An evaluation method and system based on angelman syndrome phenotype detection

Abstract

The present application relates to the technical field of wisdom monitoring, and more particularly to an evaluation method and system based on angelman syndrome phenotype detection, which comprises the following steps: S1: obtaining video data, audio data, spatial positioning, image data and social interaction data of a target object; S2: preliminarily classifying and dividing a smiling event of the target object according to the video data, audio data, spatial positioning, image data and social interaction data, and obtaining a preliminary unprovoked smiling event of the target object. Through the preliminary and deep two-level detection mechanism, high-precision identification of the unprovoked smiling event is realized; based on multiple indexes such as step symmetry, gait variation coefficient and dynamic balance margin, a normalization and weighted aggregation strategy is adopted to provide an intuitive quantitative gait abnormality score, thereby providing an accurate reference for clinical evaluation.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, and in particular to an assessment method and system based on Angelman syndrome phenotype detection. Background Technology

[0002] Angelman syndrome is a neurodevelopmental disorder caused by deletions or dysfunction in the 15q11–q13 region of chromosome 15. Its main clinical manifestations include intellectual disability, language impairment, motor coordination disorders, and a characteristic smile. Traditional diagnosis of Angelman syndrome relies on genetic testing and neuroimaging examinations; however, these methods are costly, have long testing cycles, and are difficult to implement for early and continuous monitoring. Smiling is considered one of the typical phenotypes of Angelman syndrome, but current research methods for excluding smile triggers largely rely on manual annotation or are limited by environmental control, failing to achieve automatic recognition of unprovoked smiles in natural settings. Gait abnormalities are also a key feature of the disorder, but traditional gait assessments are mostly based on high-precision laboratory equipment, making real-time monitoring in everyday environments difficult. Therefore, there is an urgent need for a method that can combine multimodal data in natural environments to perform real-time, automated, and non-invasive behavioral and motor phenotype detection on young children with Angelman syndrome who are capable of walking, and to comprehensively assess the multidimensional characteristics of smiles and gait. The "walking children" refers to those aged 1-6 years who are capable of walking. Summary of the Invention

[0003] To overcome the limitation of not being able to automatically recognize uninduced smiles in natural scenes, this invention provides an evaluation method and system based on Angelman syndrome phenotype detection.

[0004] Technical solution: An assessment method based on Angelman syndrome phenotype detection, comprising the following steps:

[0005] S1: Acquire video data, audio data, spatial positioning, image data, and social interaction data of the target object;

[0006] S2: Based on the video data, audio data, spatial positioning, image data, and social interaction data, perform preliminary classification of the target object's smile events to obtain preliminary uninduced smile events of the target object;

[0007] S3: Based on the video data, audio data, spatial positioning, image data and social interaction data, perform secondary detection on the smile events of the target object to obtain the deep uninduced smile events of the target object, and obtain all uninduced smile events based on the preliminary uninduced smile events and the deep uninduced smile events.

[0008] S4: Obtain the uninduced index or the preschool uninduced index based on the uninduced smiling events described above;

[0009] S5: Obtain the stride symmetry index, gait variation coefficient, and dynamic balance margin of the target object. Based on the stride symmetry index, gait variation coefficient, and dynamic balance margin, normalize and then perform gait weighted aggregation to obtain a comprehensive gait anomaly score.

[0010] S6: Based on the aforementioned non-inducible index or the infant non-inducible index, and combined with the aforementioned gait abnormality score, a weighted fusion is performed to obtain a comprehensive risk assessment score, and the target subject is given auxiliary diagnosis and treatment based on the comprehensive risk assessment score.

[0011] Preferably, the acquisition of video data, audio data, spatial positioning, image data, and social interaction data of the target object includes: the video data being used to detect the target object's smiling behavior and posture features; the audio data being used to detect the target object's laughter, language stimuli, and background sound sources; the spatial positioning and image data being used to determine the target object's current environmental semantics; and the social interaction data being used to perceive the interactive object and interaction quality through facial recognition, sound source quantity recognition, and Bluetooth sensing.

[0012] Preferably, the step of performing preliminary classification and division of the target object's smile events based on the video data, audio data, spatial positioning, image data, and social interaction data to obtain preliminary uninduced smile events of the target object includes:

[0013] Based on the target object's video data, audio data, spatial positioning, image data, and social interaction data;

[0014] S201: Classify the laughter of the target object as either imitative laughter or a response to external laughter, and classify it into positive stimulus audio and negative stimulus audio after detection based on the audio data, wherein the positive stimulus audio is an external sound signal that can induce or enhance the emotional response of the target object in the environment.

[0015] S202: Based on the spatial positioning and image data, perform semantic segmentation, and use an activity recognition model to classify the environment in which the target object is located into entertainment-attribute scenes and non-entertainment-attribute scenes;

[0016] S203: Obtain the identity of the interactive object and the quality of the interaction based on the video data and the social interaction data;

[0017] If the audio is not a positive stimulus, the scene is not entertainment-related, and there is no interactive object or the interaction quality is lower than the preset interaction threshold, then the target object's smile event will be marked as a preliminary uninduced smile event.

[0018] Preferably, the step of performing secondary detection on the target object's smile events based on the video data, audio data, spatial positioning, image data, and social interaction data to obtain the target object's deep uninduced smile events, and obtaining all uninduced smile events based on the preliminary uninduced smile events and deep uninduced smile events, includes: obtaining the target object's social interaction activity score, sound stimulus intensity score, and scene activity relevance score for each smile event based on the target object's video data, audio data, spatial positioning, image data, and social interaction data, and then performing a weighted sum to obtain a scene suitability score. The social interaction activity score is used to reflect the intensity and quality of the target object's interaction with outsiders during the smile event; the sound stimulus intensity score is used to measure the contribution of the sound environment to the emotional triggering in this smile event; the scene activity relevance score is used to assess the promoting or inhibiting effect of the environment in which this smile event occurs on the emotion. When the scene suitability score of the smile event is less than or equal to a preset background trigger threshold and a smile event occurs, the smile event is marked as a deep uninduced smile event, wherein the weighted summation formula is:

[0019] ;

[0020] In the formula, Score the suitability of the scene; Rate the level of social interaction activity; Rate the intensity of the sound stimulus; Rate the relevance of the scene activity; This is the weighting adjustment coefficient, and .

[0021] Preferably, obtaining the uninduced smile index or the preschool uninduced smile index based on the uninduced smile events includes: obtaining the total number of uninduced smile events within a preset time period, and obtaining the uninduced smile frequency based on the total number of events; obtaining the uninduced smile duration and average laughter decibel, and obtaining the uninduced smile index by performing normalized weighted summation based on the uninduced smile frequency, uninduced smile duration, and average laughter decibel; simultaneously obtaining the preschool uninduced smile index based on the preschool's average smile duration, average laughter volume, uninduced smile frequency, and physiological response intensity index during the uninduced smile events, and obtaining the preschool uninduced smile index by performing normalized weighted summation based on the preschool's average smile duration, average laughter volume, uninduced smile frequency, and physiological response intensity index.

[0022] Preferably, the step of obtaining the child's uninduced smile index by normalizing and weighting the sum of the child's average smile duration, average laughter volume, frequency of uninduced smiles, and physiological response intensity indicators includes: the physiological response intensity indicators include heart rate variability, instantaneous heart rate increase, and skin conductance response, which are scalars used to reflect the degree of activation of the child's autonomic nervous system before and after an uninduced smile event; the physiological response intensity index is obtained by linearly weighting the heart rate variability, instantaneous heart rate increase, and skin conductance response.

[0023] Preferably, the step of obtaining the stride symmetry index, gait variation coefficient, and dynamic balance margin of the target object, and then performing gait weighted aggregation after normalization based on the stride symmetry index, gait variation coefficient, and dynamic balance margin to obtain a comprehensive gait anomaly score, includes: obtaining the left and right stride lengths and stride intervals of the target object when walking;

[0024] A stride symmetry index is obtained by measuring the left and right strides of the target object. This stride symmetry index measures the degree of symmetry between the left and right sides of the stride. Obtain, among which Left foot length refers to the horizontal distance the body moves forward between two consecutive left foot strikes. The right foot length refers to the horizontal distance the body moves forward between two consecutive right foot strikes. This is an indicator of stride symmetry.

[0025] The gait variation coefficient is obtained through the stride interval during walking. This coefficient measures the stability of the stride interval and reflects the degree of fluctuation in gait rhythm. Obtain, among which This represents the average stride interval over a period of motion. This represents the standard deviation of the stride interval;

[0026] The pressure center position of the foot is obtained by a pressure sensor array, and the projection position of the body's center of gravity on the horizontal plane is calculated by an IMU or depth camera. The minimum distance between the two and the edge of the foot is calculated as the dynamic balance margin. The dynamic balance margin is used to reflect the safety margin between the projection of the center of gravity and the support area of ​​the foot during walking, and is used to assess the risk of falling.

[0027] Preferably, the step of normalizing and then weighting the gait to obtain a comprehensive gait anomaly score based on the stride symmetry index, gait variation coefficient, and dynamic balance margin includes: wherein the weighted gait aggregation is as follows:

[0028] ;

[0029] In the formula, Scoring for gait abnormalities; This is the normalized stride symmetry index; The normalized gait variation coefficient; This represents the normalized dynamic equilibrium margin. The weighting coefficients are used for the comprehensive gait abnormality score.

[0030] Preferably, the step of weightedly fusing the gait abnormality score with the no-inducement index or the infant no-inducement index to obtain a comprehensive risk assessment score, and then using the comprehensive risk assessment score to assist in the diagnosis and treatment of the target subject, includes: wherein

[0031] ;

[0032] In the formula The overall risk assessment score; The index is the index of no stimuli or the index of no stimuli for young children; Scoring for gait abnormalities; This is for adjusting the coefficient.

[0033] Preferably, an assessment system based on Angelman syndrome phenotype detection includes:

[0034] The data acquisition module is used to acquire video data, audio data, spatial positioning, image data, and social interaction data of the target object;

[0035] The preliminary judgment module is used to perform preliminary classification of the target object's smile events based on the video data, audio data, spatial positioning, image data, and social interaction data, and to obtain the target object's preliminary uninduced smile events.

[0036] The depth judgment module is used to perform secondary detection on the smile events of the target object based on the video data, audio data, spatial positioning, image data and social interaction data, to obtain the depth of the uninduced smile events of the target object, and to obtain all uninduced smile events based on the preliminary uninduced smile events and the depth of the uninduced smile events.

[0037] The incentive index acquisition module is used to obtain the non-incentive index or the child non-incentive index based on the non-incentive smile event.

[0038] The gait scoring module is used to acquire the stride symmetry index, gait variation coefficient, and dynamic balance margin of the target object. Based on the stride symmetry index, gait variation coefficient, and dynamic balance margin, the gait is normalized and then weighted and aggregated to obtain a comprehensive gait anomaly score.

[0039] The assessment assistance module is used to perform weighted fusion based on the non-inducible index or the infant non-inducible index, combined with the gait abnormality score, to obtain a comprehensive risk assessment score, and to provide auxiliary diagnosis and treatment for the target subject based on the comprehensive risk assessment score.

[0040] The beneficial effects of this invention are as follows:

[0041] 1. This invention utilizes multiple data sources, including video, audio, spatial positioning, images, and social interaction, to synchronously and completely collect multi-dimensional behavioral and motion data in natural scenes, thereby improving the comprehensiveness and reliability of the detection.

[0042] 2. Through a two-level detection mechanism of preliminary and deep detection, combined with semantic segmentation, audio context analysis and social interaction assessment, it can automatically filter out external stimuli and achieve high-precision recognition of unprovoked smile events.

[0043] 3. Based on multiple indicators such as stride symmetry, gait variation coefficient and dynamic balance margin, and using normalization and weighted aggregation strategies, it provides an intuitive and quantitative gait abnormality score, providing an accurate reference for clinical assessment;

[0044] 4. By integrating the Unexplained Influence Index or the Childhood Unexplained Influence Index with the Gait Abnormality Score, a comprehensive risk assessment score is obtained, providing clinicians and parents with a scientifically quantifiable auxiliary diagnostic basis. Based on the assessment results, personalized rehabilitation training or intervention plans can be developed to promote the effectiveness of early intervention. Attached Figure Description

[0045] Figure 1 This is a flowchart of an assessment method based on Angelman syndrome phenotype detection according to the present invention;

[0046] Figure 2 This is a schematic diagram of the structure of an assessment system based on Angelman syndrome phenotype detection according to the present invention. Detailed Implementation

[0047] The invention will now be described more fully below with reference to the accompanying drawings, in which presently preferred embodiments of the invention are illustrated. However, the invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided for thoroughness and completeness and to fully convey the scope of the invention to those skilled in the art.

[0048] Example 1: An assessment method based on Angelman syndrome phenotype detection, such as Figure 1 and Figure 2 As shown, it includes the following steps:

[0049] S1: Acquire video data, audio data, spatial positioning, image data, and social interaction data of the target object;

[0050] S2: Based on the video data, audio data, spatial positioning, image data, and social interaction data, perform preliminary classification of the target object's smile events to obtain preliminary uninduced smile events of the target object;

[0051] S3: Based on the video data, audio data, spatial positioning, image data and social interaction data, perform secondary detection on the smile events of the target object to obtain the deep uninduced smile events of the target object, and obtain all uninduced smile events based on the preliminary uninduced smile events and the deep uninduced smile events.

[0052] S4: Obtain the uninduced index or the preschool uninduced index based on the uninduced smiling events described above;

[0053] S5: Obtain the stride symmetry index, gait variation coefficient, and dynamic balance margin of the target object. Based on the stride symmetry index, gait variation coefficient, and dynamic balance margin, normalize and then perform gait weighted aggregation to obtain a comprehensive gait anomaly score.

[0054] S6: Based on the aforementioned non-inducible index or the infant non-inducible index, and combined with the aforementioned gait abnormality score, a weighted fusion is performed to obtain a comprehensive risk assessment score, and the target subject is given auxiliary diagnosis and treatment based on the comprehensive risk assessment score.

[0055] The video data is used to detect the target object's smiling behavior and posture features; the audio data is used to detect the target object's laughter, language stimuli, and background sound sources; the spatial positioning and image data are used to determine the target object's current environmental semantics; and the social interaction data is used to perceive the interactive object and interaction quality through facial recognition, sound source quantity recognition, and Bluetooth sensing.

[0056] First, terminal devices are deployed in the target object's activity environment. These devices include a high-definition camera, a microphone array, an indoor positioning module, a depth camera, and a Bluetooth beacon. Natural activity tracking of the target object is performed to acquire the following data: Video data: The front-end high-definition camera captures RGB video of the target object at 30fps to detect smile behavior and body posture features, and the depth camera acquires skeletal key points to assist in posture analysis; Audio data: The microphone array acquires multi-channel audio signals with a 16kHz sampling rate in real time to detect the target object's laughter spectrum features, language stimulus fragments, and the location of environmental background sound sources; Spatial positioning and image data: The indoor positioning module updates the target object's three-dimensional coordinates in real time based on UWB technology; simultaneously, panoramic images of the environment are captured, and the image data is segmented into indoor entertainment areas using a semantic segmentation model. The semantic units of the rest area are used to determine environmental attributes; social interaction data: facial recognition algorithms are used to identify the number and identity of surrounding faces, sound source number recognition technology is used to distinguish the number of interactive objects speaking at the same time, and the number of near-field devices detected by Bluetooth beacons is combined to evaluate the interaction quality; all the above data are time-series synchronized and preprocessed on local edge computing nodes to ensure that various modal data are aligned on the same time axis before being transmitted to the cloud analysis platform. When using facial recognition algorithms to identify surrounding people, the data collection scenario is a compliant scenario with the consent of the surrounding people. At the same time, after the data is captured, privacy data is deleted, only the general recognition results are retained, and neither the original biometric information is stored.

[0057] Based on the target object's video data, audio data, spatial positioning, image data, and social interaction data;

[0058] S201: Classify the laughter of the target object as either imitative laughter or a response to external laughter, and classify it into positive stimulus audio and negative stimulus audio after detection based on the audio data, wherein the positive stimulus audio is an external sound signal that can induce or enhance the emotional response of the target object in the environment.

[0059] S202: Based on the spatial positioning and image data, perform semantic segmentation, and use an activity recognition model to classify the environment in which the target object is located into entertainment-attribute scenes and non-entertainment-attribute scenes;

[0060] S203: Obtain the identity of the interactive object and the quality of the interaction based on the video data and the social interaction data;

[0061] If the audio is not a positive stimulus, the scene is not entertainment-related, and there is no interactive object or the interaction quality is lower than the preset interaction threshold, then the target object's smile event will be marked as a preliminary uninduced smile event.

[0062] Audio classification: Features of the laughter signal of the target object are extracted, including Mel frequency cepstral coefficients (MFCC) and instantaneous energy, and input into a pre-trained classification model to determine whether the laughter is "imitated laughter" or "response to external laughter". At the same time, environmental audio is classified into sound sources to distinguish between "positive stimulus audio" (such as laughter of others and toy music) and "negative stimulus audio". The potential emotional triggering effect of external sounds on the target object is judged by energy threshold.

[0063] Scene classification: Based on image data after spatial localization and semantic segmentation, an activity recognition model is used to classify the environment in which the target object is located, and output the probability distribution of entertainment-attribute scenes (such as game areas or play areas) and non-entertainment-attribute scenes (such as corridors or rest areas). The activity recognition model is a CNN model.

[0064] Social interaction assessment: The identity information of the interaction object is obtained through the face recognition results, the number of simultaneous interaction objects is obtained by combining the number of sound sources and Bluetooth beacon detection, and the interaction quality score is calculated based on the gaze duration between the interaction object and the target object and the distance change characteristics between the face and the camera.

[0065] If, in the same smile event, the audio category is "non-positive stimulus audio", the scene classification result is "non-entertainment attribute scene", and the number of detected interactive objects is zero or the interaction quality score is lower than a preset threshold, then the smile event is marked as a preliminary uninduced smile event.

[0066] Based on the target object's video data, audio data, spatial positioning, image data, and social interaction data, the social interaction activity score, sound stimulus intensity score, and scene activity relevance score for each smile event are obtained. These scores are then weighted and summed to obtain a scene suitability score. The social interaction activity score reflects the intensity and quality of the target object's interaction with others during the smile event. The sound stimulus intensity score measures the contribution of the sound environment to the emotional triggering of the smile event. The scene activity relevance score assesses the promoting or inhibiting effect of the environment on the emotion during the smile event. When the scene suitability score of a smile event is less than or equal to a preset background trigger threshold and a smile event occurs, the smile event is marked as a deep, uninduced smile event. The weighted summation formula is as follows:

[0067] ;

[0068] In the formula, Score the suitability of the scene; Rate the level of social interaction activity; Rate the intensity of the sound stimulus; Rate the relevance of the scene activity; This is the weighting adjustment coefficient, and .

[0069] How to obtain a social interaction activity score:

[0070] Interactive object detection: Using a front-end camera and face recognition module, face detection and tracking are performed on each frame of the image to obtain the number of interactive objects appearing in the field of view at the same time; combined with a sound source number recognition algorithm, the number of different sound sources that are making sounds at the same time is obtained; Bluetooth Low Energy beacon sensing is applied to obtain the number of near-field interactive devices to supplement the blind spots of face / sound source detection.

[0071] Interaction quality feature extraction: gaze duration: for each interactive object, the cumulative gaze duration of the target object is counted; facing angle: using skeletal key points or head pose estimation, the average deflection angle of the target object facing each interactive object is calculated, the smaller the angle, the higher the interaction attention; voice dialogue turn: based on speaker separation and voice activity detection, the number of turns between the target object and each interactive object is counted; spatial proximity: based on the indoor positioning system, the average distance between the target object and each interactive object is obtained, the closer the distance, the greater the interaction intensity; the above features are mapped to the [0,1] interval, and weighted summation is performed according to the weight coefficients determined by experience or cross-validation to obtain the social interaction activity score;

[0072] Obtaining the sound stimulus intensity score:

[0073] Sound source event segmentation: Based on the VAD algorithm, the continuous audio stream is framed to identify different types of sound source events, such as laughter, talking, and ambient sounds; features are extracted for each sound source event: Mel-frequency cepstral coefficients, sound pressure level, or short-time energy; positive stimulus source weighting: different sound source types are assigned preset weights (e.g., laughter or game sounds are high positive stimuli, background noise such as air conditioner sounds are low positive stimuli), denoted as... Instantaneous sound pressure level: Sound pressure level calculated for each frame. and mapped to : For all frames covered by a single smile event =1…K, calculate the sound stimulus intensity score for this event using the following formula: ; and will Mapped to ;

[0074] Obtaining the scenario activity relevance score:

[0075] Using depth cameras or panoramic cameras, semantic segmentation is performed on scene images to label different semantic units such as "entertainment area," "walking corridor," and "rest area," and the segmentation results are mapped to pixel-level label maps. ,in Given the total number of semantic categories, based on skeletal keypoint sequences or optical flow features, a pre-trained action recognition model (such as 3D-CNN or Transformer) is used to classify the current activity of the target object, outputting an activity probability distribution vector. Based on the activity level of each semantic category, and according to expert experience or clinical research, each semantic unit is assigned a weight that promotes or inhibits smiles. Extract the semantic label corresponding to the pixel position of the segmented target object. Combined with activity recognition probability With semantic weight Calculate the relevance score of the scene activity: and will Mapped to ;

[0076] When the scene suitability score of a smile event is less than or equal to the preset background trigger threshold and a smile event occurs, the smile event is marked as an unincentivized smile.

[0077] The system obtains the total number of uninduced smile events within a preset time period, and calculates the uninduced smile frequency based on the total number of events. It also obtains the duration of uninduced smiles and the average decibel level of laughter, and performs a normalized weighted sum based on the frequency, duration, and decibel level of uninduced smiles to obtain an uninduced smile index. Simultaneously, it obtains the average smile duration, average laughter volume, frequency, and physiological response intensity of children during uninduced smile events, and performs a normalized weighted sum based on these parameters to obtain the child's uninduced smile index.

[0078] The Uninduced Smile Index (ULI) is designed for a general population, including individuals of all ages, from adults and adolescents to the elderly. Its indicator system focuses on the basic characteristics of smile events, such as frequency, total duration, and average loudness, and can reflect the overall level of spontaneous smiles in natural environments. The ULI is designed without introducing physiological signals, making it easy to complete the assessment using only conventional video and audio equipment. The Uninduced Smile Index for Toddlers is specifically designed for toddlers (usually children aged 0-3 years). It adds the key dimension of physiological response intensity to the ULI to capture the degree of activation of the autonomic nervous system (such as heart rate variability, instantaneous heart rate increase, and skin conductance response) before and after smile events. Since toddlers' language abilities and social expressions are not yet mature, the physiological accompanying signals of their smiles are particularly important for diagnostic value, so it needs to be constructed separately.

[0079] The physiological response intensity index includes heart rate variability, instantaneous heart rate increase, and skin conductance response. It is a scalar measure reflecting the degree of activation of the autonomic nervous system of young children before and after an unprovoked smiling event. The physiological response intensity index is obtained by linearly weighting the heart rate variability, instantaneous heart rate increase, and skin conductance response.

[0080] Data was simultaneously acquired from a heart rate monitor and a skin conductance sensor worn by the target infant. The heart rate monitor output R-R interval sequences to calculate heart rate variability (HRV) and instantaneous heart rate increase (ΔHR); the skin conductance sensor recorded skin conductance response (SCR).

[0081] Calculation of raw values ​​of physiological response intensity:

[0082] Heart rate variability (HRV): The standard deviation SDNN is calculated for the R–R interval sequence to obtain... ;

[0083] Instantaneous heart rate increase (ΔHR): The average heart rate within 5 seconds prior to the smile event. Average heart rate within 1 second before and after the peak of smile ,calculate ;

[0084] Skin conductance response (SCR): Extracting the maximum peak amplitude during the event from the SCR waveform. As ;

[0085] right After normalization, a weighted sum is obtained to obtain the physiological response intensity index.

[0086] Obtain the left and right stride lengths and stride intervals of the target object while walking;

[0087] A stride symmetry index is obtained by measuring the left and right strides of the target object. This stride symmetry index measures the degree of symmetry between the left and right sides of the stride. Obtain, among which Left foot length refers to the horizontal distance the body moves forward between two consecutive left foot strikes. The right foot length refers to the horizontal distance the body moves forward between two consecutive right foot strikes. This is an indicator of stride symmetry.

[0088] The gait variation coefficient is obtained through the stride interval during walking. This coefficient measures the stability of the stride interval and reflects the degree of fluctuation in gait rhythm. Obtain, among which This represents the average stride interval over a period of motion. This represents the standard deviation of the stride interval;

[0089] The pressure center position of the foot is obtained by a pressure sensor array, and the projection position of the body's center of gravity on the horizontal plane is calculated by an IMU or depth camera. The minimum distance between the two and the edge of the foot is calculated as the dynamic balance margin. The dynamic balance margin is used to reflect the safety margin between the projection of the center of gravity and the support area of ​​the foot during walking, and is used to assess the risk of falling.

[0090] Left and right stride acquisition:

[0091] The pressure distribution of each metatarsal landing is recorded using a pressure sensor array. The left foot length L is calculated by detecting the change in the position of the left foot landing between two consecutive left foot landings. Similarly, the right foot length R is obtained by calculating the change in the position of the right foot landing between two consecutive right foot landings. This process continues during walking, and a pair of (L,R) data is obtained after each complete left-right alternation gait.

[0092] The gait-weighted aggregation is as follows:

[0093] ;

[0094] In the formula, Scoring for gait abnormalities; This is the normalized stride symmetry index; The normalized gait variation coefficient; This represents the normalized dynamic equilibrium margin. The weighting coefficients are used for the comprehensive gait abnormality score.

[0095] For normalized indicators , , After performing min-max normalization, we obtain The larger the DSM, the higher the balance margin and the lower the risk of anomalies; conversely, the smaller the DSM, the higher the risk.

[0096] in

[0097] ;

[0098] In the formula The overall risk assessment score; The index is the index of no stimuli or the index of no stimuli for young children; Scoring for gait abnormalities; This is for adjusting the coefficient.

[0099] Example 2: Based on Example 1, an assessment system based on Angelman syndrome phenotype detection, such as... Figure 2 As shown, it includes:

[0100] The data acquisition module is used to acquire video data, audio data, spatial positioning, image data, and social interaction data of the target object;

[0101] The preliminary judgment module is used to perform preliminary classification of the target object's smile events based on the video data, audio data, spatial positioning, image data, and social interaction data, and to obtain the target object's preliminary uninduced smile events.

[0102] The depth judgment module is used to perform secondary detection on the smile events of the target object based on the video data, audio data, spatial positioning, image data and social interaction data, to obtain the depth of the uninduced smile events of the target object, and to obtain all uninduced smile events based on the preliminary uninduced smile events and the depth of the uninduced smile events.

[0103] The incentive index acquisition module is used to obtain the non-incentive index or the child non-incentive index based on the non-incentive smile event.

[0104] The gait scoring module is used to acquire the stride symmetry index, gait variation coefficient, and dynamic balance margin of the target object. Based on the stride symmetry index, gait variation coefficient, and dynamic balance margin, the gait is normalized and then weighted and aggregated to obtain a comprehensive gait anomaly score.

[0105] The assessment assistance module is used to perform weighted fusion based on the non-inducible index or the infant non-inducible index, combined with the gait abnormality score, to obtain a comprehensive risk assessment score, and to provide auxiliary diagnosis and treatment for the target subject based on the comprehensive risk assessment score.

[0106] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An assessment method based on Angelman syndrome phenotype detection, characterized in that, Includes the following steps: S1: Acquire video data, audio data, spatial positioning, image data, and social interaction data of the target object; S2: Based on the video data, audio data, spatial positioning, image data, and social interaction data, perform preliminary classification of the target object's smile events to obtain preliminary uninduced smile events of the target object; S3: Based on the video data, audio data, spatial positioning, image data and social interaction data, perform secondary detection on the smile events of the target object to obtain the deep uninduced smile events of the target object, and obtain all uninduced smile events based on the preliminary uninduced smile events and the deep uninduced smile events. S4: Obtain the uninduced index or the preschool uninduced index based on the uninduced smiling events described above; S5: Obtain the stride symmetry index, gait variation coefficient, and dynamic balance margin of the target object. Based on the stride symmetry index, gait variation coefficient, and dynamic balance margin, normalize and then perform gait weighted aggregation to obtain a comprehensive gait anomaly score. S6: Based on the aforementioned no-inducement index or the infant no-inducement index, and combined with the aforementioned gait abnormality score, a weighted fusion is performed to obtain a comprehensive risk assessment score, and the target subject is given auxiliary diagnosis and treatment based on the comprehensive risk assessment score; The step of performing a preliminary classification of the target object's smile events based on the video data, audio data, spatial positioning, image data, and social interaction data to obtain preliminary uninduced smile events of the target object includes: Based on the target object's video data, audio data, spatial positioning, image data, and social interaction data; S201: Classify the target object's laughter as either imitative laughter or a response to external laughter, and classify it into positive stimulus audio and negative stimulus audio after detection based on the audio data, wherein the positive stimulus audio is an external sound signal that can trigger or enhance the target object's emotional response in the environment; S202: Based on the spatial positioning and image data, perform semantic segmentation, and use an activity recognition model to classify the environment in which the target object is located into entertainment-attribute scenes and non-entertainment-attribute scenes; S203: Obtain the identity of the interactive object and the quality of the interaction based on the video data and the social interaction data; If the audio is not positively stimulating, the scene is not entertainment-related, and there is no interactive object or the interaction quality is lower than the preset interaction threshold, then the target object's smile event will be marked as a preliminary uninduced smile event. The process involves secondary detection of the target object's smile events based on the video data, audio data, spatial positioning, image data, and social interaction data to obtain deep, uninduced smile events. Based on the initial and deep uninduced smile events, all uninduced smile events are obtained, including: obtaining a social interaction activity score, a sound stimulus intensity score, and a scene activity relevance score for each smile event based on the target object's video data, audio data, spatial positioning, image data, and social interaction data; and then weighted summing these scores to obtain a scene suitability score. The social interaction activity score reflects the intensity and quality of the target object's interaction with others during the smile event; the sound stimulus intensity score measures the contribution of the sound environment to the emotional triggering of the smile event; and the scene activity relevance score assesses the promoting or inhibiting effect of the environment on the emotion during the smile event. When the scene suitability score of a smile event is less than or equal to a preset background trigger threshold and a smile event occurs, the smile event is marked as a deep uninduced smile event. The weighted summation formula is as follows: ； In the formula, Score the suitability of the scene; Rate the level of social interaction activity; Rate the intensity of the sound stimulus; Rate the relevance of the scene activity; This is the weighting adjustment coefficient, and ; The step of obtaining the uninduced smile index or the preschool uninduced smile index based on the uninduced smile events includes: obtaining the total number of uninduced smile events within a preset time period, and obtaining the uninduced smile frequency based on the total number of events; obtaining the uninduced smile duration and average laughter decibel, and obtaining the uninduced smile index by performing normalized weighted summation based on the uninduced smile frequency, uninduced smile duration, and average laughter decibel; simultaneously obtaining the preschool uninduced smile index based on the preschool's average smile duration, average laughter volume, uninduced smile frequency, and physiological response intensity index during the uninduced smile events, and obtaining the preschool uninduced smile index by performing normalized weighted summation based on the preschool's average smile duration, average laughter volume, uninduced smile frequency, and physiological response intensity index. The steps involve acquiring the stride symmetry index, gait variation coefficient, and dynamic balance margin of the target object, and then normalizing and weighting the stride symmetry index, gait variation coefficient, and dynamic balance margin to obtain a comprehensive gait anomaly score, including: acquiring the left and right stride lengths and stride intervals of the target object while walking. A stride symmetry index is obtained by measuring the left and right strides of the target object. This stride symmetry index measures the degree of symmetry between the left and right sides of the stride. Obtain, among which Left foot length refers to the horizontal distance the body moves forward between two consecutive left foot strikes. The right foot length refers to the horizontal distance the body moves forward between two consecutive right foot strikes. This is an indicator of stride symmetry. The gait variation coefficient is obtained through the stride interval during walking. This coefficient measures the stability of the stride interval and reflects the degree of fluctuation in gait rhythm. Obtain, among which This represents the average stride interval over a period of motion. This represents the standard deviation of the stride interval; The pressure center position of the foot is obtained by a pressure sensor array, and the projection position of the body's center of gravity on the horizontal plane is calculated by an IMU or depth camera. The minimum distance between the two and the edge of the foot is calculated as the dynamic balance margin. The dynamic balance margin is used to reflect the safety margin between the projection of the center of gravity and the support area of ​​the foot during walking, and is used to assess the risk of falling.

2. The assessment method based on Angelman syndrome phenotype detection according to claim 1, characterized in that, The acquisition of video data, audio data, spatial positioning, image data, and social interaction data of the target object includes: the video data being used to detect the target object's smiling behavior and posture features; the audio data being used to detect the target object's laughter, language stimuli, and background sound sources; the spatial positioning and image data being used to determine the target object's current environmental semantics; and the social interaction data being used to perceive the interactive object and interaction quality through facial recognition, sound source quantity recognition, and Bluetooth sensing.

3. The assessment method based on Angelman syndrome phenotype detection according to claim 1, characterized in that, The method of obtaining the child's uninduced smile index by normalizing and weighting the sum of the child's average smile duration, average laughter volume, frequency of uninduced smiles, and physiological response intensity indicators includes: the physiological response intensity indicators include heart rate variability, instantaneous heart rate increase, and skin conductance response, which are scalars used to reflect the degree of activation of the child's autonomic nervous system before and after an uninduced smile event; the physiological response intensity index is obtained by linearly weighting the heart rate variability, instantaneous heart rate increase, and skin conductance response.

4. The assessment method based on Angelman syndrome phenotype detection according to claim 1, characterized in that, The comprehensive gait anomaly score is obtained by normalizing and then weighting the gait based on the stride symmetry index, gait variation coefficient, and dynamic balance margin, including: where the weighted gait aggregation is: ； In the formula, Scoring for gait abnormalities; This is the normalized stride symmetry index; The normalized gait variation coefficient; This represents the normalized dynamic equilibrium margin. The weighting coefficients are used for the comprehensive gait abnormality score.

5. The assessment method based on Angelman syndrome phenotype detection according to claim 1, characterized in that, The step involves weighted fusion of the non-inducible index or the infant non-inducible index with the gait abnormality score to obtain a comprehensive risk assessment score, and then using this comprehensive risk assessment score to provide auxiliary diagnosis and treatment for the target individual, including: wherein... ； In the formula The overall risk assessment score; The index is the index of no stimuli or the index of no stimuli for young children; Scoring for gait abnormalities; This is for adjusting the coefficient.

6. An assessment system based on Angelman syndrome phenotype detection, used to implement the assessment method based on Angelman syndrome phenotype detection as described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to acquire video data, audio data, spatial positioning, image data, and social interaction data of the target object; The preliminary judgment module is used to perform preliminary classification of the target object's smile events based on the video data, audio data, spatial positioning, image data, and social interaction data, and to obtain the target object's preliminary uninduced smile events. The depth judgment module is used to perform secondary detection on the smile events of the target object based on the video data, audio data, spatial positioning, image data and social interaction data, to obtain the depth of the uninduced smile events of the target object, and to obtain all uninduced smile events based on the preliminary uninduced smile events and the depth of the uninduced smile events. The incentive index acquisition module is used to obtain the non-incentive index or the child non-incentive index based on the non-incentive smile event. The gait scoring module is used to acquire the stride symmetry index, gait variation coefficient, and dynamic balance margin of the target object. Based on the stride symmetry index, gait variation coefficient, and dynamic balance margin, the gait is normalized and then weighted and aggregated to obtain a comprehensive gait anomaly score. The assessment assistance module is used to perform weighted fusion based on the non-inducible index or the infant non-inducible index, combined with the gait abnormality score, to obtain a comprehensive risk assessment score, and to provide auxiliary diagnosis and treatment for the target subject based on the comprehensive risk assessment score.

Images