyperplane parameters, calculates the vertical normal distance from the multi-dimensional feature points to the plane corresponding to the state classification boundary hyperplane parameters, and generates the interval distance value. The interval calculation submodule calls the acoustic spectrum feature set and maps the two-dimensional frequency cepstral coefficients in the acoustic spectrum feature set into multi-dimensional feature points through the radial basis kernel function. These multi-dimensional feature points contain coordinate values ​​in 300 dimensions. It retrieves the state classification boundary hyperplane parameters obtained by training historical normal breathing audio features and crying audio features under a support vector classification architecture. These hyperplane parameters contain normal vector values ​​in 300 dimensions and one intercept parameter. It reads the coordinate values ​​of each dimension of the multi-dimensional feature point item by item, performs multiplication calculations with the corresponding dimension parameters in the normal vector values, accumulates all product results to obtain the normal vector inner product sum, adds the intercept parameter to this sum of normal vector inner products, and obtains the absolute projection value of the multi-dimensional feature point in the normal vector direction. It calculates and accumulates the square values ​​of all dimensions of the normal vector parameters, and performs a square root operation on the accumulated sum of squares to obtain the normal vector magnitude. Divide the absolute projection value by the normal vector magnitude to calculate the perpendicular normal distance from the multidimensional feature point to the plane corresponding to the state classification boundary hyperplane parameter. Assign this distance value to the data storage node. Assume that the multidimensional feature points of a certain feature frame are extracted into three key dimension coordinates: 1.5, 2.0, and 0.8. The corresponding normal vector values ​​in the obtained state classification boundary hyperplane parameters are 0.6, 0.8, and 0.0, with an intercept parameter of -1.2. Multiply 1.5 by 0.6 to get 0.9, multiply 2.0 by 0.8 to get 1.6, and multiply 0.8 by 0.0 to get 0.0. Add 0.9, 1.6, and 0.0 to get the normal vector dot product sum of 2.5. Add the dot product sum of 2.5 to the intercept parameter -1.2 to get the absolute projection value of 1.3. Square the normal vector values ​​0.6 and 0.8 respectively to get 0.36 and 0.64, sum them to get 1.0, and take the square root of the result. The normal vector magnitude is 1.0. Dividing the absolute projection value of 1.3 by the modulus of 1.0, the final interval distance value is 1.3.

[0025] The state coding submodule calls the interval distance value, calculates the absolute difference between the interval distance value and the preset baseline classification threshold, obtains the difference data volume, inputs the difference data volume into a discrete integer matrix to retrieve the corresponding identifier bit, and generates the emotion state code. The state coding submodule calls the interval distance value and reads the pre-configured preset baseline classification threshold from internal storage. The preset baseline classification threshold is calculated by adding twice the standard deviation to the arithmetic mean of the interval distance values ​​of the previous 1000 historical stable audio data, and its fixed value is set to 1.0. It obtains the interval distance value and the preset baseline classification threshold, subtracts the preset baseline classification threshold from the interval distance value to obtain the difference value, and takes the absolute value of the difference value to obtain the difference data volume. The state coding submodule has a built-in two-dimensional discrete integer matrix, which is divided into 4 data interval columns and corresponding identifier rows. It compares the difference data volume with the upper and lower limits of each interval in the matrix, and inputs the difference data volume into the discrete integer matrix to retrieve the corresponding identifier bit matching its interval. When the difference in data volume is less than 0.5, the system locates the first flag bit and outputs 1; when the difference in data volume is greater than or equal to 0.5 and less than 1.2, the system locates the second flag bit and outputs 2; when the difference in data volume is greater than or equal to 1.2 and less than 2.5, the system locates the third flag bit and outputs 3; and when the difference in data volume is greater than or equal to 2.5, the system locates the fourth flag bit and outputs 4. The extracted flag bit values ​​are combined into an 8-bit binary data stream. In the previous example, the interval distance value was 1.3, and the preset baseline classification threshold was 1.0. Subtracting 1.0 from 1.3 yields a difference value of 0.3. Taking the absolute value gives the difference in data volume as 0.3. The difference in data volume 0.3 is input into a discrete integer matrix for lookup. Since 0.3 is less than 0.5, the system locates the first flag bit, finding a corresponding flag bit value of 1. The value 1 is converted into the corresponding binary data, generating the emotion state code 00000001.

[0026] Specifically, such as Figure 2 , 4 As shown, the respiratory airflow monitoring module includes: The signal acquisition submodule acquires respiratory airflow sensing signals through a flow sensor, identifies the extreme position coordinates and start and end time points of the respiratory airflow sensing signals, extracts the corresponding baseline peak amplitude and single ventilation time, and merges the baseline peak amplitude and single ventilation time in order to establish a deep and long breathing characteristic sequence. The signal acquisition submodule continuously detects changes in airflow temperature using a miniature hot-wire flow sensor connected to the front of the breathing mask, thereby converting these changes into respiratory airflow sensing signals. The sensor acquires 200 analog signal values ​​per second, which are then converted into discrete digital airflow sequences using an analog-to-digital converter. A sliding window with five sampling points is used to traverse this sequence, comparing the values ​​at the center point with those of the adjacent points on either side. When the center point value is simultaneously greater than all node values ​​on both the left and right sides and greater than a preset baseline noise level of 0.2 liters per minute, it is identified as the peak extreme value of the respiratory airflow sensing signal. Based on this peak extreme value coordinate, a first derivative is calculated to the left and right. When the absolute value of the derivative first falls below 0.05 liters per square minute, that point is defined as a trough node, thus extracting the start and end times of a single breath waveform. The signal acquisition submodule uses the flow velocity value at the peak as the baseline peak amplitude to extract data nodes, and subtracts the start time from the end time to obtain the single breath exchange time. After calculating 15 consecutive breaths within the current test cycle, the extracted baseline peak amplitude and single-breath time are sequentially combined and merged. Assuming the peak is identified at 250 milliseconds using a sliding window in the first breath, the corresponding flow rate is recorded as 3.5 liters per minute, which is used as the baseline peak amplitude. The signal acquisition submodule calculates derivatives to the left and right, locking the start time at 100 milliseconds and the end time at 600 milliseconds. Subtracting 100 from 600 yields a breath time of 500 milliseconds. Subsequently, in the second breath, the peak corresponds to a flow rate of 4.1 liters per minute, with a single-breath time of 650 milliseconds. Following the chronological order, 3.5 and 500 are merged as the first element, and 4.1 and 650 are merged as the second element, and these are sequentially accumulated and entered into an array to establish a deep and long breathing feature sequence.

[0027] The interval analysis submodule calls the deep and long breathing feature sequence, iterates through the adjacent peak timestamps of the recorded elements in the deep and long breathing feature sequence one by one, calculates the absolute time difference component of the two timestamps, accumulates all the absolute time difference components in the group to define the span parameter, and generates the peak interval value. The interval analysis submodule calls the deep and long breathing feature sequence and reads all data representing the time of peak occurrence from the record elements within the sequence as peak timestamps. An index pointer is set, and the adjacent peak timestamps of the record elements within the deep and long breathing feature sequence are traversed item by item. For the nth and n+1th records in the sequence, the peak timestamp of the latter record is subtracted from the peak timestamp of the former record to obtain the difference. The absolute value of this difference is then calculated to obtain the absolute time difference component. The 10 consecutive absolute time difference components within the current analysis period are assigned to the same accumulation group. All absolute time difference components within this group are summed to obtain the total difference of the group. This total difference is then divided by the number of components involved in the accumulation, 10, to define the span parameter. This span parameter is directly assigned to the corresponding peak interval numerical data bits. Assume that the peak timestamps of the first three breaths in the deep and long breathing feature sequence are 250 milliseconds, 1200 milliseconds, and 2350 milliseconds, respectively. The interval analysis submodule iterates through the first and second items, subtracting 250 from 1200 to obtain a difference of 950 milliseconds. Taking the absolute value, we get the first absolute time difference component, which is 950 milliseconds. Continuing to iterate through the second and third items, subtracting 1200 from 2350 to obtain a difference of 1150 milliseconds. Taking the absolute value, we get the second absolute time difference component, which is also 1150 milliseconds. Assuming the current accumulation group only contains these two absolute time difference components for calculation, adding 1150 to 950 gives a total group difference of 2100 milliseconds. Dividing 2100 by the number of items involved in the accumulation (2) yields a span parameter of 1050 milliseconds. The value 1050 is then written to a register, generating the final peak interval value of 1050 milliseconds.

[0028] The duration calculation submodule compares the peak interval value with the inhalation cycle determination benchmark value, selects specific peak interval data segments that are less than the inhalation cycle determination benchmark value, and performs a summation operation on the span parameters corresponding to all specific peak interval data segments to obtain the duration of the inhalation phase. The duration calculation submodule retrieves the peak interval values ​​and reads the inspiratory cycle determination benchmark value from the built-in calibration storage unit. This inspiratory cycle determination benchmark value is calculated and set based on the standard age-appropriate children's quiet breathing rate, and here it is set to 1500 milliseconds. For the record array containing 20 peak interval values, a comparison loop instruction is established to compare the size relationship between each peak interval value and the inspiratory cycle determination benchmark value. When the peak interval value is less than the inspiratory cycle determination benchmark value of 1500 milliseconds, the peak interval value and its index position in the array are extracted as a specific peak interval data segment. When the peak interval value is greater than or equal to the inspiratory cycle determination benchmark value, the value is discarded and not extracted. After traversing all data, the span parameter corresponding to all selected specific peak interval data segments is summarized. A summation operation is performed on the span parameter to calculate the total value, and the total value result is assigned to the inspiratory phase duration node. Assume the input duration calculation submodule records three peak interval values ​​in its array: 1050 milliseconds, 1600 milliseconds, and 1250 milliseconds. The baseline value for determining the inhalation cycle is 1500 milliseconds. The first peak interval value of 1050 is compared; since 1050 is less than 1500, it is selected as the first specific peak interval data segment. The second value of 1600 is compared; since 1600 is greater than 1500, it is not extracted. The third value of 1250 is compared; since 1250 is less than 1500, it is selected as the second specific peak interval data segment. The selected values ​​of 1050 and 1250 are summed, resulting in 2300 milliseconds. The duration calculation submodule then determines the inhalation phase duration as 2300 milliseconds.

[0029] Table 1: Historical Statistical Table of Respiratory Duration

[0030] Table 1 shows the parameter values ​​for each acquisition time point in the historical dataset of some deep and long breathing feature sequences.

[0031] Specifically, such as Figure 2 , 5 As shown, the rhythm feature quantification module includes: The difference calculation submodule calls the inhalation phase duration, compares the inhalation phase duration with the preset baseline duration threshold, calculates the duration difference between the inhalation phase duration and the preset baseline duration threshold, extracts the absolute value of the duration difference, and generates the duration deviation. The difference calculation submodule calls the inspiratory phase duration function and retrieves a preset baseline duration threshold from the system's global variable cache. The preset baseline duration threshold is obtained by statistically processing historical data sets of inspiratory phase durations corresponding to deep breathing characteristic sequences, ordered by record timestamps. The process involves arithmetically summing the duration parameters of the first 50 stable states in the historical data shown in Table 1 and dividing by 50 to obtain the average value. This average value serves as the baseline duration parameter, which remains fixed within the same 5-minute sampling period; assuming the calculated value is 1200 milliseconds. The currently acquired inspiratory phase duration is retrieved and compared sequentially with the preset baseline duration threshold of 1200 milliseconds. The inspiratory phase duration is subtracted from the preset baseline duration threshold to calculate the duration difference. The absolute value of this duration difference is extracted. Assuming the current inspiratory phase duration called by the difference calculation submodule is 1050 milliseconds and the preset baseline duration threshold is fixed at 1200 milliseconds... Subtracting 1200 from 1050 yields a duration difference of -150 milliseconds. Extracting the absolute value of -150 gives 150 milliseconds. This 150 milliseconds is the duration deviation.

[0032] The baseline extraction submodule, for the deep and long breathing feature sequence, sequentially traverses the peak amplitude elements within the deep and long breathing feature sequence, accumulates the peak amplitude elements and divides them by the total number of peak amplitude elements to obtain the waveform average parameter, assigns a constant value to the waveform average parameter, and obtains the baseline amplitude value. The benchmark extraction submodule reads the internal storage area of ​​the deep and long breathing feature sequence and iterates through all peak amplitude elements within the sequence using a loop index instruction. An accumulator register with an initial value of 0 is set. During the traversal, the airflow value corresponding to each extracted peak amplitude element is added up progressively, and the total cumulative waveform amplitude value is obtained after accumulating the peak amplitude elements. The total number of peak amplitude elements in the current sequence is counted, and the total cumulative waveform amplitude value is divided by the total number of peak amplitude elements to obtain the waveform average parameter. The benchmark extraction submodule disconnects the time-dependent data connection of this waveform average parameter, staticizes it, assigns a constant value to the waveform average parameter, and stores it in a read-only register to obtain the benchmark amplitude value. Assuming the deep and long breathing feature sequence contains four peak amplitude elements with recorded airflow values ​​of 3.0 liters per minute, 3.2 liters per minute, 2.8 liters per minute, and 3.4 liters per minute, the benchmark extraction submodule iterates through these four values ​​and accumulates them. Adding 3.0 to 3.2 gives 6.2, adding 2.8 gives 9.0, and adding 3.4 gives a total cumulative waveform amplitude of 12.4 liters per minute. The total number of elements is 4. Dividing 12.4 by 4 gives an average waveform parameter of 3.1 liters per minute. Setting this value of 3.1 liters per minute as a constant and cutting off subsequent data refresh successfully obtains the baseline amplitude value of 3.1 liters per minute.

[0033] The matrix building submodule retrieves the duration deviation, baseline amplitude value, and emotional state code based on the recorded timestamp, aligns the time nodes of the duration deviation, baseline amplitude value, and emotional state code by column, and concatenates the various parameters to build a feature rhythm deviation matrix. The matrix construction submodule reads the timestamp synchronization pulse signal from the central control clock. Based on this recorded timestamp, it retrieves the duration deviation from the data bus in parallel, calculating the duration deviation, the baseline amplitude from the baseline extraction submodule, and the emotional state code from the state coding submodule. The matrix construction submodule constructs a three-column, multi-row blank two-dimensional array structure. The duration deviation is assigned to column 1, the baseline amplitude to column 2, and the emotional state code, parsed into decimal value, is assigned to column 3. Using the millisecond value of the timestamp as the row index, the time nodes of the duration deviation, baseline amplitude, and emotional state code are aligned column-wise. For data items at the same time node, each parameter is concatenated to create a row vector, which is then filled into the blank two-dimensional array row by row as the timestamp progresses. For example, at the 1000th millisecond of the recorded timestamp, the decimal value corresponding to a duration deviation of 150, a baseline amplitude of 3.1, and an emotional state code of 00000001 is retrieved. Fill 150 into the first column of the current row, 3.1 into the second column, and 1 into the third column, concatenating them to form a row vector. At the 2000-millisecond node, the duration deviation is found to be 180, the baseline amplitude value is 3.1 (remaining constant), and the emotional state is encoded as the value 2. Fill these values ​​into columns 1 to 3 of the second row, respectively. After concatenation, output the feature rhythm deviation matrix containing the above multiple rows of data.

[0034] Specifically, such as Figure 2 , 6 As shown, the breathing behavior guidance module includes: The weighted operation submodule extracts the distribution elements of each row and column of the rhythm deviation feature matrix, assigns the corresponding preset feature weight coefficients to the distribution elements in sequence, and performs numerical multiplication calculation between the distribution elements and the preset feature weight coefficients to generate weighted distribution element values. The weighted operation submodule retrieves the feature rhythm deviation matrix and extracts the distribution elements of each row and column within it. It then sequentially assigns preset feature weight coefficients to the distribution elements from a pre-defined weight configuration table. Specifically, the preset feature weight coefficient is set to 0.5 for the distribution element in column 1 representing the duration deviation; 0.3 for the distribution element in column 2 representing the baseline amplitude; and 0.2 for the distribution element in column 3 representing the emotion state encoding. The distribution elements in each column are then multiplied by the assigned preset feature weight coefficients, and the product is filled into a new data space according to the original matrix coordinates. For example, if the weighted operation submodule extracts the distribution elements from the first row of the feature rhythm deviation matrix, where the value in column 1 is 150, column 2 is 3.1, and column 3 is 1, then preset feature weight coefficients of 0.5, 0.3, and 0.2 are assigned to them respectively. For the first column's distribution element 150, multiply 150 by 0.5 and perform a numerical product calculation to obtain 75. For the second column's distribution element 3.1, multiply it by 0.3 to obtain 0.93. For the third column's distribution element 1, multiply it by 0.2 to obtain 0.2. The weighted operation submodule stores the calculated 75, 0.93, and 0.2 as weighted distribution element values ​​in the corresponding positions.

[0035] The compensation extraction submodule calls the weighted distribution element value, extracts and filters all weighted distribution element values ​​located in the same row vector interval, performs an accumulation operation on all weighted distribution element values ​​after classification, and obtains the behavior guidance compensation amount. The compensation extraction submodule calls the weighted distribution element values ​​to extract all numerical items within the newly generated matrix space. Using row index as the classification basis, the submodule filters all weighted distribution element values ​​located within the same row vector interval (i.e., the same timestamp). It calls the adder to perform a cumulative operation on all weighted distribution element values ​​within the same row after classification, adding the values ​​calculated from different columns to obtain the comprehensive sum for that row. This comprehensive sum is directly output to obtain the behavior guidance compensation amount. Combining the weighted distribution element values ​​generated in the above steps, the submodule calls the weighted distribution element values ​​of the first row vector interval, extracting all weighted distribution element values ​​within that same row vector interval: 75, 0.93, and 0.2. These three values ​​are then cumulatively added: 75 plus 0.93 equals 75.93, and 75.93 plus 0.2 equals 76.13. After calculation, the comprehensive sum, 76.13, is used as the behavior guidance compensation amount corresponding to the current time node. The advantage of this operational logic is that it collapses multi-dimensional weighted discrete indices into a single control dimension variable by accumulating the same row vectors, which greatly simplifies the complexity of the controller's input interface.

[0036] The offset conversion submodule, based on the behavior guidance compensation amount, collects the built-in visual animation coordinate nodes in the terminal display interface, performs coordinate conversion of the internal node trajectory offset of the visual animation coordinate nodes according to the deflection direction of the behavior guidance compensation amount, and generates dynamic guidance feedback parameters. The offset conversion submodule, based on the behavior guidance compensation amount, collects the built-in visual animation coordinate nodes of the terminal display interface directly in front of the child through a data interface. It determines the deflection direction and movement step size according to the positive / negative characteristics and magnitude of the acquired behavior guidance compensation amount. Using a constant of 5 as a conversion factor, the behavior guidance compensation amount is multiplied by this factor to obtain the offset of the node trajectory within the Cartesian coordinate system. This offset is directly added to the initial horizontal coordinate of the visual animation coordinate node, and the coordinate conversion processing of the node trajectory offset within the Cartesian coordinate system is performed, encapsulating the offset horizontal and vertical coordinate values ​​into a structure. Assuming the offset conversion submodule collects the initial Cartesian coordinates of the visual animation coordinate node numbered 1 in Table 2, with a horizontal value of 500 pixels and a vertical value of 400 pixels, and the received behavior guidance compensation amount of 76.13, multiplying 76.13 by the constant 5 yields a node trajectory offset of 380.65 pixels. Since this value is positive, the deflection direction is defined as a rightward offset. The initial horizontal value of 500 is increased by an offset of 380.65, and a coordinate transformation is performed to obtain a new horizontal coordinate value of 880.65 pixels. The vertical coordinate remains unchanged at 400 pixels, and the values ​​of 880.65 and 400 are combined and packaged to generate dynamic guidance feedback parameters.

[0037] Table 2: Correspondence Table of Screen Animation Coordinate Transformation Parameters

[0038] Table 2 lists the initial parameter settings for the coordinate nodes used for screen guidance feedback display.

[0039] Specifically, such as Figure 2 , 7 As shown, the atomized drug release control module includes: The feature fusion submodule extracts discrete numerical sequences of dynamic guidance feedback parameters and emotional state codes, aligns the dynamic guidance feedback parameters and emotional state codes by columns, performs matrix row and column concatenation operations, and generates a comprehensive state feature vector. The feature fusion submodule establishes a data extraction channel for the dynamic guidance feedback parameters generated in the preceding steps and the emotional state code provided by the state coding submodule. It extracts data from the horizontal and vertical coordinates of the dynamic guidance feedback parameter structure and extracts the corresponding decimal state bit data from the emotional state code. These are then converted to floating-point numbers to form discrete numerical sequences of the dynamic guidance feedback parameters and emotional state codes. Following the sampling time-series pulse train to which the data belongs, the discrete numerical sequences corresponding to the dynamic guidance feedback parameters and emotional state codes are aligned column-wise. The feature fusion submodule uses the two data columns containing the dynamic guidance feedback parameters as the left matrix and the one data column of the emotional state code as the right matrix, performing a matrix row and column concatenation operation horizontally to fuse the original sequences of different data dimensions into a single array containing three columns of feature attributes. Assume that the discrete numerical sequence of the dynamic guidance feedback parameters extracted by the feature fusion submodule at the current moment has a horizontal coordinate of 880.65 and a vertical coordinate of 400.0. The extracted discrete numerical sequence of the emotional state code has a state bit value of 1.0. Align the data column-wise, using 880.65 as the first column element, 400.0 as the second column element, and 1.0 as the third column element. Perform a matrix row-column concatenation operation on these three discrete numerical sequence elements, placing them in the same row vector. After concatenation, a comprehensive state feature vector is generated, which contains an array of the three elements: 880.65, 400.0, and 1.0.

[0040] The duty cycle conversion submodule calls the comprehensive state feature vector, calculates the deviation ratio, error integral, and error differential terms between the comprehensive state feature vector and the preset reference state vector, accumulates the deviation ratio, error integral, and error differential terms to perform a linear duty cycle mapping transformation, and generates the duty cycle mapping parameters. The duty cycle conversion submodule calls the parameter representing the lateral feedback offset in the first column of the comprehensive state feature vector, and simultaneously calls the internally stored preset baseline state vector. This preset baseline state vector is calculated by performing a statistical mean operation on the discrete numerical sequences of the dynamic guidance feedback parameters and emotion state codes over the past 10-day historical operating cycle. The average parameter is obtained by summing the first column elements of all historical sequences and dividing by the number of elements, resulting in an average parameter of 500.0. Abnormal data deviating from the average parameter by more than two standard deviations are removed, and standard deviation constraints are applied to generate a stable state data set. The average parameter in this stable state data set is then normalized by dividing by the maximum display width of 1000 to construct the preset baseline state vector. Here, the first column of the baseline parameter is assigned a value of 0.5. The same normalization operation is performed by dividing the first column parameter of the current comprehensive state feature vector by 1000 to obtain the measured parameter. The measured parameter is subtracted from the baseline parameter and then multiplied by the proportional gain coefficient 2.0 to calculate the deviation ratio between the comprehensive state feature vector and the preset baseline state vector. The duty cycle conversion submodule accumulates the differences between the measured parameters and the reference parameters from each iteration, multiplies this sum by an integral gain coefficient of 0.1, and calculates the integral error term. It then subtracts the previous difference from the current difference and multiplies this sum by a differential gain coefficient of 0.5 to calculate the differential error term. The cumulative deviation proportional term, integral error term, and differential error term yield the total control output value. A linear duty cycle mapping transformation is performed on this total value, limiting it to between 0 and 100. Assuming the first column of the comprehensive state feature vector has a horizontal coordinate parameter of 880.65, dividing by 1000 gives a measured parameter of 0.88065. The preset reference state vector corresponds to a reference parameter of 0.5 in its first column. Subtracting 0.5 from 0.88065 yields a deviation difference of 0.38065. Multiplying 0.38065 by a proportional gain coefficient of 2.0 gives a deviation proportional term of 0.7613. Assuming the cumulative difference is 1.5, multiplying it by 0.1 yields an integral error term of 0.15. Assuming the current difference is 0.38065, subtracting the previous difference of 0.3 gives 0.08065, multiplying this by 0.5 yields a differential error term of 0.040325. The duty cycle conversion submodule adds these three terms, resulting in 0.7613 plus 0.15 plus 0.040325, yielding 0.951625. The duty cycle conversion submodule then multiplies 0.951625 by the scaling factor of 100, generating the final duty cycle mapping parameter of 95.16.

[0041] The adjustment and control submodule, based on the duty cycle mapping parameter, compares the duty cycle mapping parameter with the preset operating limit threshold in turn, selects the duty cycle mapping parameter that does not exceed the preset operating limit threshold, performs low-level format encoding conversion processing, and generates atomization rate adjustment command. The regulation and control submodule obtains the duty cycle mapping parameter based on the system transmission bus and retrieves the preset operating limit threshold from the internal parameter library. The preset operating limit threshold is derived by dividing the duty cycle mapping parameter into intervals based on the maximum and minimum values ​​over the past 6 months of historical operating cycles. These intervals are divided into a basic operating range of 20 to 80 and a high-speed operating range of 80 to 95. A preset safety margin constant of 2.0 is used to converge the upper and lower boundaries of these intervals, ultimately setting the upper safety threshold to 93.0. The duty cycle mapping parameter is then compared to this upper safety threshold. When the duty cycle mapping parameter is less than or equal to the preset operating limit threshold, it is deemed qualified, and the duty cycle mapping parameter that does not exceed the preset operating limit threshold is selected. The corresponding decimal value of the duty cycle mapping parameter is then converted into a binary high / low level control bit sequence according to the pulse width modulation protocol standard. Assume the current value of the duty cycle mapping parameter is 85.5, and the upper safety threshold of the preset operating limit threshold is 93.0. The adjustment control submodule compares 85.5 with 93.0 and determines that 85.5 is less than 93.0. It then filters out the duty cycle mapping parameter 85.5, which does not exceed the preset operating threshold. This value is multiplied by the microprocessor's internal resolution constant of 10 to obtain 855. A low-level format encoding conversion process is then performed on 855, converting the decimal value 855 into a binary sequence stream recognizable by the microcontroller's timer comparator register. This generates the final atomization rate adjustment command, which is then input to the hardware driver.

[0042] Please see Figure 8 The method for controlling nebulized rhinitis treatment in children based on behavior guidance and intelligent monitoring is implemented based on the aforementioned nebulized rhinitis treatment control system for children based on behavior guidance and intelligent monitoring, and includes the following steps: S1: Collect children's speech audio sequences, extract the frequency cepstral coefficients of the children's speech audio sequences to establish an acoustic spectrum feature set, use support vector machine to calculate the interval distance between the acoustic spectrum feature set and the state classification boundary, and generate emotional state codes; S2: Collect respiratory airflow sensing signals through a flow sensor, extract the peak amplitude and duration of the respiratory airflow sensing signals to establish a deep and long breathing feature sequence, calculate the time interval between adjacent peaks in the deep and long breathing feature sequence, and obtain the duration of the inspiratory phase. S3: Calculate the absolute value of the difference between the duration of the inhalation phase and the preset baseline duration threshold to obtain the duration deviation. Set the mean amplitude of the deep breathing feature sequence as the baseline amplitude value. Based on the timestamp, perform alignment and combination calculations on the duration deviation, the baseline amplitude value, and the emotional state code to construct a rhythm deviation feature matrix. S4: Perform a weighted summation calculation on the distribution elements of the rhythm deviation feature matrix to obtain the behavior guidance compensation amount. Based on the behavior guidance compensation amount, perform trajectory offset conversion on the visual animation coordinate nodes of the terminal display interface to generate dynamic guidance feedback parameters. S5: The dynamic guidance feedback parameters and emotional state codes are spliced ​​and fused to construct a comprehensive state feature vector. The proportional-integral-differential algorithm is called to perform duty cycle mapping transformation calculation based on the comprehensive state feature vector to generate atomization rate adjustment instructions.

[0043] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the described technical solutions.

Claims

1. A nebulized treatment control system for pediatric rhinitis based on behavior guidance and intelligent monitoring, characterized in that, The system includes: The voice emotion analysis module collects children's voice audio sequences, extracts the frequency cepstral coefficients of the children's voice audio sequences to establish an acoustic spectral feature set, and uses a support vector machine to calculate the distance between the acoustic spectral feature set and the state classification boundary to generate an emotion state code. The respiratory airflow monitoring module collects respiratory airflow sensing signals through a flow sensor, extracts the peak amplitude and duration of the respiratory airflow sensing signals to establish a deep and long breathing feature sequence, calculates the time interval between adjacent peaks within the deep and long breathing feature sequence, and obtains the duration of the inspiratory phase. The rhythm feature quantification module calculates the absolute value of the difference between the duration of the inhalation phase and the preset baseline duration threshold to obtain the duration deviation. It sets the mean amplitude of the deep and long breathing feature sequence as the baseline amplitude value. Based on the timestamp, it performs alignment and combination calculations on the duration deviation, the baseline amplitude value, and the emotional state code to construct a rhythm deviation feature matrix. The breathing behavior guidance module performs a weighted summation calculation on the distribution elements of the rhythm deviation feature matrix to obtain the behavior guidance compensation amount, and performs trajectory offset conversion on the visual animation coordinate nodes of the terminal display interface based on the behavior guidance compensation amount to generate dynamic guidance feedback parameters. The nebulized drug delivery control module concatenates and fuses the dynamic guidance feedback parameters with the emotional state code to construct a comprehensive state feature vector. It then calls the proportional-integral-differential algorithm to perform duty cycle mapping transformation calculation based on the comprehensive state feature vector, generating a nebulization rate adjustment command.

2. The pediatric rhinitis nebulization treatment control system based on behavior guidance and intelligent monitoring according to claim 1, characterized in that, The emotional state encoding includes anger encoding, anxiety encoding, and calm encoding; the deep and long breathing feature sequence includes peak amplitude, duration, and time interval; the rhythm deviation feature matrix includes duration deviation, baseline amplitude value, and emotional state encoding; the dynamic guidance feedback parameters include coordinate offset, refresh frame rate, and guidance trajectory direction; and the nebulization rate adjustment command includes duty cycle value, nebulization frequency, and drug release sequence.

3. The pediatric rhinitis nebulization treatment control system based on behavior guidance and intelligent monitoring according to claim 1, characterized in that, The voice emotion analysis module: The feature extraction submodule collects children's speech audio sequences, extracts instantaneous frequency components within the children's speech audio sequences, extracts frequency cepstral coefficients by performing logarithmic and cosine transforms on the instantaneous frequency components, merges the frequency cepstral coefficients, and establishes an acoustic spectrum feature set. The interval calculation submodule calls the acoustic spectrum feature set, maps the acoustic spectrum feature set into multi-dimensional feature points, obtains the state classification boundary hyperplane parameter, calculates the vertical normal distance from the multi-dimensional feature points to the plane corresponding to the state classification boundary hyperplane parameter, and generates the interval distance value. The state coding submodule calls the interval distance value, calculates the absolute difference between the interval distance value and the preset benchmark classification threshold, obtains the difference data volume, inputs the difference data volume into a discrete integer matrix to retrieve the corresponding identifier bit, and generates an emotion state code.

4. The pediatric rhinitis nebulization treatment control system based on behavior guidance and intelligent monitoring according to claim 1, characterized in that, The respiratory airflow monitoring module includes: The signal acquisition submodule acquires respiratory airflow sensing signals through a flow sensor, identifies the extreme position coordinates and start and end time points of the respiratory airflow sensing signals, extracts the corresponding baseline peak amplitude and single ventilation time, and merges the baseline peak amplitude and single ventilation time in order to establish a deep and long breathing characteristic sequence. The interval analysis submodule calls the deep and long breathing feature sequence, iterates through the adjacent peak timestamps of the recorded elements in the deep and long breathing feature sequence one by one, calculates the absolute time difference component of the two timestamps, accumulates all the absolute time difference components in the group to define the span parameter, and generates the peak interval value. The duration calculation submodule compares the peak interval value with the inhalation cycle determination benchmark value, selects specific peak interval data segments that are less than the inhalation cycle determination benchmark value, and performs a summation operation on the span parameters corresponding to all specific peak interval data segments to obtain the duration of the inhalation phase.

5. The pediatric rhinitis nebulization treatment control system based on behavior guidance and intelligent monitoring according to claim 1, characterized in that, The rhythm feature quantification module includes: The difference calculation submodule calls the duration of the inhalation phase, compares the duration of the inhalation phase with the preset reference duration threshold, calculates the duration difference between the duration of the inhalation phase and the preset reference duration threshold, extracts the absolute value of the duration difference, and generates the duration deviation. The benchmark extraction submodule, for the deep and long breathing feature sequence, sequentially traverses the peak amplitude elements inside the deep and long breathing feature sequence, accumulates the peak amplitude elements and divides them by the total number of peak amplitude elements to obtain the waveform average parameter, assigns a constant value to the waveform average parameter, and obtains the benchmark amplitude value. The matrix building submodule retrieves the duration deviation, the baseline amplitude value, and the emotion state code based on the recorded timestamp, aligns the time nodes of the duration deviation, the baseline amplitude value, and the emotion state code by column, and concatenates the various parameters to build a feature rhythm deviation matrix.

6. The pediatric rhinitis nebulization treatment control system based on behavior guidance and intelligent monitoring according to claim 5, characterized in that, The preset baseline duration threshold is a baseline duration parameter obtained by statistically processing the historical data set of the duration of the inspiratory phase corresponding to the deep and long breathing characteristic sequence in the order of recorded timestamps, and the baseline duration parameter remains a fixed value within the same sampling period.

7. The pediatric rhinitis nebulization treatment control system based on behavior guidance and intelligent monitoring according to claim 1, characterized in that, The breathing behavior guidance module includes: The weighted operation submodule extracts the record distribution elements of each row and column inside the rhythm deviation feature matrix, sequentially assigns the distribution elements to the corresponding preset feature weight coefficients, and performs numerical multiplication calculation between the distribution elements and the preset feature weight coefficients to generate weighted distribution element values. The compensation extraction submodule calls the weighted distribution element value, extracts and filters all weighted distribution element values ​​located in the same row vector interval, performs an accumulation operation on all weighted distribution element values ​​after classification, and obtains the behavior guidance compensation amount. The offset conversion submodule, based on the behavior guidance compensation amount, collects the built-in visual animation coordinate nodes of the terminal display interface, performs coordinate conversion processing of the internal node trajectory offset of the visual animation coordinate nodes according to the deflection direction of the behavior guidance compensation amount, and generates dynamic guidance feedback parameters.

8. The pediatric rhinitis nebulization treatment control system based on behavior guidance and intelligent monitoring according to claim 1, characterized in that, The atomized drug release control module includes: The feature fusion submodule extracts discrete numerical sequences of the dynamic guidance feedback parameters and the emotional state codes, aligns the dynamic guidance feedback parameters and the emotional state codes by columns, performs matrix row and column concatenation operations, and generates a comprehensive state feature vector. The duty cycle conversion submodule calls the comprehensive state feature vector, calculates the deviation ratio term, error integral term, and error differential term between the comprehensive state feature vector and the preset reference state vector, accumulates the deviation ratio term, error integral term, and error differential term to perform a linear duty cycle mapping transformation, and generates duty cycle mapping parameters. The adjustment control submodule, based on the duty cycle mapping parameter, sequentially compares the duty cycle mapping parameter with the preset operating limit threshold, selects the duty cycle mapping parameter that does not exceed the preset operating limit threshold, performs underlying format encoding conversion processing, and generates atomization rate adjustment command.

9. The pediatric rhinitis nebulization treatment control system based on behavior guidance and intelligent monitoring according to claim 8, characterized in that, The preset benchmark state vector is a set of stable state data generated by performing statistical mean calculation and standard deviation constraint filtering on the discrete numerical sequence of dynamic guidance feedback parameters and emotional state codes in the historical operating cycle. The preset benchmark state vector is constructed after normalizing the set of stable state data. The preset operating limit threshold is formed by dividing the range of the maximum and minimum values ​​of the duty cycle mapping parameter within the historical operating cycle, and then performing numerical convergence processing on the upper and lower boundaries of the range in combination with the preset safety margin.

Images