An Online Learning Emotional Intervention System for the Elderly Based on Sentiment Analysis

Abstract

This application discloses an emotion intervention system for online learning among the elderly based on emotion analysis, including a recording module, a face recognition module, an emotion feature analysis module, a time-based emotion mapping module, a behavior dynamic transfer module, and an intervention strategy module. The recording module automatically records facial videos of the elderly during online learning to obtain facial emotion data. The face recognition module performs statistical analysis on the facial emotion data of all elderly individuals. This application belongs to the field of emotion intervention technology, and its purpose is to solve the problem that existing technologies cannot accurately capture facial expression data of the elderly in online learning scenarios, leading to low accuracy in subsequent emotion recognition and an inability to provide effective intervention measures. The achieved technical effect is: facilitating accurate capture of facial expression data of the elderly in online learning scenarios, improving the accuracy of subsequent emotion recognition, and enabling the provision of effective intervention measures.

Description

Technical Field

[0001] This application belongs to the field of emotion intervention technology, specifically relating to an online learning emotion intervention system for the elderly based on emotion analysis. Background Technology

[0002] Old age is a period when emotional needs are most prominent. Online learning for the elderly aims to enhance their emotional experience and social participation. Online learning provides a natural setting for large-scale analysis of the emotions of older adults, and emotion is a key focus in gerontology and online learning for the elderly. Research shows that online learning can significantly improve the emotional state of older adults, increase their subjective well-being, positively impact their mental health, and alleviate depressive symptoms.

[0003] The effectiveness of online learning for the elderly lies in two aspects: firstly, the acquisition of knowledge and skills increases their opportunities for self-actualization; secondly, it provides them with pathways to social participation, becoming an important channel for emotional support. For elderly individuals experiencing emotional deprivation or exhibiting depressive tendencies, designing and implementing targeted educational interventions can help improve their mental health. Therefore, affective analysis occupies an important position in gerontology and online learning for the elderly, contributing to the precision of curriculum design and providing a basis for interventions addressing negative emotions in the elderly. Analyzing the emotions of the elderly and conducting educational interventions based on affective computing technology is of contemporary significance in improving negative emotions among the elderly.

[0004] Facial expressions, as a natural channel for human emotional expression, provide relatively authentic and easily accessible information, thus becoming a commonly used data modality in affective computing. The natural scenarios of online learning can objectively reflect the true emotional state of the elderly, providing a convenient channel for large-scale capture and analysis of their facial expressions.

[0005] However, most current online learning emotion intervention systems for the elderly cannot accurately capture facial expression data of the elderly in online learning scenarios, resulting in low accuracy in subsequent emotion recognition, making it impossible to accurately judge the emotional state of the elderly and thus unable to provide effective intervention measures. Summary of the Invention

[0006] This application provides an emotion intervention system for online learning for the elderly based on emotion analysis, aiming to solve the problem that existing technologies cannot accurately capture facial expression data of the elderly in online learning scenarios, resulting in low accuracy of subsequent emotion recognition and thus failing to provide effective intervention measures.

[0007] An online learning and emotional intervention system for the elderly based on sentiment analysis includes a recording module, a face recognition module, a sentiment feature analysis module, a time-sensory sentiment mapping module, a behavior dynamic transfer module, and an intervention strategy module.

[0008] The recording module is used to automatically record facial videos of the elderly during online learning and obtain facial emotional data;

[0009] The facial recognition module is used to perform statistical analysis on the facial emotion data of all elderly people, and to analyze the average frequency of facial emotions that appear in the elderly people in online learning scenarios.

[0010] The emotion feature analysis module is used to analyze the structural features of facial expressions of the elderly using machine learning data mining methods.

[0011] The time-emotional graph drawing module is used to draw a smooth line graph reflecting the emotional changes of the elderly over time. The time-emotional graph drawing module includes a data preprocessing unit.

[0012] The behavior dynamic transfer module is used to perform lag sequence analysis on the facial expression behavior of the elderly, obtain a behavior transfer frequency table and a Z-score table, and derive the negative emotions of the elderly and their transfer patterns.

[0013] The intervention strategy module is used to determine the richness of emotional and behavioral transfer in different types of elderly people based on the Z-score table, and automatically match and push differentiated intervention content based on the richness of transfer.

[0014] Furthermore, the recording module has a recording duration of 1 hour during the online learning process, and allows interactive behaviors such as pausing and jumping back and forth during the online learning process.

[0015] Furthermore, the face recognition module is built on FaceReader software, and the training database of FaceReader software contains facial expression data of the elderly and East Asians.

[0016] Furthermore, the machine learning data mining method is cluster analysis, which is used to calculate the sum of clustering errors for all samples of facial emotion data of the elderly.

[0017] Furthermore, the smoothed line chart is drawn using detailed facial expression recognition data exported from FaceReade software, employing the Python language and the Matplotlib library.

[0018] Furthermore, the data preprocessing unit is used to normalize and remove noise from the facial emotion data.

[0019] Furthermore, the intervention strategy module includes a preset unit and a matching unit. The preset unit is used to define the richness classification of emotional behavior transfer based on the Z-score table.

[0020] Furthermore, the matching unit automatically identifies the emotional types of the elderly based on a preset classification of the richness of emotional behavior transfer, and accurately matches the corresponding intervention content.

[0021] Compared with the prior art, this application has at least the following beneficial effects:

[0022] Based on further analysis and research into existing technological problems, this application utilizes a collaborative recording module and a facial recognition module to accurately capture facial expression data of elderly individuals in online learning scenarios. By using this facial data as a key information source and conducting in-depth analysis, it reveals the emotional distribution patterns of the elderly. These patterns not only provide a new perspective for understanding the emotional state of the elderly during the learning process but also help to deliver corresponding emotional intervention strategies, thereby more effectively improving the impact of emotional support. Attached Figure Description

[0023] Figure 1 A schematic diagram of a module of an online learning and emotional intervention system for the elderly based on emotion analysis, provided as an embodiment of this application;

[0024] Figure 2 A schematic diagram illustrating the SSE value changes of an online learning emotion intervention system for the elderly based on emotion analysis, provided as an embodiment of this application; Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments.

[0026] Example 1

[0027] like Figure 1 and Figure 2 As shown, this application provides an online learning emotion intervention system for the elderly based on emotion analysis, including a recording module, a face recognition module, an emotion feature analysis module, a time-emotional mapping module, a behavior dynamic transfer module, and an intervention strategy module.

[0028] The recording module automatically records facial videos of seniors during online learning to capture facial emotional data. Recording time is one hour to avoid fatigue. Seniors are allowed to pause, skip ahead, and navigate during self-study, with their facial expressions recorded via computer webcam.

[0029] The study chose online learning for seniors as its research scenario for several reasons. Firstly, online learning is an important and common method of education for the elderly. Secondly, online learning has less interference from human factors, allowing for the capture of more natural emotional expressions from seniors. Finally, online methods enable large-scale and automated data collection, facilitating intelligent analysis.

[0030] The facial recognition module is used to statistically analyze the facial emotion data of all elderly people, analyzing the average frequency of facial emotions exhibited by the elderly in online learning scenarios. The module is built upon FaceReader software, which is designed based on Ekman's basic emotion theory and can analyze seven emotions with high accuracy. Chinese scholars have verified the system's ability to recognize facial expressions in Chinese people.

[0031] Meanwhile, the FaceReader software's training database contains facial expression data of the elderly and East Asians, which can effectively complete the emotion recognition task required for this study.

[0032] The emotion feature analysis module utilizes machine learning data mining methods to analyze the structural features of facial expressions in the elderly. The machine learning data mining method used is cluster analysis. Cluster analysis mathematically divides research samples into different categories, maximizing the similarity within each category and minimizing the similarity between different categories. It does not pre-determine the number or structure of clusters but objectively classifies them based on data characteristics, effectively avoiding human interference and ensuring the classification results more closely reflect the true distribution of the data. K-means clustering is a typical unsupervised clustering algorithm. Using relevant libraries in Python, cluster analysis can be performed on facial emotion data of the elderly. The calculation formula is shown below:

[0033]

[0034] Where SSE (Sum of the Squared Error) is the sum of the clustering errors of all samples, i represents the i-th cluster, p is the number of sample points, and m is the sum of the clustering errors of all samples. i SSE is the mean of the samples in family i. The lower the SSE value, the better the clustering effect. As the number of clusters K increases, the sample division becomes more refined, and the samples within each cluster become denser, thus reducing the SSE. However, when K is less than the actual number of clusters, increasing K will cause the SSE to decrease rapidly; when K is greater than the actual number of clusters, the rate of decrease in SSE caused by increasing K will slow down. Therefore, the inflection point of the slope of the SSE curve is the optimal K value.

[0035] Under the K-means clustering method, the SSE value of facial expression data changes with different K values.

[0036] like Figure 2 As shown, when K is greater than 5, the rate at which the SSE value decreases slows down as K increases, therefore 5 is the optimal number of clusters.

[0037] The Time-Based Emotional Mapping Module is used to draw smooth line graphs reflecting the epochal changes in emotions of older adults. It utilizes detailed facial expression recognition data exported from FaceReade software and employs Python and the Matplotlib library to draw smooth line graphs reflecting the epochal changes in emotions of older adults.

[0038] Meanwhile, the time-based emotion mapping module also includes a data preprocessing unit, which is used to normalize and remove noise from facial emotion data. The specific details are as follows:

[0039] Because the total duration of emotional recordings varies slightly for each elderly person, facial emotion data is normalized to allow for comparison of emotional changes among different learners at different time scales. To eliminate data noise, an exponentially weighted moving average is used to smooth the data, giving higher weight to recent data to better capture trends and thus improve interpretability.

[0040] The behavior dynamic transfer module is built on the interactive behavior analysis software GSEQ (General Sequential Querier) to perform lag sequence analysis on the facial expression behavior of the elderly, obtain a behavior transfer frequency table and a Z-score table, and derive the negative emotions of the elderly and their transfer patterns. This provides guidance for designing effective educational intervention measures and thus improves the emotional state of the elderly.

[0041] The intervention strategy module is used to determine the richness of emotional and behavioral shifts among different types of elderly people based on the Z-score, and automatically matches and pushes differentiated intervention content according to the richness of shifts. The intervention strategy module includes a preset unit and a matching unit;

[0042] The preset unit is used to define the richness classification of emotional behavior transfer based on the Z-score table.

[0043] For example, based on the Z-score scale, the emotional types of the elderly can be classified into three categories: anxious elderly, cheerful elderly, and serious elderly, and the corresponding intervention content can be entered.

[0044] The matching unit automatically identifies the emotional types of the elderly based on a preset classification of the richness of emotional behavior transfer, accurately matches the corresponding intervention content, and pushes differentiated intervention strategies through the system.

[0045] The aforementioned emotion intervention system for online learning for the elderly, based on sentiment analysis, accurately captures facial expression data of seniors in online learning scenarios through the collaborative work of a recording module and a facial recognition module. Using this facial data as a key information source, in-depth analysis reveals patterns in the emotional distribution of the elderly. These patterns not only provide a new perspective for understanding the emotional state of the elderly during the learning process but also help to deliver corresponding emotional intervention strategies, thereby more effectively improving the impact of emotional support.

[0046] Example 2

[0047] Thirty-two elderly participants were randomly recruited as participants in the experiment (8 men and 24 women) at a senior citizens' college in Beijing. The online learning platform chosen was the National Public Service Platform for Elderly Education, built by the National University for the Elderly. Participants selected the video course "How Much Do You Know About Health?", which consists of 10 lessons, each with a different theme.

[0048] For example, topics such as "Do I have any illnesses?" and "Who can die from a cold?" align with the learning preferences of older adults. Recording time was one hour to avoid fatigue. Adults were allowed to pause, jump around, and engage in other interactive activities during their online learning. Facial expressions were recorded via webcam. Researchers provided technical support throughout the process, ensuring a smooth experiment without any technical barriers affecting the participants' emotional well-being.

[0049] The facial videos of elderly people in the study were analyzed using a facial recognition module. Table 1 shows a summary of facial emotion data of some of the elderly people in the study.

[0050] Table 1 shows the duration of facial emotions observed in some study subjects.

[0051]

[0052] Statistical analysis was performed on the facial emotion data of all elderly people, and the results are shown in Table 2.

[0053] Table 2. Facial Emotion Statistics of All Study Subjects

[0054]

[0055] In online learning scenarios, the average frequency of facial emotions observed in older adults, from highest to lowest, was: neutral (61.6%), sadness (15.2%), anger (7.5%), joy (4.3%), disgust (3.0%), surprise (1.5%), and fear (0.2%). This indicates that neutral emotions were predominant among older adults. Furthermore, sadness occurred significantly more frequently than the other five emotions.

[0056] Clinical medical research shows that sad expressions are related to the degree of depression; depressed patients are significantly more likely to exhibit sad expressions than the general population, and this probability increases with the severity of depression. Therefore, the elderly have a higher risk of developing depression than the general population.

[0057] Further investigation was conducted into gender differences in the frequency of facial emotions among older adults. The results of the independent samples t-test are shown in Table 3.

[0058] Table 3. Mean frequency of facial expressions in different gender groups

[0059]

[0060] Women showed more neutral expressions (65.4% > 50.2%) and expressions of disgust (3.9% > 0.4%), while men showed more expressions of anger (16.50% > 4.56%), reaching a statistically significant level.

[0061] Based on the facial expression data of 32 elderly people, cluster analysis was used to analyze the structural features of facial expressions of the elderly.

[0062] Based on the above affective computing technology and clustering analysis algorithm, the emotional characteristics of the elderly in online learning scenarios can be clustered into five types. A radar chart was generated based on the average frequency of the six expressions (excluding neutral) for each group, yielding the following results:

[0063] The first group of elderly people mainly showed neutral emotions, while the average frequency of the other six types of expressions was less than 2%.

[0064] Especially in the other four groups;

[0065] While the overall level of sadness among the elderly was high, this group of elderly individuals exhibited a lower frequency of sad expressions (1.39%), indicating that they experienced less emotional fluctuation and maintained a relatively calm state of mind. This group can be termed the "calm type," and accounted for 21.88% of the elderly participants.

[0066] The second group of elderly participants showed a significantly higher frequency of anger (22.53%) compared to the other four groups, and also exhibited a higher frequency of sadness (21.37%). This can be termed the anxiety type, and accounted for 25% of the elderly participants.

[0067] The third group of elderly participants showed a significantly higher average frequency of disgust (14.3%) than the other four groups, accompanied by a higher frequency of sadness (20.88%). Expressions of pleasure (2.2%), surprise (0%), fear (0%), and anger (2.74%) were less frequent. This can be termed the "boredom" type, accounting for 15.63% of the elderly participants.

[0068] The fourth group of elderly people exhibited a higher proportion of sad expressions (17.04%), while the average frequency of the other five expression categories was less than 5%. Based on the scatter distribution of the clustering results, the main difference between this group and the annoyance type is that this group showed a lower frequency of disgust expressions. This group can be termed the serious type, accounting for 34.38% of the elderly participants, the highest percentage.

[0069] The fifth group of elderly individuals exhibited a higher frequency of cheerful expressions (21%), while expressions of surprise (7.2%) appeared significantly more frequently than in the other four groups. Although a certain percentage of sad expressions (14.9%) appeared, their frequency was lower compared to the other groups. Negative expressions such as fear (5.4%), disgust (0%), and anger (0%) appeared less frequently. This group showed a relatively higher proportion of positive expressions, termed the cheerful type, accounting for 3.13% of the participants, the lowest percentage.

[0070] The Time-Based Emotional Mapping Module is used to draw smooth line graphs that reflect the emotional epochal changes of the elderly, and to display the characteristics of emotional epochal changes in different types of elderly people.

[0071] In the smooth line graph of the emotional diachronic changes of calm-type elderly people, except for neutral emotions, the frequency of other emotions is low and basically unchanged; the curve is flat and the fluctuations are small.

[0072] In the smooth line graph of the emotional curve of anxious elderly people, sadness and anger occur with a significantly high frequency, and the frequency of these two emotions shows a slight upward trend over time; the curve is flat with little fluctuation.

[0073] In the smooth line graph of the emotional epoch change of the boredom type of elderly people, sadness and disgust appeared with significantly high frequency; the curve was not flat and fluctuated.

[0074] In the smooth line graph of the emotional diachrony of serious-type elderly people, sadness occurs significantly more frequently, while other types of emotions occur less frequently; the curve is flat and has little fluctuation.

[0075] In the smooth line graph of the emotional diachrony of happy older adults, multiple emotions coexist, but happy emotions are slightly higher than other emotions; the emotions change in a wave-like manner with large fluctuations.

[0076] As shown in Table 4, a table of coded letters and facial expressions is set up to facilitate subsequent research on negative emotions and their transfer patterns in the elderly, and to provide guidance for designing effective educational interventions.

[0077] Table 4. Correspondence between coded letters and facial expressions

[0078]

[0079] As shown in Tables 5 and 6, GSEQ analysis analyzed the frequency of emotion shifts and Z-scores in all elderly participants. "Initiating behavior" refers to the action that is initiated first, and "Later behavior" refers to the action that follows. For example, the intersection of the initiating behavior "N" and the delayed behavior "SU" table means that expression N was followed immediately by expression SU.

[0080] Table 5. Frequency of Facial Emotional Transfer in the Elderly

[0081]

[0082] As shown in Table 6, behavioral sequences with statistical significance were selected. If the adjusted residual Z-score was greater than 1.96, the behavioral transfer was considered significant. Overall, the Z-scores for H→SU, SU→H, SA→A, and A→SA were all greater than 1.96, indicating that the emotional transfer pattern in older adults was statistically significant for the four expression sequences: from pleasure to surprise, surprise to pleasure, sadness to anger, and anger to sadness. However, the transfer between other expression sequences was not significant. Furthermore, although the transfer frequency between neutral expressions and other expression types was extremely high, the Z-scores were all less than 1.96, indicating that these behavioral transfers were not significant, suggesting that neutral expressions are a common expression.

[0083] Table 6. Z-scores for facial emotion transfer sequences in the elderly.

[0084]

[0085] Further exploration of the emotional transfer patterns of different types of elderly people yielded significant transfer behaviors, as shown in Table 7.

[0086] Table 7. Emotional and behavioral transfer and Z-score scores of different categories of elderly people.

[0087]

[0088] The following can be derived from Table 7:

[0089] (1) Anxious older adults exhibit a wide range of emotional and behavioral shifts. There are significant bidirectional behavioral shifts between pleasure and surprise, sadness and anger, and sadness and fear. In addition, there are significant behavioral shifts from pleasure to aversion and from aversion to surprise.

[0090] (2) The only significant behavioral shift from sadness to happiness was observed in the happy type of older adults.

[0091] (3) In addition to the significant bidirectional behavioral shift between pleasure and surprise, serious older adults also exhibit a significant behavioral shift from sadness to fear.

[0092] It is evident that behavioral shifts between negative emotions occur in older adults, exhibiting a bidirectional cycle of shifts between sadness and anger, and between sadness and fear, particularly in those with anxiety patterns. This phenomenon is not observed in other age groups. The discovery of this behavioral shift pattern between negative emotions in anxious older adults is significant, indicating that their negative emotions can continuously shift between each other, leading to a persistent and difficult-to-interrupt cycle of negative emotions, resulting in accumulated negative feelings. Anxious older adults are more prone to depressive tendencies, thus requiring special attention and support for this group.

[0093] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. An emotion analysis-based emotional intervention system for online learning of the elderly, characterized by, The system comprises a recording module, a face recognition module, an emotional feature analysis module, a time emotional map drawing module, a behavior dynamic shift module, and an intervention strategy module. The recording module is used to automatically record the facial video of the elderly during online learning and obtain facial emotional data. The face recognition module is used to statistically analyze the facial emotional data of all the elderly and analyze the average frequency of facial emotions of the elderly in the online learning scenario. The emotional feature analysis module is used to analyze the structural features of the facial expressions of the elderly by using a machine learning data mining method. The time emotional map drawing module is used to draw a smooth broken line graph reflecting the duration change of the emotions of the elderly, and the time emotional map drawing module comprises a data preprocessing unit. The behavior dynamic shift module is used to perform lag sequence analysis on the facial expression behavior of the elderly, obtain a behavior shift frequency table and a Z-score score table, and derive the negative emotions of the elderly and their shift rules. The intervention strategy module is used to determine the richness of the emotional behavior shift of different types of elderly according to the Z-score score table, and automatically match and push differentiated intervention content according to the richness of the shift.

2. The emotion analysis based online learning emotional intervention system for the elderly according to claim 1, wherein, The recording duration of the recording module during online learning is 1 hour, and the online learning process allows interactive behaviors such as pausing, forward and backward skipping.

3. The emotion analysis based online learning emotional intervention system for the elderly as claimed in claim 1, wherein The face recognition module is constructed based on the FaceReader software, and the training database of the FaceReader software contains facial expression data of the elderly and East Asians.

4. The emotion analysis based online learning emotional intervention system for the elderly as claimed in claim 1, wherein, The machine learning data mining method is clustering analysis, which is used to calculate the sum of clustering errors of all samples of the facial emotional data of the elderly.

5. The emotion analysis based online learning emotional intervention system for the elderly as claimed in claim 1, wherein The smooth broken line graph is drawn by using the detailed expression recognition data derived from the FaceReader software, using Python language and Matplotlib library.

6. The emotion analysis based online learning emotional intervention system for the elderly as claimed in claim 1, wherein The data preprocessing unit is used to normalize and eliminate noise of the facial emotional data.

7. The emotion analysis based online learning emotional intervention system for the elderly as claimed in claim 1, wherein The intervention strategy module comprises a preset unit and a matching unit, the preset unit is used to define the emotional behavior shift richness classification based on the Z-score score table.

8. The emotion analysis based online learning emotional intervention system for the elderly as claimed in claim 7, wherein, The matching unit automatically identifies the emotional type of the elderly based on the preset emotional behavior shift richness classification, and accurately matches the corresponding intervention content.

Images