Pediatric pain assessment system

Abstract

The invention relates to the technical field of pain assessment, in particular to a paediatric pain assessment system which comprises a muscle group rhythm recognition module, a behavior linkage screening module, a periodic fluctuation extraction module, a behavior physiological difference module and a pain grading judgment module. According to the method, a linkage offset identification path is constructed through an angle relation between facial muscle groups and a starting time sequence, multi-muscle cooperative imbalance performance is identified, a region-level dynamic behavior judgment basis is formed, and a behavior cooperative judgment mechanism is established by utilizing response overlapping conditions of multiple behavior characteristics in the same time period; path signals which do not form effective cooperation are eliminated, the effectiveness of behavior characteristics is enhanced, a continuous trend structure of peak time in rhythm change is introduced, a periodic response sudden change monitoring channel is constructed, the recognition sensitivity of a short-time fluctuation concentration section is improved, and a response mismatch section is extracted based on a trigger time dislocation state between behavior and physiological response, so that the recognition sensitivity of the short-time fluctuation concentration section is improved. And a pain judgment key point is positioned in the time sequence overlapping region.

Description

Technical Field

[0001] The present invention relates to the technical field of pain assessment, and particularly to a pediatric pain assessment system. Background Art

[0002] Pain assessment technology is a key branch in medicine and nursing, mainly studying how to objectively and accurately identify and quantify patients' pain experiences. This field covers various methods such as physiological parameter monitoring, behavioral observation scales, self-report tools, image recognition, and AI-assisted analysis, and is widely used in clinical diagnosis and treatment, chronic disease management, postoperative recovery, and first aid scenarios.

[0003] Among them, the pediatric pain assessment system is a pain recognition and assessment system specifically designed for children, aiming to overcome problems such as limited expression ability and difficulty in subjective description of children. By integrating technologies such as facial expression recognition, cry analysis, physiological parameter monitoring, and multi-dimensional behavior assessment scales, this system helps medical staff accurately judge the pain level of children. Its main uses include pain assessment during clinical treatment, pain management during routine operations such as postoperative care and vaccine injection, and continuous monitoring of children with chronic diseases, thus providing precise and scientific support for pediatric medical services.

[0004] The prior art often makes threshold determination through single-frame facial images or cry amplitude curves, ignoring the coupling state of the path angles between multiple muscles and the action initiation sequence, making it difficult to identify the rhythm misalignment phenomenon between actions. An interactive judgment framework has not been constructed between behavioral characteristics, resulting in a lack of a linkage reference benchmark for different types of behavioral signals. The paths of linkage failure cannot be effectively identified and incorporated into the assessment. Physiological curves and behavioral events belong to independent trajectories, and the rhythm mutation response cannot be monitored in advance through the evolution of peak trends, resulting in the inability to completely extract the cycle structure. The misalignment state where behavior precedes or lags behind physiological signals lacks differential inversion judgment within a time period, and the abnormal initiation sequence existing in pain expression is misjudged or undetected. The assessment results lack a stable cross-mapping basis and are difficult to achieve hierarchical output under the background of multi-dimensional complex data, affecting the adaptability accuracy and response timeliness of pain management measures. Summary of the Invention

[0005] To address the technical problems in the prior art, such as threshold determination often being carried out through single-frame facial images or cry amplitude curves, ignoring the coupling state of the path angles and action start sequences between multiple muscles, making it difficult to identify the rhythm misalignment phenomenon between actions, no interactive judgment framework being constructed between behavioral features, resulting in a lack of a linkage reference benchmark for different types of behavioral signals, the paths of linkage failure being unable to be effectively identified and incorporated into the evaluation, the physiological curve and behavioral events belonging to independent trajectories, the rhythm mutation response not being able to be monitored in advance through the evolution of peak trends, resulting in the inability to completely extract the cycle structure, the misalignment state of the behavior preceding or lagging behind the physiological signal lacking differential inversion judgment within a time period, the abnormal start sequence in pain expression being misjudged or undetected, and the evaluation result lacking a stable cross-mapping basis, making it difficult to achieve hierarchical output in the context of multi-dimensional complex data, which affects the adaptation accuracy and response timeliness of pain management measures, the embodiments of the present invention provide a pediatric pain assessment system. The technical solutions are as follows: On the one hand, a pediatric pain assessment system is provided, including: Based on the collected sequence of children's facial images, the muscle group rhythm recognition module extracts the starting positions and movement directions of the zygomatic major muscle, frontal muscle, and depressor anguli oris muscle, calculates the angles and start intervals between the muscles, identifies the situation of movement rhythm misalignment, screens the combinations of different steps, and obtains the facial linkage angle offset; Based on the facial linkage angle offset, the behavior linkage screening module obtains the start and duration of the cry frequency, chin muscle stretch, body vibration, and blink behavior responses, analyzes the response overlap between multiple behaviors, and eliminates the behavior channels lacking linkage to obtain invalid behavior path data; The cycle fluctuation extraction module calls the invalid behavior path data, draws the facial rhythm, heart rate fluctuation, and muscle activity curves, identifies the change trend of the signal peak spacing in adjacent time periods, screens the regions with concentrated mutations, and obtains the rhythm mutation response section; According to the rhythm mutation response section, the behavior-physiology difference module analyzes the start times of the cry and heart rate responses, calculates the time difference between them, identifies the misalignment interval between the behavior and physiological responses, and obtains the behavior-physiology misalignment section.

[0006] On the other hand, the facial linkage angle offset includes the included angle of the muscle group movement trajectory, the action start time label, and the linkage coordination abnormality mark. The invalid behavior path data includes the behavior response time distribution, the channel linkage density index, and the abnormal channel number set. The rhythm mutation response section includes the signal peak time interval sequence, the cycle fluctuation amplitude data, and the mutation cycle segment index. The behavior-physiology misalignment section includes the start response time comparison group, the synchronous offset result between the behavior and physiology, and the cross-type signal misalignment section label.

[0007] On the other hand, the muscle group rhythm recognition module includes: The muscle point positioning sub-module analyzes the starting regions of the zygomatic major muscle, frontal muscle, and depressor anguli oris muscle labeled in consecutive frames of the collected children's facial image sequence. By comparing the spatial position changes of each muscle region in adjacent frames, it determines the coordinate correspondence relationship of the same muscle in different frames and establishes a muscle position data sequence; The rhythm parameter extraction sub-module calculates the change in the movement direction angle of each muscle within consecutive frames based on the relative displacement of the muscle positions between each frame in the muscle position data sequence, determines the frame order relationship when the muscle starts to move, and obtains the muscle movement rhythm parameters by sorting out the change trends of each group of muscles in the image sequence; The linkage rhythm screening sub-module screens the relationship between the direction angle trends and starting frames of multiple groups of muscles in the muscle movement rhythm parameters, judges whether the direction trends between different muscles are consistent and whether the movement start order is coordinated, and counts the combinations that do not have isotropy or action synchronization to obtain the facial linkage angle offset.

[0008] On the other hand, the behavior linkage screening module includes: The cry frequency analysis sub-module obtains the corresponding image sequence time period based on the facial linkage angle offset, analyzes the starting frame and ending frame of the vocalization signal in the cry frequency during this period, calculates the frame distance distribution between consecutive vocalization segments, judges whether there are interval changes in the frame distance fluctuation, screens the consecutive paragraphs with the characteristic of vocalization aggregation, and obtains the vocalization dense distribution amount; The facial stretching determination sub-module analyzes the vertical displacement direction of the chin area in consecutive image frames according to the image paragraph covered by the vocalization dense distribution amount, compares the direction consistency of the displacement trajectories between the starting frame and the ending frame, judges whether a continuous upward or downward trend is formed, and screens the image segments with a continuously one-way offset trajectory to obtain the stretching trajectory offset degree; The behavior overlap detection sub-module calls the time period corresponding to the stretching trajectory offset degree, calculates the first and last occurrence frame labels of the blinking action in the frame sequence, performs a time-axis cross-judgment with the body vibration frame label interval in the same segment, and screens the behavior channel data of the non-overlapping segments to obtain the invalid behavior path data.

[0009] On the other hand, the periodic fluctuation extraction module includes: The rhythm curve construction sub-module calls the image frame segment covered by the invalid behavior path data, analyzes the facial action frame order, heart rate signal density, and muscle activity frequency band, calculates the corresponding change trajectory positions of the three types of signals in consecutive frames respectively, draws a continuous curve in chronological order, and obtains the signal rhythm distribution information; The peak distance change recognition sub-module calculates the frame sequence interval between consecutive peak points according to the change positions of the curve trajectories in the signal rhythm distribution information, compares the variation amplitudes of adjacent intervals in consecutive frame segments, screens the frame segment sequences with key changes occurring continuously, and obtains the curve interval fluctuation trend. The abnormal fluctuation screening sub-module determines whether there are inconsistent change directions or delayed response distributions in the three types of curves of the face, heart rate, and muscles according to the time periods marked by the curve interval fluctuation trend, analyzes the concentrated time periods of synchronous deviation in multiple segments of data, and obtains the rhythm mutation response section.

[0010] On the other hand, when analyzing the concentrated time periods of synchronous deviation in multiple segments of data, the following formula is used: ; Calculate the synchronous deviation value , and obtain the rhythm mutation response section, where represents the number of time periods, represents the heart rate signal in the th time period, represents the mean value of the heart rate signals in all time periods, represents the standard deviation of the heart rate signals in all time periods, represents the muscle signal in the th time period, represents the mean value of the muscle signals in all time periods, represents the standard deviation of the muscle signals in all time periods, represents the face signal in the th time period, represents the mean value of the face signals in all time periods, represents the standard deviation of the face signals in all time periods, represents the time mark of the th time period, represents the mean value of the time marks in all time periods, represents a non-zero constant used to smooth the square root term to prevent instability caused by being zero.

[0011] On the other hand, the behavior-physiology difference module includes: The response time extraction sub-module analyzes the starting points of consecutive sound waves in the vocalization trajectory according to the frame sequence range of the rhythm mutation response section, synchronously detects the first densely fluctuating paragraph in the heart rate data, compares the arrangement order of the first frames of the two segments on the time axis, determines their sequence relationship and establishes a frame position correspondence item to obtain the starting offset of vocalization and heart rate. The time period difference calculation sub-module analyzes the continuous fluctuation segments corresponding to the cry and heart rate in the image frame axis based on the frame sequence position in the starting heart rate offset of the vocalization, compares the time spans between the start and end frames of the two types of signals, and screens the regions with relatively stable change directions of the frame segments to obtain the behavior duration comparison quantity; The signal consistency judgment sub-module compares the fluctuation directions in the cry amplitude rising segment and the heart rate change trajectory according to the fluctuation frame segments in the behavior duration comparison quantity, determines whether they show synchronous characteristics within the same time, and screens out the intervals with reverse changes or fluctuation faults to obtain the behavior physiological misalignment section.

[0012] On the other hand, when comparing the fluctuation directions in the cry amplitude rising segment and the heart rate change trajectory, the formula: ; is used to calculate the fluctuation synchronization index and determine whether it shows synchronous characteristics within the same time. Among them, represents the amplitude in the rising segment of the cry amplitude of the th frame, represents the heart rate in the heart rate change trajectory of the th frame, represents the mean value of the amplitudes in the entire rising segment of the cry amplitude, represents the mean value of the heart rates in the entire heart rate change trajectory, represents the total number of frames in the current comparison frame segment, represents a non-zero constant used to smooth the square root term to prevent instability caused by zero.

[0013] On the other hand, the system further includes: The pain grading determination module monitors whether there are simultaneous changes in facial, behavioral, and physiological signals within the monitored time period based on the behavior physiological misalignment section, identifies the overlapping regions of the three types of signals, analyzes whether they appear at the characteristic positions in the pain judgment criteria, and determines the pain degree grading of the corresponding regions to obtain the multi-source pain grading regions; The multi-source pain grading regions include multi-dimensional signal fusion results, pain level classification numbers, and reference standard mapping indexes.

[0014] On the other hand, the pain grading determination module includes: The signal change detection sub-module analyzes the continuous expression contour changes in the facial image, the vocalization fluctuations in the sound trajectory, and the starting points of the fluctuations in the heart rate curve based on the frame sequence interval covered by the behavior physiological misalignment section, and determines whether the signals produce linkage changes within the same frame segment range to obtain the signal change synchronization quantity; The overlapping segment screening sub-module, based on the frame segment range involved in the signal change synchronization quantity, compares the starting trajectories of facial, behavioral, and physiological signals at the same frame sequence, screens out the frame segment fragments that have intersections and continuously extend in the time sequence direction, and then counts the number of overlapping ranges in the continuously distributed segments to obtain the total amount of overlapping distribution segments; The feature level matching sub-module analyzes whether the corresponding positions are located in the key segments set in the pain level reference sequence according to the signal coincidence structure and frame segment duration characteristics in the total amount of overlapping distribution segments, judges the matching degree between the linkage form presented by the frame segments and the level standard, determines the corresponding pain degree grading, and obtains the multi-source pain grading area.

[0015] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include: By constructing a linkage offset recognition path through the angular relationship and starting time sequence between facial muscle groups, identifying the performance of multi-muscle coordination imbalance, forming a basis for regional dynamic behavior judgment, using the response overlap situation of multiple behavior characteristics in the same time period, establishing a behavior coordination discrimination mechanism, excluding the path signals that do not form effective cooperation, enhancing the effectiveness of behavior characteristics, introducing the continuous trend structure of the peak time in the rhythm change, constructing a periodic response mutation monitoring channel, improving the recognition sensitivity to the short-term fluctuation concentration segment, based on the trigger time misalignment state between behavior and physiological responses, extracting the response mismatch segment, and locating the pain judgment key point in the time sequence overlap area, delimiting the grade mapping block, and overall taking the time sequence difference, channel coordination, and periodic change as the combined judgment core, constituting a multi-source fusion analysis chain across the behavior and physiological dimensions, improving the recognition clarity and grade discrimination ability of pain expression in non-standard behavior states. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a system schematic diagram of the present invention; Figure 2 It is a system framework schematic diagram of the present invention; Figure 3 It is a flowchart of the muscle group rhythm recognition module of the present invention; Figure 4 It is a flowchart of the behavior linkage screening module of the present invention; Figure 5 It is a flowchart of the periodic fluctuation extraction module of the present invention; Figure 6Flow chart of the behavioral physiology pair difference module of the present invention; Figure 7 Flow chart of the pain grading determination module of the present invention. Detailed implementation manners

[0018] The technical solutions in the present invention will be described below with reference to the accompanying drawings.

[0019] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two can be selected.

[0020] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0021] In the embodiments of the present invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0022] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0023] The embodiments of the present invention provide a pediatric pain assessment system, as Figure 1 shown, the system includes: The muscle group rhythm recognition module is based on the collected sequence of children's facial images, calls the starting positions and movement directions of the zygomatic major muscle, frontal muscle and depressor anguli oris muscle in the image sequence, calculates the movement angles and movement time intervals between each pair of muscles, determines whether the movement rhythms between multiple muscles are misaligned or out of sync, screens inconsistent movement combinations, and obtains the facial linkage angle offset; The behavioral linkage screening module is based on the facial linkage angle offset, obtains the cry frequency, the stretching ratio of the chin muscle, the body vibration change and the number of blinks within a time period, counts the start time and duration of each behavioral response, analyzes the number of overlaps in time between different behavioral responses, screens the behavioral channels that do not participate in the linkage as invalid information, and obtains the invalid behavioral path data; The periodic fluctuation extraction module calls the invalid behavior path data, draws the change curves of the facial action rhythm, heart rate fluctuation trajectory, and muscle activity duration, checks the highest point positions of each curve in adjacent time periods, records the change trend of the time difference between the highest points, determines which time periods have unstable or mutated signal distributions, screens the time periods with fluctuating mutations as abnormal segments, and obtains the rhythm mutation response sections; The behavioral physiology pair difference module analyzes the start times of crying and heart rate responses in the mutated segment according to the rhythm mutation response section, calculates the time difference between the two types of data, identifies the time periods when the behavioral response is earlier or later than the physiological response, and determines whether there is an inconsistent signal type difference in this time period, obtaining the behavioral physiology misalignment section; The pain grading determination module monitors whether there are simultaneous changes in facial, behavioral, and physiological signals within the time period based on the behavioral physiology misalignment section, identifies the overlapping regions of the three types of signals, checks the quantity and duration of the signal overlapping parts, analyzes whether they appear at the characteristic positions in the pain judgment criteria, determines the pain degree grading of the corresponding regions, and obtains the multi-source pain grading regions.

[0024] The facial linkage angle offset includes the included angle of muscle group movement trajectories, action start time tags, and linkage coordination abnormality marks. The invalid behavior path data includes behavioral response time distributions, channel linkage density indicators, and abnormal channel number sets. The rhythm mutation response section includes signal peak time interval sequences, periodic fluctuation amplitude data, and mutant cycle segment indexes. The behavioral physiology misalignment section includes start response time comparison groups, synchronization offset results between behavior and physiology, and cross-type signal misalignment section tags. The multi-source pain grading region includes multi-dimensional signal fusion results, pain level classification numbers, and reference standard mapping indexes.

[0025] A behavioral channel refers to each independent source of behavioral signals in the system, which serves as a separate data input path to reflect the non-verbal behavioral responses of children within a specific time period. For example, the crying channel (crying frequency, intensity, spectral distribution), the mentalis muscle channel (mentalis muscle stretching ratio, muscle tension state), the body vibration channel (frequency of slight limb movements or torso tremors), and the blinking channel (frequency of blinking, time interval).

[0026] As Figure 2 and Figure 3 shown, the muscle group rhythm recognition module includes: The muscle point positioning sub-module analyzes the starting regions of the zygomatic major muscle, frontalis muscle, and depressor anguli oris muscle marked in consecutive frames of the collected child facial image sequence based on the collected child facial image sequence. By comparing the spatial position changes of each muscle region in adjacent frames, it determines the coordinate correspondence relationship of the same muscle in different frames and establishes a muscle position data sequence; During the image preprocessing stage, regional recognition is performed on each frame of the image. By extracting the central point coordinates of the target muscle group regions marked in the image, a list of muscle coordinate information indexed by the frame number is formed. For example, in the 5th frame of the image, it is recognized that the zygomatic major muscle is located at the horizontal coordinate of 120 pixels and the vertical coordinate of 80 pixels in the image, recorded as (120, 80). In the 6th frame, if the recognized coordinates of this muscle change to (125, 85), it means that the muscle has moved in the upper right direction in terms of spatial position between two adjacent frames. Repeat this process to traverse the changes in the central points of all target muscle regions in each frame of the entire image sequence, and store the results of the spatial coordinate differences of each target muscle between consecutive frames to form structured spatial movement data. Then, by comparing the position relationships of the same muscle in different images in adjacent frames, that is, comparing the position changes of the muscle in the front and back frames from front to back according to the frame sequence index, and traversing the entire frame image in this way. During the continuous comparison process, to ensure the tracking continuity of the same muscle, only the cases where the position change amplitude of the marked area is within a reasonable range are recorded. If the position drift between two frames exceeds three times the diameter of the preset recognition area, the muscle positioning in this frame fails, and the data of this frame needs to be excluded from the continuous sequence. Finally, a coordinate trajectory linked list data set of each muscle spanning all valid frames in the image sequence is formed, which is the muscle position data sequence.

[0027] Based on the relative displacement of the muscle positions between each frame in the muscle position data sequence, the rhythm parameter extraction sub-module calculates the change in the movement direction angle of each muscle within consecutive frames, determines the frame position sequence relationship when the muscle starts to move, and obtains the muscle movement rhythm parameters by sorting out the change trends of each group of muscles in the image sequence. Sequentially read the coordinate points of the same muscle in the front and back frames, and evaluate the change trend in the spatial direction through the displacement direction line segment formed by connecting this pair of coordinate points. This operation is performed for any two adjacent frames to obtain a series of direction angle change information. On this basis, traverse the movement direction trend sequence formed by each muscle in the entire image sequence, extract the time periods with relatively large direction changes or stable directions, and at the same time, record the frame number when each muscle first undergoes an obvious displacement in the image sequence as the starting frame. The judgment criterion is that the position offset between two frames exceeds 5 pixels of the preset pixel difference and there is a moving trend with the same direction for at least 3 consecutive frames, which is regarded as the starting point of movement for this frame. For example, if the displacement of the zygomatic major muscle in the 8th frame first meets this criterion, the starting frame number is 8. Subsequently, extract the direction change sequences and starting frame numbers of the frontal muscle and the depressor anguli oris muscle respectively in the same way, sort out the direction trend data and starting time sequences corresponding to each muscle, and finally, output the muscle movement rhythm parameters for subsequent screening and analysis.

[0028] The linkage rhythm screening sub-module screens the relationship between the direction angle trends of multiple groups of muscles and the starting frame in the muscle movement rhythm parameters, judges whether the direction trends between different muscles are consistent and whether the movement start order is coordinated, counts the combinations that do not have isotropy or action synchronization, and obtains the facial linkage angle offset; The three target muscles are combined in pairs to form three muscle pairs. The changing trends of the movement directions and the starting frame sequence information of the two muscles in each group are compared in turn. First, for each combination, the direction change data of the two muscles are compared frame by frame in the image sequence to judge whether they show similar direction trends in the same time period. The situation where the frame-to-frame direction difference does not exceed 10 degrees is defined as a consistent direction trend. If this trend continuously appears for more than three consecutive frames, it is recorded as having isotropy. At the same time, the starting frame numbers of the two muscles are compared. If the difference between the two numbers does not exceed 2 frames, that is, the starting action appears within adjacent frames or the time difference is less than or equal to 66 milliseconds, it is considered that the starting time sequence is coordinated. If the direction trend difference in a certain combination exceeds 15 degrees, or the difference in the starting frame numbers is greater than 3 frames, it is determined as a non-synchronous linkage combination. All muscle combinations that do not meet the linkage conditions are counted and marked with a quantity, and at the same time, the corresponding direction difference range and the difference in the starting frame numbers are recorded. In the system initialization stage, the threshold of 10 degrees set by this judgment criterion comes from the median result of the common direction fluctuations during the normal muscle movement of children's faces. The direction angle difference interval of 0 to 5 degrees is high synchronization, 5 to 10 degrees is medium synchronization, 10 to 15 degrees is low synchronization, and more than 15 degrees is asynchronous. Finally, all muscle combinations that do not meet the direction consistency and starting time sequence coordination and their analysis results are formed into a structured list and output as the facial linkage angle offset.

[0029] As Figure 2 and Figure 4 shown, the behavior linkage screening module includes: The cry frequency analysis sub-module obtains the corresponding image sequence time period based on the facial linkage angle offset, analyzes the starting frame and the ending frame of the vocalization signal in the cry frequency during this period, calculates the frame distance distribution between consecutive vocalization segments, judges whether there are interval changes in the frame distance fluctuations, screens out the consecutive paragraphs with the characteristics of vocalization aggregation, and obtains the vocalization dense distribution quantity; Extract the crying frequency data synchronized with this time period, identify all audio segments with obvious sound wave energy fluctuations therefrom, align the audio track along the time axis according to the image frame rate, extract the instantaneous energy value of the audio signal for each frame, and identify the frames with energy values significantly higher than the reference energy of the background noise by comparing the energy value of each frame with the reference energy of the background noise. For example, set the reference as the mean value of the audio energy of the first 100 frames plus the standard deviation in the mute state. If the energy of a certain frame is at least 20% higher than this value, then this frame is recorded as the starting frame of vocalization. Continuously detect the starting and ending frames of each vocalization segment and record their numbers as a vocalization paragraph. Subsequently, analyze the difference in the inter-frame distance between each vocalization segment to obtain a list of the inter-frame distance distributions between vocalization paragraphs, and determine whether there is a dense segment with an inter-frame distance within 50 frames of the lower limit of the set vocalization aggregation threshold. When the inter-frame distance difference of consecutive vocalization segments is less than 50 frames and the number of consecutive occurrences reaches three or more times, it is defined as a vocalization aggregation segment. For example, the inter-frame distances between three consecutive segments from frame 100 to frame 130, from frame 160 to frame 185, and from frame 200 to frame 230 are 30 frames and 25 frames respectively, then it forms a vocalization aggregation group as a whole. Count all the audio segments that meet this standard to obtain the vocalization dense distribution quantity.

[0030] The facial stretching determination sub-module analyzes the vertical displacement direction of the chin area in consecutive image frames according to the image paragraph covered by the vocalization dense distribution quantity, compares the direction consistency of the displacement trajectories between the starting frame and the ending frame, determines whether a continuous upward or downward trend is formed, and filters out the image segments with a continuously one-way offset trajectory to obtain the stretching trajectory offset degree. First, extract the central position coordinates of the chin area in each frame of the image sequence and arrange them in frame order to form a coordinate trajectory list. Calculate the vertical coordinate difference between every two adjacent frames to obtain the frame-by-frame vertical displacement value of the chin center position. Then, by summarizing the positive and negative directions of all displacement values between the starting frame and the ending frame, analyze whether the chin movement shows continuous one-wayness during this time period. The judgment criterion is: if the vertical coordinates in all consecutive frames within the entire paragraph continuously increase, it is judged as a downward trend; conversely, if the vertical coordinates continuously decrease, it is judged as an upward trend. If the direction of more than 20% of the frame segments is reversed, it does not form a one-way trend. For example, from frame 100 to frame 150, the chin coordinates gradually increase from 90 to 135, and only 8 frames have decreasing coordinates in this paragraph, with a proportion of 16%, which is less than the 20% threshold, so it can be regarded as forming a stable downward trend. The 20% direction reversal threshold is derived from the median frequency of analyzing the coordinate fluctuation value distributions in 50 actual vocalization images of children's faces to reduce misjudgment caused by jitter or image blur. Filter out the image paragraphs with continuous one-way vertical displacement characteristics, and quantify the vertical distance between the starting and ending positions in this paragraph as the stretching trajectory offset degree of this paragraph.

[0031] The behavior overlap detection sub-module calls the time period corresponding to the stretching trajectory offset, calculates the first and last occurrence frame labels of the blinking action in the frame sequence, makes a time-axis intersection judgment with the body vibration frame label interval within the same segment, filters the behavior channel data of the non-overlapping segments, and obtains the invalid behavior path data; Extract the frame sequence information of the blinking action and body vibration within this time period. First, mark the blinking state of each frame in the image sequence, record the start frame and end frame of blinking, and form the start and end frame numbers of each blinking action segment. For example, a complete blinking process is recognized from frame 102 to frame 108. Organize the frame segment list of all blinking actions within the current time period in this way, then extract the action frame interval output by the body vibration recognition module, construct the set of vibration segment frame numbers, and then make a cross-judgment on the frame segments of the two behavior channels. The method is as follows: for each blinking frame segment, perform an interval overlap check on the start frame and end frame with the body vibration frame segment. If there is no time-axis overlap with the current blinking segment in any vibration segment, that is, the end frame of the blinking frame segment is earlier than the start frame of the vibration frame segment, or the start frame of the blinking frame segment is later than the end frame of the vibration frame segment, it is regarded as no behavior overlap. Count the blinking and vibration combinations of all such non-overlapping behavior paragraphs to form a non-overlapping behavior channel set, and label the above non-intersecting behavior time periods and their behavior types as invalid behavior path data.

[0032] As Figure 2 and Figure 5 shown, the periodic fluctuation extraction module includes: The rhythm curve construction sub-module calls the image frame segment covered by the invalid behavior path data, analyzes the facial action frame sequence, heart rate signal density, and muscle activity frequency band, calculates the corresponding change trajectory positions of the three types of signals in consecutive frames respectively, draws a continuous curve in chronological order, and obtains the signal rhythm distribution information; Extract three types of signal data within this time range in sequence, namely facial action frame sequence, heart rate signal density, and muscle activity frequency band. For the facial action frame sequence, first obtain the movement direction and amplitude of facial key points in each frame image within the corresponding time period. By counting the vertical and horizontal displacement quantities of the zygomatic major muscle, frontal muscle, and depressor anguli oris muscle in consecutive frames, convert them into a sequence of facial activity intensity values per frame. For example, within frames 110 to 130, if the zygomatic major muscle moves upward by 1 pixel per frame, the action intensity value is 1; for the heart rate signal density, collect the pulse wave signal synchronized with the image frames, and convert the heart rate signal per second into the average number of pulses between image frames. For example, if the video is 30 frames per second and the heart rate within this second is 90 beats per minute, the average density per frame is 1.5 times. Generate a sequence of heart rate density change values for all frame segments in this way; for the muscle activity frequency band, analyze the position and energy magnitude of the main frequency band in each frame of electromyogram signal, count the number of changes in the main frequency within every 10 frames and normalize it to a frequency value. Construct continuous curves with the frame number corresponding to the three types of numerical values as the horizontal axis (frame sequence) and the intensity value as the vertical axis. By connecting the numerical changes of the three types of signals point by point, form three types of continuous rhythm change curves for the face, heart rate, and muscles, and draw and output the signal rhythm distribution information.

[0033] The peak distance change recognition sub-module calculates the frame sequence interval between consecutive peak points according to the change position of the curve trajectory in the signal rhythm distribution information, compares the change amplitude of adjacent intervals in consecutive frame segments, screens the frame segment sequence with key changes occurring continuously, and obtains the curve interval fluctuation trend; Identify all local maximum positions in each curve, record them as the peak point position sequence of the signal, calculate the frame sequence difference between adjacent two peak points segment by segment to form a frame distance sequence, and then compare each item in this frame distance sequence adjacent to each other to obtain the difference between each frame distance segment and the subsequent frame distance segment. For example, if peak points appear at frames 110, 130, 152, and 177 in a certain signal curve, the frame distances are 20, 22, and 25 frames in sequence, and the adjacent frame distance differences are 2 and 3 frames. Judge whether each difference is greater than the frame distance fluctuation threshold. The fluctuation threshold is set to 5 frames according to the signal frequency stability in the normal physiological rhythm of children. If three consecutive frame distance differences are all greater than this value, then determine that this segment is a key change segment. For example, if the frame distances are 20, 28, 36, and 45 respectively, the adjacent differences are 8, 8, and 9 frames, which meet the conditions. Output the set of frame segment numbers that meet the change amplitude judgment conditions as the curve interval fluctuation trend.

[0034] The abnormal fluctuation screening sub-module judges whether there are inconsistent change directions or delayed response distributions in the three types of curves for the face, heart rate, and muscles according to the time period marked by the curve interval fluctuation trend, analyzes the concentrated time periods of synchronous deviation in multiple segments of data, and obtains the rhythm mutation response section; Analyze the concentrated time periods of synchronous deviation in multiple segments of data, using the formula: ; Calculate the synchronization deviation value , where the synchronization deviation value is a quantitative index used to measure the relative coordination and temporal consistency deviation degree of facial signals, heart rate signals, and muscle signals within a certain period of time, and is used to analyze the concentrated periods of synchronization deviation in multiple data segments to obtain the rhythm mutation response section. Among them, represents the number of time periods, represents the heart rate signal within the th time period, represents the mean value of the heart rate signals in all time periods, represents the standard deviation of the heart rate signals in all time periods, represents the muscle signal within the th time period, represents the mean value of the muscle signals in all time periods, represents the standard deviation of the muscle signals in all time periods, represents the facial signal within the th time period, represents the mean value of the facial signals in all time periods, represents the standard deviation of the facial signals in all time periods, represents the time mark of the th time period, represents the mean value of the time marks in all time periods, represents a non-zero constant used to smooth the square root term to prevent instability caused by zero; Let , and the data acquisition frequency is recorded once every 10 seconds. The physiological sensor monitoring results are as follows: Heart rate signal (unit: bpm): , , ; Muscle signal (unit: μV, using EMG RMS): , , ; Facial signal (unit: μV): , , ; Time mark (unit: s): , , ; Smoothing constant: (a constant value set to maintain stability); Mean value: ; ; ; ; Standard deviation: ; ; ; Calculate item by item: Item 1: ; Item 2: ; Item 3: ; Sum and take the average: ; The result shows that the value is 0, and the deviation trends of the three types of signals in the three time periods are overall completely offset under the calculated weights, there is no obvious overall synchronization offset, and the synchronization state is relatively balanced within this window.

[0035] As Figure 2 and Figure 6 shown, the behavioral physiology difference module includes: The response time extraction sub-module analyzes the starting points of consecutive sound waves in the vocalization trajectory according to the frame sequence range of the rhythm mutation response section, synchronously detects the first dense fluctuation paragraph in the heart rate data, compares the arrangement order of the first frames of the two paragraphs on the time axis, judges their sequence relationship and establishes the frame position corresponding item to obtain the starting offset of vocalization heart rate; According to the frame sequence range of the rhythm mutation response segment, the audio data and heart rate data in the target time period are extracted, and the sound wave envelope analysis is performed on the audio data. The frame position of the first sound wave rising edge in the continuous sound signal is extracted as the sound start frame. For example, if the audio energy rises for three consecutive frames for the first time in the 210th frame and exceeds the silent energy benchmark, then the 210th frame is recorded as the starting point of the sound trajectory. At the same time, the fluctuation density detection is performed on the heart rate data in the same segment, and the starting position with the most intensive pulse waveform changes is selected as the first heart rate response frame. Specifically, the heart rate mutation per second is analyzed. The frequency of occurrence of variable values ​​is as follows: if a high-frequency fluctuation segment appears within 5 seconds after the start of a certain frame and the fluctuation amplitude exceeds the average amplitude of the previous cycle by more than 20%, the frame is marked as the starting frame of intensive heart rate fluctuation. For example, if continuous heart rate jump points begin to appear at frame 218, frame 218 is regarded as the starting point of the heart rate response. The numbering sequence of the utterance start frame and the heart rate response frame is compared, and the one with the smaller number is recorded as the first response signal. At the same time, the frame sequence positions of the two are correspondingly marked as a set of position matching items, which ultimately form the starting response offset data of the utterance and heart rate signals in all rhythm mutation segments.

[0036] The time period difference calculation submodule analyzes the continuous fluctuation segments corresponding to crying and heart rate in the image frame axis based on the frame sequence position in the starting offset of the vocalization heart rate, compares the time span between the start and end frames of the two types of signals, and selects the area with relatively stable frame segment change direction to obtain the behavior duration comparison; According to each group of paired items, the complete fluctuation segments of the crying and heart rate signals in the image frame axis are extracted, and the frame numbers corresponding to the first rising point and the last peak point of each type of signal in the segment are recorded to form the start and end time intervals of each signal, and their respective time spans are calculated as the behavior duration. For example, the crying fluctuation segment is detected in the 210th to 245th frames, and the heart rate fluctuation segment is detected in the 218th to 252nd frames, corresponding to durations of 35 frames and 34 frames respectively. Next, the relative positions between the start and end frames of the two signals are cross-compared. If the starting frame spacing of the two does not exceed 10 frames, the ending frame spacing does not exceed 15 frames, and there is no drastic reversal of the signal waveform trend within the frame segment, then the segment is considered to be a behavior cycle with relatively stable direction. The criterion for drastic reversal is that the waveform has more than two reverse jumps or the change amplitude exceeds 20% within 10 frames. All time periods that meet the stability condition are counted and the time span comparison values ​​of the two types of signals are sorted out in turn, and the numerical values ​​and time period positions of all behavior duration comparison quantities are output.

[0037] The signal consistency judgment submodule compares the fluctuation frame segments within the behavior duration comparison amount, compares the rising segment of the crying amplitude with the fluctuation direction in the heart rate change trajectory, and determines whether they present synchronous characteristics within the same time, and screens out the intervals with reverse changes or fluctuation faults to obtain the behavior and physiological dislocation segments; Compare the fluctuation direction in the rising segment of the cry amplitude with that in the heart rate change trajectory, and use the formula: ; Calculate the fluctuation synchronization index , where the fluctuation synchronization index is an index used to measure whether the cry amplitude change and the heart rate change synchronously fluctuate within the same time period. The fluctuation synchronization index generally compares and to evaluate whether they show consistency in the change trend, that is, whether they increase or decrease simultaneously in the same time period, and then judge whether they show synchronous characteristics within the same time. Among them, represents the amplitude in the rising segment of the cry amplitude of the th frame, represents the heart rate in the heart rate change trajectory of the th frame, represents the mean value of the amplitude in the entire rising segment of the cry amplitude, represents the mean value of the heart rate in the entire heart rate change trajectory, represents the total number of frames in the current comparison frame segment, represents a non-zero constant used to smooth the square root term to prevent instability caused by zero; The sampling frequency of the comparison frame segment is set to 100 frames per second, and the total duration is 5 seconds, which is uniformly controlled by the sampling device. The number of frames obtained is . The cry signal is collected by a microphone, and the amplitude value is calculated every 10 ms using the short-time energy method. After preprocessing and filtering the data, it is standardized. Five frames, namely the 5th frame, the 120th frame, the 240th frame, the 360th frame, and the 480th frame, are selected as representatives within the comparison time period. The data is as follows: The cry amplitude of the 5th frame , and the heart rate ; The cry amplitude of the 120th frame , and the heart rate ; The cry amplitude of the 240th frame , and the heart rate ; The cry amplitude of the 360th frame , and the heart rate ; The cry amplitude of the 480th frame , and the heart rate .

[0038] The mean value of the cry amplitude is calculated by averaging the amplitudes of the selected frames: ; The mean value of the heart rate is calculated in the same way: ; Smoothing parameter Used to prevent the denominator from being zero, with a value of 0.01, set according to the minimum theoretical fluctuation of the denominator after signal normalization. When the standard amplitude change is less than 0.05, Ensure that the denominator is non-zero.

[0039] Calculate the synchronization term for each frame: Frame 5: ; Frame 120: ; Frame 240: ; Frame 360: ; Frame 480: ; Substitute the calculated values of 5 frames into the overall average: ; The result shows that the instantaneous fluctuation synchronization index between the cry amplitude and heart rate fluctuation within the comparison frame segment is 68.445. The numerical value represents the matching degree between the current behavioral characteristics and physiological signals. This value is comprehensively obtained from multiple synchronization terms, forming a quantitative description of the overall behavioral and physiological fluctuation relationship of the current frame segment.

[0040] Such as Figure 2 and Figure 7 shown, the pain grading determination module includes: The signal change detection sub-module analyzes the starting points of fluctuations in the continuous expression contours in the facial image, the vocalization undulation patterns in the sound track, and the heart rate curve based on the frame sequence interval covered by the behavioral and physiological misalignment section, determines whether the signals generate linkage changes within the same frame segment range, and obtains the signal change synchronization quantity; Extract facial image data, voice tracks, and heart rate curve information in sequence. Perform facial contour extraction operations on each frame of the facial image data. By comparing the contour point coordinate changes of parameters such as eyelid opening and closing, mouth corner position, and eyebrow height in consecutive frames, form a sequence of facial expression morphological changes. Perform energy analysis and frequency band identification on the voice tracks in the same section, extract the audio energy value and main frequency structure of each frame, record the frame number sequence where the energy values of three consecutive frames are higher than the background reference as the starting point of sound production, and count the number of fluctuations in the sound production frequency band with every 5 frames as a window section. Perform jump point extraction operations on the heart rate curve within this frame section range, record the starting points of continuous upward or downward trends in the pulse curve, and form a sequence of fluctuation starting points. Subsequently, extract the frame numbers of the change starting points for the three types of signals respectively, compare them with the frame sequence axis, and determine whether significant changes occur simultaneously in the same frame section for the three types of signals. Set the judgment criterion as the maximum frame difference between the change starting frames of the three signals not exceeding 4 frames, that is, starting to change simultaneously within approximately 133 milliseconds, and it is considered a linked change frame section. Statistically record all frame sections that meet the conditions in the behavioral and physiological misalignment section according to this standard, and summarize them as the signal change synchronization quantity.

[0041] Based on the frame section range involved in the signal change synchronization quantity, the overlapping section screening sub-module compares the starting trajectories of facial, behavioral, and physiological signals at the same frame sequence, screens frame section segments that have intersections and continue to extend in the time sequence direction, and then counts the number of overlapping ranges in the continuously distributed segments to obtain the total number of overlapping distribution segments; Extract the starting trajectories of changes in facial, behavioral, and physiological signals in sequence within the time period, compare the starting frames and continuation trends of the three types of signals on the same frame axis, and judge whether each signal starts to continuously fluctuate within this frame section from the starting frame, that is, the change value maintains an increasing or decreasing trend in at least 5 consecutive frames. Perform time axis superposition operations according to the frame section span. Consider the frame section where the starting frames of fluctuations of the three types of signals are in the same frame section and there is at least 8-frame trend extension thereafter as a linked continuation section, record the starting and ending frames of each signal in the corresponding frame section, construct a frame section coincidence set, and perform an intersection judgment on all the starting frame and ending frame sequences in this set. If the overlapping frame sequence range of the frame sections of the three types of signals is not less than 6 frames, record it as an effective overlapping distribution segment, and count the total number of all such overlapping segments one by one as the total number of overlapping distribution segments. The minimum intersection frame number of 6 frames set in this coincidence judgment is set according to the median value of clinical data on the synchronization probability of facial reactions and heart rate changes, and output the total number of overlapping distribution segments.

[0042] Based on the signal coincidence structure and frame section duration characteristics in the total number of overlapping distribution segments, the feature level matching sub-module analyzes whether its corresponding position is in the key section set in the pain level reference sequence, judges the matching degree between the linked form presented by the frame section and the level standard, determines the corresponding pain degree classification, and obtains the multi-source pain classification area; Compare the peak facial action intensity, the rising speed of the cry amplitude, and the amplitude of the heart rate fluctuation in each segment, and map them to the key section positions set in the pain level reference sequence according to the frame sequence position. In each frame segment, compare item by item whether the change patterns of the three types of signals match the characteristic curve types defined in the level standard. For example, if within the 250th frame to the 270th frame, all three types of signals show synchronous increase, and the peak facial action intensity exceeds 1.5 times the action reference amplitude, the average cry energy is higher than twice the silent reference, and the heart rate fluctuation shows a jump frequency exceeding 10% for 8 consecutive frames, then mark this segment as a high-matching segment, and then map it according to the level area in the reference sequence where this segment is located. If this segment is in the third-level pain section, it is recorded as the corresponding frame segment of the third-level pain. Sequentially confirm and label the matching levels for all frame segments that meet the mapping conditions, and output the frame sequence positions and their level values corresponding to all pain levels to form a multi-source pain grading area.

[0043] It should be understood that the term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. Additionally, the character " / " in this article generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context before and after.

[0044] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following items" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0045] It should be understood that in various embodiments of the present invention, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0046] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0047] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0048] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0049] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0050] In addition, in each embodiment of the present invention, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0051] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0052] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims described above.

Claims

1. A pediatric pain assessment system, characterized in that, The system includes: Based on the collected sequence of children's facial images, the muscle group rhythm recognition module extracts the starting positions and movement directions of the zygomatic major muscle, frontalis muscle, and depressor anguli oris muscle, calculates the angles and starting intervals between the muscles, identifies the situation of movement rhythm dislocation, screens the combinations of different steps, and obtains the facial linkage angle offset; Based on the facial linkage angle offset, the behavior linkage screening module obtains the starting and duration times of the cry frequency, chin muscle stretching, body vibration, and blinking behavior responses, analyzes the response overlap among multiple behaviors, and eliminates the behavior channels lacking linkage to obtain the invalid behavior path data; The periodic fluctuation extraction module calls the invalid behavior path data, draws the facial rhythm, heart rate fluctuation, and muscle activity curves, identifies the change trend of the signal peak interval in adjacent time periods, screens the regions with concentrated mutations, and obtains the rhythm mutation response section; Based on the rhythm mutation response section, the behavior-physiology pair difference module analyzes the starting times of the cry and heart rate responses, calculates the time difference between them, identifies the interval of misalignment between behavior and physiological responses, and obtains the behavior-physiology misalignment section.

2. The pediatric pain assessment system according to claim 1, characterized in that, The facial linkage angle offset includes the included angle of the muscle group movement trajectory, the action start time label, and the linkage coordination abnormality mark. The invalid behavior path data includes the behavior response time distribution, the channel linkage density index, and the abnormal channel number set. The rhythm mutation response section includes the signal peak time interval sequence, the periodic fluctuation amplitude data, and the mutation cycle segment index. The behavior-physiology misalignment section includes the starting response time comparison group, the synchronous offset result between behavior and physiology, and the cross-type signal misalignment section label.

3. The pediatric pain assessment system according to claim 1, characterized in that, The muscle group rhythm recognition module includes: Based on the collected sequence of children's facial images, the muscle point positioning sub-module analyzes the starting regions of the zygomatic major muscle, frontalis muscle, and depressor anguli oris muscle marked in consecutive frames of the image sequence. By comparing the spatial position changes of each muscle region in adjacent frames, it determines the coordinate correspondence relationship of the same muscle in different frames and establishes a muscle position data sequence; Based on the relative displacement of the muscle positions between each frame in the muscle position data sequence, the rhythm parameter extraction sub-module calculates the change in the movement direction angle of each muscle within consecutive frames, determines the frame position order relationship of the start of muscle movement, and obtains the muscle movement rhythm parameters by sorting out the change trends of each group of muscles in the image sequence; The linkage rhythm screening sub-module screens the relationship between the direction angle trend and the starting frame of multiple groups of muscles in the muscle movement rhythm parameters, judges whether the direction trends between different muscles are consistent and whether the movement start order is coordinated, and counts the combinations lacking isotropy or action synchrony to obtain the facial linkage angle offset.

4. The pediatric pain assessment system according to claim 1, characterized in that, The behavior linkage screening module includes: Based on the facial linkage angle offset, the cry frequency analysis sub-module obtains the corresponding time period of the image sequence, analyzes the starting frame and ending frame of the vocalization signal in the cry frequency during this period, calculates the frame distance distribution between consecutive vocalization segments, judges whether there is an interval change in the frame distance fluctuation, screens the consecutive paragraphs with the characteristics of vocalization aggregation, and obtains the vocalization dense distribution quantity; The facial stretching measurement sub-module analyzes the vertical displacement direction of the chin area in consecutive image frames according to the image segment covered by the vocalization density distribution amount, compares the direction consistency of the displacement trajectories between the starting frame and the ending frame, determines whether a continuous upward or downward trend is formed, filters the image segments with continuously unidirectional offset trajectories, and obtains the stretching trajectory offset degree. The behavior overlap detection sub-module calls the time period corresponding to the stretching trajectory offset degree, calculates the first and last occurrence frame labels of the blinking action in the frame sequence, performs a time-axis cross-judgment with the body vibration frame label interval within the same segment, filters the behavior channel data without an overlapping segment, and obtains the invalid behavior path data.

5. The pediatric pain assessment system according to claim 1, characterized in that, The periodic fluctuation extraction module includes: The rhythm curve construction sub-module calls the image frame segment covered by the invalid behavior path data, analyzes the facial action frame sequence, heart rate signal density, and muscle activity frequency band, calculates the corresponding change trajectory positions of the three types of signals in consecutive frames respectively, draws a continuous curve in chronological order, and obtains the signal rhythm distribution information. The peak distance change recognition sub-module calculates the frame sequence interval between consecutive peak points according to the change positions of the curve trajectories in the signal rhythm distribution information, compares the variation amplitude of adjacent intervals in consecutive frame segments, filters the frame segment sequences with key changes occurring continuously, and obtains the curve interval fluctuation trend. The abnormal fluctuation screening sub-module determines whether there are inconsistent change directions or delayed response distributions in the three types of curves of the face, heart rate, and muscles according to the time period marked by the curve interval fluctuation trend, analyzes the concentrated time periods of synchronous deviation in multiple segments of data, and obtains the rhythm mutation response section.

6. The pediatric pain assessment system according to claim 5, characterized in that, When analyzing the concentrated time periods of synchronous deviation in multiple segments of data, the formula is used: ； Calculate the synchronization deviation value , obtain the rhythm mutation response section, where represents the number of time periods represents the heart rate signal within the th time period, represents the mean of the heart rate signals for all time periods represents the standard deviation of the heart rate signals for all time periods represents the muscle signal within the th time period, represents the mean of the muscle signals for all time periods represents the standard deviation of the muscle signals for all time periods represents the time mark for the th time period, represents a non-zero constant used to smooth the square root term to prevent instability caused by zero 7. The pediatric pain assessment system according to claim 1, characterized in that, The behavior-physiology pair difference module includes: The response time extraction sub-module analyzes the starting points of consecutive sound waves in the vocalization trajectory according to the frame sequence range of the rhythm mutation response section, synchronously detects the first dense fluctuation section in the heart rate data, compares the arrangement order of the first frames of the two sections on the time axis, determines their sequence relationship and establishes a frame position corresponding item, and obtains the vocalization-heart rate starting offset. The time period difference calculation sub-module analyzes the continuous fluctuation sections corresponding to the cry and heart rate on the image frame axis based on the frame sequence position in the vocalization-heart rate starting offset, compares the time spans between the start and end frames of the two types of signals, filters the regions with relatively stable change directions of the frame segments, and obtains the behavior duration comparison amount. The signal consistency judgment sub-module compares the fluctuation directions in the rising section of the cry amplitude and the heart rate change trajectory according to the fluctuation frame segments within the behavior duration comparison amount, determines whether they show synchronous characteristics at the same time, filters the intervals with reverse changes or fluctuation breaks, and obtains the behavior-physiology misalignment section.

8. The pediatric pain assessment system according to claim 7, characterized in that, When comparing the fluctuation directions in the rising section of the cry amplitude and the heart rate change trajectory, the formula is used: ； Calculate the fluctuation synchronization index , and determine whether it shows synchronous characteristics within the same time. Among them, represents the amplitude in the rising segment of the -th frame of the cry amplitude, represents the heart rate in the heart rate change trajectory of the -th frame, represents the mean value of the amplitudes in the entire rising segment of the cry amplitude, represents the mean value of the heart rates in the entire heart rate change trajectory, represents the total number of frames in the current comparison frame segment, represents a non-zero constant used to smooth the square root term to prevent instability caused by zero.

9. The pediatric pain assessment system according to claim 1, characterized in that, The system further includes: The pain level determination module monitors whether there are changes in facial, behavioral, and physiological signals simultaneously within the time period based on the behavior-physiology misalignment section, identifies the overlapping regions of the three types of signals, analyzes whether they appear at the characteristic positions in the pain judgment criteria, determines the pain level classification of the corresponding regions, and obtains the multi-source pain classification regions. The multi-source pain grading region includes the multi-dimensional signal fusion result, the pain level classification number, and the reference standard mapping index.

10. The pediatric pain assessment system according to claim 9, wherein, The pain grading determination module includes: Based on the frame sequence interval covered by the behavioral and physiological misalignment section, the signal change detection sub-module analyzes the continuous expression contour changes in the facial image, the vocalization fluctuations in the sound track, and the starting points of the fluctuations in the heart rate curve, determines whether the signals generate linkage changes within the same frame segment range, and obtains the signal change synchronization amount; Based on the frame segment range involved in the signal change synchronization amount, the overlapping segment screening sub-module compares the starting trajectories of the facial, behavioral, and physiological signals at the same frame sequence, screens out the frame segment fragments that have intersections and continuously extend in the time sequence direction, and then counts the number of overlapping ranges in the continuous distribution segment to obtain the total amount of overlapping distribution segments; According to the signal coincidence structure and the frame segment duration characteristics in the total amount of overlapping distribution segments, the feature level matching sub-module analyzes whether their corresponding positions are located in the key sections set in the pain level reference sequence, determines the matching degree between the linkage form presented by the frame segment and the level standard, determines the corresponding pain degree grading, and obtains the multi-source pain grading region.

Images