Search

CN-122024298-A - Special children emotion recognition and suitability content recommendation image analysis system

CN122024298ACN 122024298 ACN122024298 ACN 122024298ACN-122024298-A

Abstract

The application discloses a special child emotion recognition and suitability content recommendation image analysis system which comprises a basic feature quantification module, a correlation analysis evaluation module, an initial pressure calculation module, a scene risk monitoring module, a comprehensive matching decision module and a feedback control execution module, wherein the basic feature quantification module is used for acquiring a real-time facial image and depth point cloud of a child to obtain a facial micro-expression defect value, the correlation analysis evaluation module is used for acquiring an acousto-optic interference index of the current environment of the special child, carrying out space-time correlation analysis on the acousto-optic interference index and the real-time facial image to obtain an emotion fluctuation abnormal value, the initial pressure calculation module is used for acquiring a real-time emotion pressure initial value according to the facial micro-expression defect value and the emotion fluctuation abnormal value, the scene risk monitoring module is used for acquiring current data of the special child and combining parameters of the current content to obtain a content rejection risk value, the comprehensive matching decision module is used for obtaining a content adaptation degree evaluation value through the real-time emotion pressure initial value and the content rejection risk value, and the feedback control execution module is used for executing corresponding content switching according to a comparison result of the content adaptation degree evaluation value and a preset threshold.

Inventors

  • ZHANG ZHONGLIANG

Assignees

  • 金乡县特殊教育学校

Dates

Publication Date
20260512
Application Date
20260128

Claims (9)

  1. 1. A special child emotion recognition and suitability content recommendation image analysis system, comprising: The basic feature quantization module is used for acquiring real-time facial images and depth point cloud data of special children in real time through the camera, and analyzing and obtaining facial micro-expression defect values according to the real-time facial images and the depth point cloud data; The association analysis evaluation module is used for acquiring the acousto-optic interference index of the current environment of the special child, and carrying out space-time association analysis on the acousto-optic interference index and the face real-time image so as to acquire the abnormal value of the emotion fluctuation; the initial pressure calculation module is used for carrying out weighted calculation according to the facial microexpressive defect value and the emotion fluctuation abnormal value to obtain a real-time emotion pressure initial value; The scene risk monitoring module is used for acquiring current behavior participation data and historical emotion tolerance threshold values of special children in real time, and analyzing and acquiring content rejection risk values by combining parameters of the current teaching content intensity; The comprehensive matching decision module is used for carrying out difference operation on the real-time emotion pressure initial value and the content rejection risk value to obtain a content adaptation degree evaluation value; and the feedback control execution module is used for automatically executing corresponding content switching according to the comparison result of the content adaptation degree evaluation value and a preset threshold value.
  2. 2. The special children emotion recognition and suitability content recommendation image analysis system of claim 1, wherein the specific analysis steps of the facial microexpressive defect values are as follows: extracting values of pupil diameter, eyebrow spacing and mouth angle offset in the facial image; Comparing the extracted numerical value with the reference facial features of the child in a calm state, and calculating the deviation ratio of each feature; and carrying out linear weighted calculation on each offset ratio to obtain a facial micro-expression defect value.
  3. 3. The special child emotion recognition and suitability content recommendation image analysis system according to claim 1, wherein the emotion fluctuation abnormal value acquisition step is: Constructing an emotion classification model based on a convolutional neural network, collecting a facial image dataset containing multiple emotions of a special child, dividing the dataset into a training set and a testing set, and performing iterative training to obtain the emotion classification model with highest emotion analysis accuracy; collecting acousto-optic interference indexes of the current environment as compensation parameters, constructing multi-mode input feature vectors together with the real-time images of the faces of the special children, inputting the multi-mode input feature vectors into a emotion classification model, outputting positive and negative probability distribution of the emotion of the special children through the model, and obtaining abnormal values for calculating emotion fluctuation.
  4. 4. The special child emotion recognition and suitability content recommendation image analysis system according to claim 1, wherein the calculation step of the real-time emotion pressure initial value is: Carrying out normalization processing on the facial microexpressive defect value and the emotion fluctuation abnormal value data; setting the facial feature weight as Setting the environment association weight as And (2) and ; And carrying out summation operation on the product of the normalized facial micro-expression defect value and the facial feature weight and the product of the emotion fluctuation abnormal value and the environment association weight to obtain a real-time emotion pressure initial value.
  5. 5. The special child emotion recognition and suitability content recommendation image analysis system according to claim 1, wherein the content exclusion risk value acquisition step is: the method comprises the steps of monitoring the frequency of the child's sight leaving a screen and the frequency of the limbs swinging greatly in real time, dividing the frequency by corresponding behavior intervention thresholds respectively, and obtaining behavior risk factors; The historical rehabilitation file of the child is called, the average tolerance duration of the child to the current teaching strength is obtained, the ratio of the current duration to the tolerance duration is calculated, and the time risk factor is obtained; and carrying out weighted summation on the behavior risk factors and the time risk factors to obtain a content rejection risk value.
  6. 6. The system for analyzing the emotion recognition and suitability content recommendation image of a specific child according to claim 5, wherein the behavior intervention threshold is set by calling an average behavior fluctuation rule of the group from an expert knowledge base according to the age group and pathology classification of the specific child, and setting the average behavior fluctuation rule as the behavior intervention threshold.
  7. 7. The special children emotion recognition and suitability content recommendation image analysis system of claim 1, wherein automatically performing the corresponding content switching specifically comprises the steps of: if the content adaptation degree evaluation value is greater than or equal to a preset threshold value, maintaining the current teaching content; And if the content adaptation degree evaluation value is smaller than a preset threshold value, automatically extracting the matched audio and video content from a preset pacifying library for switching.
  8. 8. The special child emotion recognition and suitability content recommendation image analysis system of claim 1, wherein the acousto-optic interference index comprises an acoustic index and an optical index; Acquiring indoor static sound data and illumination data before acquiring the acoustic index and the optical index, wherein the sound baseline and the illumination baseline are respectively acquired by the acoustic index and the optical index; the acoustic offset is the difference between the instantaneous sound intensity and the sound baseline, and the optical offset is the difference between the illumination data and the illumination baseline.
  9. 9. The special child emotion recognition and suitability content recommendation image analysis system of claim 8, wherein the acousto-optic interference indicator The calculation formula of (2) is as follows: Wherein, the method comprises the steps of, ; As an acoustic interference factor (as an acoustic interference factor), , The instantaneous sound intensity is collected for the microphone array, As a baseline for the sound of the person, Is a decibel tolerance limit threshold for a particular child; As an optical interference factor, a light source is used, , As the difference between the instantaneous illuminance and the illumination baseline, Is the illumination variation frequency coefficient.

Description

Special children emotion recognition and suitability content recommendation image analysis system Technical Field The application relates to the technical field of content recommendation, in particular to a special child emotion recognition and suitability content recommendation image analysis system. Background Special children (such as autism spectrum disorder, attention deficit hyperactivity disorder and the like) have obvious differences in emotion expression, perception and regulation, and education and teaching processes of the special children have extremely high requirements on personalized and dynamically adapted intervention strategies. In the prior art, emotion recognition assistance systems rely mostly on single modality visual analysis. For example, a child's facial expression is captured by a camera, and a pre-trained generic Facial Expression Recognition (FER) model is utilized to determine that it belongs to basic emotion categories such as happiness, sadness, anger, etc., and accordingly trigger simple preset feedback. However, the scheme has obvious defects that firstly, the recognition accuracy of the general model on the special children is often lower, because the facial microexpressive features of the special children are often different from those of the typical nerve development crowd and the expression is more darkly and changeable, secondly, the existing scheme only reacts according to the instantaneous emotional state, ignores the direct influence of environmental interference (such as burst noise and light change) on the emotional fluctuation of the children, and does not consider the dynamic relation between the historical behavior mode of the children and the current teaching task intensity, so that the recommended content is often untimely, even anxiety or interference of the children can be caused, and the intervention effect is limited. Disclosure of Invention The present application aims to solve at least one of the technical problems in the related art to some extent. To achieve the above object, an embodiment of a first aspect of the present application provides a special child emotion recognition and suitability content recommendation image analysis system, including: The basic feature quantization module is used for acquiring real-time facial images and depth point cloud data of special children in real time through the camera, and analyzing and obtaining facial micro-expression defect values according to the real-time facial images and the depth point cloud data; The association analysis evaluation module is used for acquiring the acousto-optic interference index of the current environment of the special child, and carrying out space-time association analysis on the acousto-optic interference index and the face real-time image so as to acquire the abnormal value of the emotion fluctuation; the initial pressure calculation module is used for carrying out weighted calculation according to the facial microexpressive defect value and the emotion fluctuation abnormal value to obtain a real-time emotion pressure initial value; The scene risk monitoring module is used for acquiring current behavior participation data and historical emotion tolerance threshold values of special children in real time, and analyzing and acquiring content rejection risk values by combining parameters of the current teaching content intensity; The comprehensive matching decision module is used for carrying out difference operation on the real-time emotion pressure initial value and the content rejection risk value to obtain a content adaptation degree evaluation value; and the feedback control execution module is used for automatically executing corresponding content switching according to the comparison result of the content adaptation degree evaluation value and a preset threshold value. In addition, the special children emotion recognition and suitability content recommendation image analysis system provided by the application can also have the following additional technical characteristics: In one embodiment of the present application, the specific analysis steps of the facial micro-expression defect value are: extracting values of pupil diameter, eyebrow spacing and mouth angle offset in the facial image; Comparing the extracted numerical value with the reference facial features of the child in a calm state, and calculating the deviation ratio of each feature; and carrying out linear weighted calculation on each offset ratio to obtain a facial micro-expression defect value. In one embodiment of the present application, the obtaining step of the mood swings outlier includes: Constructing an emotion classification model based on a convolutional neural network, collecting a facial image dataset containing multiple emotions of a special child, dividing the dataset into a training set and a testing set, and performing iterative training to obtain the emotion classification model with highest emoti