Intelligent evaluation system for elderly nursing education

Abstract

The invention belongs to the field of teaching evaluation, and provides an intelligent evaluation system for elderly nursing education, which comprises the steps of acquiring and storing historical video data in an elderly nursing process; performing image processing on the historical video, and extracting at least one frame of image containing significant expressions, postures or motion features of the old as a key frame image; presenting the key frame image to an evaluated person, and receiving judgment information of the evaluated person on the current state of the old person based on the key frame image and a corresponding nursing treatment scheme; after judgment information and a nursing treatment scheme input by an evaluated person are received, a complete video clip corresponding to the key frame image is played, the judgment information and the nursing treatment scheme of the evaluated person are compared and analyzed with a preset standard answer, and a scoring result is generated in combination with complete video content; and feeding back the scoring result to the evaluated person in a graphical or text form, and providing an improvement suggestion.

Description

Technical Field

[0001] The present invention belongs to the field of teaching evaluation, and more particularly relates to an intelligent evaluation system for elderly care education. Background Art

[0002] Elderly care is an important part of the modern social health service system. The elderly population has diverse and complex care needs in terms of physiology, psychology, cognition, etc. To meet the growing demand for high-quality care services and improve the professional capabilities of nursing staff, many higher medical colleges, vocational nursing institutions, and elderly care service units have made elderly care education the core teaching content, promoting the transformation from knowledge transfer to ability cultivation and the upgrade from skill training to situation response.

[0003] The existing elderly care education methods mainly rely on traditional teaching means such as classroom lectures, case analysis, skill training, and simulation exercises. Although some institutions have introduced clinical training humanoid models or scenario simulation systems, there are generally the following technical problems and deficiencies:

[0004] In current elderly care teaching, the evaluation of students' learning effects still mainly relies on subjective scoring by teachers, lacking a systematic, objective, and structured quantitative evaluation standard, and it is particularly difficult to accurately evaluate learners' observation, judgment, and handling abilities in emergency situations.

[0005] Most teaching platforms lack diverse cases based on real elderly care processes and are difficult to cover high-frequency emergencies such as falls, loss of consciousness, and pain attacks. The teaching content is disconnected from clinical practice.

[0006] The existing system has a multiple result-oriented evaluation mode, lacking the excavation of students' response processes, thinking paths, and logical chains, and it is difficult to provide targeted feedback suggestions and ability improvement paths.

[0007] Some teaching tools use static graphics or linear videos, which are difficult to simulate dynamic, complex, and real elderly care scenarios, resulting in low student participation and limited training effectiveness. Summary of the Invention

[0008] To solve the problems in the prior art, the present invention provides an intelligent evaluation system for elderly care education, including:

[0009] A video acquisition module for acquiring and storing historical video data during the elderly care process;

[0010] A key frame extraction module that performs image processing on the historical video and extracts at least one image containing significant facial expressions, postures, or action features of the elderly as a key frame image;

[0011] Image display and answering interaction module, which is used to present the key frame images to the person being evaluated and receive the judgment information of the current state of the elderly based on the key frame images by the person being evaluated and the corresponding nursing treatment plan;

[0012] Video playback module, which is used to play the complete video segment corresponding to the key frame images after receiving the judgment information and nursing treatment plan input by the person being evaluated;

[0013] Intelligent evaluation module, which is used to compare and analyze the judgment information and nursing treatment plan of the person being evaluated with the preset standard answers and generate a scoring result in combination with the complete video content;

[0014] Result display and feedback module, which is used to feedback the scoring result to the person being evaluated in a graphical or text form and provide improvement suggestions.

[0015] Furthermore, the video acquisition module includes a video input unit, a video parsing unit and a data storage unit. The video input unit is used to obtain video data from pension institutions, hospital nursing on-site teaching, clinical simulation drills, remote monitoring devices and historical nursing case databases.

[0016] Furthermore, the key frame extraction module includes a video frame decoding unit, a candidate frame detection unit, a posture mutation detection unit, a facial expression saliency detection unit, a key frame discrimination and screening unit and a key frame storage unit.

[0017] Furthermore, the facial expression saliency detection unit uses an expression recognition model to obtain the probability distribution of facial expressions and calculates the maximum expression confidence. If this confidence exceeds the preset threshold, it indicates that the frame image has significant expression features.

[0018] Furthermore, the image display and answering interaction module includes an image visualization unit, a answering prompt generation unit, a user interaction interface unit and an input parsing and data structuring unit.

[0019] Furthermore, the user interaction interface unit supports voice input and transcription functions and can provide standardized nursing term recommendations to assist users in quickly inputting nursing suggestions.

[0020] Furthermore, the video playback module includes a key frame index positioning unit, a video segment extraction unit, a playback control unit and a visualization annotation enhancement unit.

[0021] Furthermore, the visualization annotation enhancement unit is used to overlay skeleton trajectories, facial expression labels and subtitle prompts during playback to enhance the teaching evaluation effect.

[0022] Further, the intelligent evaluation module includes an answer preprocessing unit, a standard answer matching unit, a semantic understanding and similarity calculation unit, a video comparison and analysis unit, a scoring generation unit, and a feedback explanation generation unit.

[0023] Further, the result display and feedback module includes a result visualization unit, a text comment generation unit, a personalized improvement suggestion generation unit, a result storage and learning tracking unit, and a user interaction display unit.

[0024] An intelligent evaluation system for elderly care education provided by the present invention has the following beneficial effects:

[0025] By combining the display of key-frame images with the playback of the complete video, the present invention guides the person being evaluated to judge the current state of the elderly based on single-frame images and give a nursing treatment plan, and then conducts intelligent scoring by comparing with the complete video scenario, avoiding the drawback of relying too much on the subjective experience of teachers for scoring in previous nursing teaching, and realizing the objective and quantitative evaluation of judgment ability and decision-making rationality.

[0026] This system can comprehensively score the person being evaluated in multiple dimensions such as state recognition, logical reasoning, nursing treatment, and consistency with actual results, and provide a visual score structure and text comments through the result display and feedback module. At the same time, it supports the storage and tracking of answering records, and has good interpretability and traceability of the evaluation process.

[0027] The present invention uses historical video clips of elderly care as evaluation materials, combines image processing and behavior recognition technologies, extracts typical frame images from actual clinical scenarios, and constructs a highly simulated judgment training scenario, enabling learners to be in a real case background, which helps to improve their observation ability, judgment ability, and actual combat response ability, and enhance the immersive and scenario-based effects of teaching.

[0028] After the system completes the evaluation, it automatically generates personalized learning suggestions based on the scoring results, including error location, improvement strategies, and learning path recommendations, forming an intelligent teaching closed-loop of "judgment - feedback - improvement", which is beneficial to individualized teaching, hierarchical teaching, and the cultivation of autonomous learning ability. Description of the Drawings

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0030] Figure 1 It is the system diagram of the method of the present invention. Detailed Implementation Modes

[0031] Next, with reference to the accompanying drawings and specific implementation modes, a preferred description of the invention will be given.

[0032] This embodiment solves the above problems through the following steps:

[0033] In one embodiment, referring to Figure 1 , the present invention provides an intelligent evaluation system for elderly care education. The system collects image or video information during the elderly care process, combines multimodal analysis with an expert knowledge base, constructs a teaching evaluation model based on scene perception and semantic understanding, and can quantitatively score and provide personalized feedback on the coping abilities of nursing students in aspects such as identifying the behavioral characteristics of the elderly, judging nursing risks, and formulating nursing plans, so as to achieve the intelligent goal of "promoting learning through evaluation and promoting improvement through evaluation" in the elderly care teaching process.

[0034] Specifically, the system includes:

[0035] A video acquisition module for acquiring and storing historical video data during the elderly care process. It includes a video input unit, a video parsing unit, and a data storage unit, where:

[0036] The video input unit is used to obtain video data during the elderly care process from multiple data sources. The data sources include but are not limited to elderly care institutions, on-site teaching in hospital nursing, clinical simulation drills, remote monitoring devices, and historical nursing case databases, and support multiple input methods such as local file import, real-time camera capture, and cloud interface synchronization.

[0037] The video parsing unit performs format decoding, time axis calibration, and basic cleaning processing on the acquired video data, extracts image sequences, audio-video synchronization information, key event tags, etc. related to the elderly care behavior, and ensures the integrity and processability of the video content;

[0038] The data storage unit structurally stores the processed original video and its parsing information, classifies and manages it using timestamp indexing and multi-dimensional label management methods, including patient identity anonymization processing, event type classification (such as falls, turning over, feeding medicine, talking, etc.), and nursing scene classification (such as in-bed care, pre-meal care, toilet care, etc.), and supports on-demand call and version update, facilitating subsequent key frame extraction, training sample construction, and teaching scenario generation.

[0039] A key frame extraction module that performs image processing on the historical video and extracts at least one frame of an image containing significant facial expressions, postures, or action characteristics of the elderly as a key frame image.

[0040] This module includes a video frame decoding unit, a candidate frame detection unit, a posture mutation detection unit, a facial expression saliency detection unit, a key frame discrimination and screening unit, and a key frame storage unit.

[0041] The video frame decoding unit is used to decode the input nursing teaching video V into a frame sequence:

[0042] F = {f1, f2, …, f T}

[0043] Where:

[0044] f t represents the t-th frame image in the video;

[0045] T represents the total number of frames of this video segment;

[0046] The video frame rate is denoted as r (unit: frame / second);

[0047] To reduce the computational amount, the frame interval step size is set to Δt, and the extracted frame sequence is:

[0048] F′ = {f1, f 1+Δt , f 1+2Δt ,..., f 1+kΔt}

[0049] Where Preferably, take Δt = 5 (that is, extract one frame every 5 frames).

[0050] The candidate frame detection unit is used to estimate the pixel-level displacement between consecutive frames by a method and obtain the average motion energy E t of each frame, and is used to detect whether there is violent motion:

[0051]

[0052] Where:

[0053] I t (x, y) represents the grayscale or RGB value of the t-th frame image at the pixel position (x, y);

[0054] N is the total number of pixels in the image;

[0055] E t The larger it is, the more violent the action change between frames;

[0056] Set the motion threshold τ1. If E t > τ1, it is considered that this frame contains obvious actions.

[0057] The posture mutation detection unit is used to detect whether there is a significant change in the human posture.

[0058] Extract the coordinates of n key points of the human body using a pose recognition model (such as OpenPose or MediaPipe):

[0059]

[0060] Calculate the Euclidean distance transformation amount between adjacent frames:

[0061]

[0062] Where:

[0063] is the coordinate of the i-th key point in the t-th frame;

[0064] D[[ID=2X]] t represents the overall change degree of the pose structure;

[0065] Set the threshold τ2. If D t > τ2, it is considered that a significant pose transformation has occurred.

[0066] The facial expression saliency detection unit is used to perform face detection and expression classification on each frame of image. Use the EmoNet model to obtain the probability distribution of facial expressions:

[0067]

[0068] Where:

[0069] m is the preset number of emotion types (such as pain, surprise, anger, smile, indifference, etc.); represents the probability that the t-th frame belongs to the j-th type of expression;

[0070] Define the maximum expression confidence as:

[0071] C t = max(P t )

[0072] Set the confidence threshold τ3. If C t > τ3, it indicates that the expression features of this frame are significant.

[0073] The key frame discrimination and screening unit synthesizes the three features and constructs a key frame discrimination function:

[0074] Where:

[0075] is an indicator function, which takes the value of 1 when the condition is met, otherwise 0;

[0076] If K(f t ) ≥ 2, it is considered that the t-th frame meets the key frame requirements and is extracted.

[0077] The key frame storage unit is used for structurally annotating the screened key frame images, including: key frame numbers and timestamps;

[0078] Expression categories and confidence levels;

[0079] Posture mutation types (such as falling, sideward movement, turning over);

[0080] Scene categories (such as in bed, using the toilet, eating);

[0081] They are stored in the key frame image database for subsequent calls by the image display and interactive evaluation module.

[0082] The image display and answer interaction module is used to present the key frame images to the person being evaluated and receive the judgment information of the person being evaluated on the current state of the elderly based on the key frame images and the corresponding nursing treatment plans.

[0083] This module includes four parts: an image visualization unit, a question answering prompt generation unit, a user interaction interface unit, and an input parsing and data structuring unit.

[0084] The image visualization unit performs standardization processing on the images output by the key frame extraction module, including size adjustment, border annotation (such as head, hand, face areas), and label overlay (such as "pain expression");

[0085] Furthermore, it supports multi - image combined display (such as providing front - back comparison of upper and lower frames) or image magnification for viewing;

[0086] Furthermore, it supports displaying the basic information of the elderly (such as gender, age, past medical history) below the image;

[0087] Furthermore, it supports providing an image loading progress prompt and a task status bar.

[0088] The question answering prompt generation unit automatically generates observation task instructions and question answering guidance information for the key frame image;

[0089] Exemplarily, the prompt format can include:

[0090] "Please observe the facial expressions, body postures, and behavior states of the elderly in the image";

[0091] "Please judge the possible health risks the elderly may face currently and put forward specific nursing suggestions."

[0092] Furthermore, the task type can be configured (open - ended question, multiple - choice, operation sorting, etc.).

[0093] The user interaction interface unit provides a human - machine interaction input interface, supporting the following input forms:

[0094] Status description input box: The user fills in the judgment of the current status of the elderly (such as "confused consciousness", "suspected pre - fall state", "psychological anxiety");

[0095] Disposal plan input box: Fill in nursing response suggestions (such as "immediately support", "verbal comfort", "check pulse", "call doctor");

[0096] Furthermore, it may include a voice input and transcription module: The user answers questions through voice, and the system automatically recognizes and converts them into text;

[0097] Furthermore, it may include standardized recommended vocabulary prompts: For the user to quickly select common nursing terms to improve accuracy.

[0098] Input parsing and data structuring unit

[0099] Parse the user input content into a structured JSON format:

[0100] Exemplarily:

[0101]

[0102] Furthermore, a natural language processing model can be called to perform semantic verification, denoising processing, and keyword extraction on the user input content;

[0103] Attach a unique identifier to the input content for corresponding storage with subsequent scoring results.

[0104] A complete example:

[0105] Key - frame image: An elderly woman is sitting sideways on the edge of the bed, with a painful facial expression and her hand covering her chest

[0106] System prompt:

[0107] "Please judge the current status of the elderly based on the image and give a disposal plan."

[0108] User input:

[0109] Observation and judgment: Suspected angina attack

[0110] Nursing advice: Immediately assist her to lie flat, monitor breathing and pulse, and notify medical staff.

[0111] The system then calls the original video playback module to play the complete segment of this event and enters the intelligent scoring process.

[0112] Video playback module, which is used to play the complete video segment corresponding to the key frame image after receiving the judgment information and the nursing treatment plan input by the person being evaluated. Automatically playing the complete video segment of the nursing process corresponding to this key frame image can help the person being evaluated compare and analyze the differences between their judgment and the actual situation, and at the same time provide a reference basis for the intelligent scoring module.

[0113] This module includes a key frame index positioning unit, a video segment extraction unit, a playback control unit, and a visualization annotation enhancement unit.

[0114] The key frame index positioning unit retrieves the time stamp T corresponding to this frame in the original video according to the key frame number f of the currently evaluated image t , in the original video; t ;

[0115] Define the playback window length ΔT, and determine the start time T start = T t -ΔT1, and the end time T end = T t +ΔT2;

[0116] Among them, ΔT1 and ΔT2 are the left and right time extension windows (for example: 5 seconds before to 10 seconds after);

[0117] Furthermore, it also supports intercepting the segments of the required intervals from the original video database as the playback content as needed.

[0118] The video segment extraction unit calls the video decoder to extract the video frames within the specified time interval [T start , T end ;

[0119] If the video frame rate is r frames per second, then the extracted frame set is:

[0120]

[0121] The playback control unit provides a video playback control interface for the person being evaluated, supporting:

[0122] Play, pause, and speed change;

[0123] Drag the time axis;

[0124] Quickly locate to the key frame position;

[0125] Furthermore, the following information can be displayed during the playback:

[0126] The current playback frame number and time stamp;

[0127] Key frame marking prompt (such as highlighting with a red dot on the time axis);

[0128] Nursing situation labels (such as "elderly person falls", "asks for help when sitting up", etc.).

[0129] When the visual annotation enhancement unit plays a video clip, it superimposes auxiliary information to enhance the teaching and evaluation effects, including:

[0130] Select the elderly area by drawing a box;

[0131] Visualize the real-time pose trajectory (skeleton lines);

[0132] Facial expression labels and probability distribution heat maps (such as "pain: 87%");

[0133] Add subtitle prompts or background explanations.

[0134] Intelligent evaluation module, which is used to compare and analyze the judgment information of the person to be evaluated with the nursing treatment plan and the preset standard answer, and generate a scoring result in combination with the complete video content.

[0135] This module includes an answer preprocessing unit, a standard answer matching unit, a semantic understanding and similarity calculation unit, a video comparison and analysis unit, a scoring generation unit and a feedback explanation generation unit.

[0136] The answer preprocessing unit formats the judgment text and treatment plan input by the person to be evaluated; it includes word segmentation, part-of-speech tagging, syntactic parsing, synonym normalization, negation word judgment, etc.; natural language processing tools (such as SpaCy, BERT-Tokenizer, etc.) can be used to generate vectorized representations.

[0137] Standard answer matching unit

[0138] The system pre-establishes a "standard answer library" covering different situations, and each standard answer includes:

[0139] Exemplarily:

[0140]

[0141] According to the current key frame number and situation type, filter out the set of reference answers with the highest matching degree.

[0142] Semantic understanding and similarity calculation unit

[0143] Embed the user's answer and the standard answer into the semantic space respectively, and use pre-trained models such as BERT, SBERT, RoBERTa to obtain sentence vectors:

[0144]

[0145] Among them:

[0146] U t is the user input statement;

[0147] S t is the standard answer statement;

[0148] is the corresponding sentence vector representation.

[0149] Calculate the cosine similarity score:

[0150]

[0151] The score range is [0, 1], and the closer to 1 indicates a higher semantic consistency.

[0152] The video comparison and analysis unit performs behavior recognition on the actual events that occur in the complete video (such as whether the elderly person falls, groans painfully, or calls for help); uses action recognition models (such as I3D, SlowFast, TSM) to judge the event type; combines the video results with the user's prediction to perform result deviation analysis and quantify the prediction accuracy.

[0153] The scoring generation unit is used to generate the total score.

[0154] The total score is:

[0155] S total = α1S status + α2S logic + α3S plan + α4S actual

[0156] Where:

[0157] S status represents the accuracy score for the elderly person's state recognition;

[0158] S logic represents the logical score for the judgment process;

[0159] S plan represents the scientific and feasible score for the nursing treatment suggestions;

[0160] S actual represents the consistency score with the results of the complete video events;

[0161] α i is the weighting coefficient, satisfying ∑α i = 1 (such as 0.3, 0.2, 0.3, 0.2).

[0162] Description of each score:

[0163] The state recognition score S status :

[0164] Determine whether the physical abnormalities of the elderly (such as pain, imbalance) are recognized;

[0165] Those with a semantic matching degree ≥ threshold (such as 0.7) with the standard answer get full marks.

[0166] Logical score S logic :

[0167] Determine whether there are errors or disorder in the nursing process;

[0168] Use a rule engine to check the "observation → judgment → execution" process.

[0169] Disposal plan score S plan :

[0170] Determine whether it complies with the nursing norms and whether there are safety risks;

[0171] Deduct points for risky operations such as "directly picking up the elderly".

[0172] Result deviation score S actual :

[0173] Degree of matching with the actual situation of the video;

[0174] If the user predicts "fall" and it actually occurs in the video, full marks will be obtained.

[0175] The feedback explanation generation unit automatically generates a structured evaluation feedback according to the scores of each item:

[0176] Result summary (such as: "You successfully recognized that the elderly may have angina pectoris, and the disposal suggestions are relatively comprehensive, but the on-site environmental safety factors were not considered.");

[0177] Gap analysis chart (such as a radar chart comparing with the standard answer);

[0178] Optimization suggestions (such as: "Next time, you can first check the consciousness state to avoid direct moving.");

[0179] Evaluation level (such as: excellent, qualified, need improvement);

[0180] The result display and feedback module is used to feedback the scoring results to the evaluated person in a graphical or text form and provide improvement suggestions.

[0181] Specifically, this module includes a result visualization unit, a text comment generation unit, a personalized improvement suggestion generation unit, a result storage and learning tracking unit, and a user interaction display unit.

[0182] The result visualization unit displays the dimensional indicators of the scoring results in a graphical manner. Common graphs include:

[0183] Radar chart: showing the sub - scores of "status judgment, logic, handling rationality, result deviation";

[0184] Bar chart / Comparative bar chart: comparing the scores of the evaluated person with the class average and the standard answer score;

[0185] Heat map: marking the key scoring words / misjudged words in text answers;

[0186] Trend curve chart: showing the learning and growth trend of multiple answers.

[0187] Text comment generation unit

[0188] Automatically generate text comments using a rule engine based on template filling + conditional combination:

[0189] Exemplarily, the output is:

[0190] "You accurately identified the abnormal behavior of the elderly, with clear judgment logic and relatively comprehensive handling suggestions. However, the expression is slightly vague, and the specific steps of vital sign monitoring are not clearly defined. It is recommended to strengthen the standardization of the emergency response process in future operations."

[0191] Or call an NLP fine - tuning model (such as T5 / BART) to generate a natural - language summary of the scoring dimensions; the text content includes: overall evaluation, sub - item analysis, key error prompts, behavior suggestions, etc.

[0192] The personalized improvement suggestion generation unit generates specific operation suggestions based on error labels and low - scoring dimensions, and the output format includes: action improvement suggestions (such as "The consciousness state should be evaluated first before moving");

[0193] Communication suggestions (such as "You can try to establish communication using a soft tone");

[0194] Expression suggestions (such as "There is a lack of clear terms in the answer, it is recommended to strengthen language precision"); Recommended practice content (such as "It is recommended to review the content of Chapter 3 on emergency response").

[0195] If a knowledge graph is accessed, the task recommendation engine can be called to dynamically push corresponding learning resource links.

[0196] The result storage and learning tracking unit writes the current scoring result, input content, and system feedback into the database uniformly, with the following structure:

[0197]

[0198] The user interaction display unit provides a graphic - text mixed feedback interface, and the module includes:

[0199] Scoring Scorecard (Total Score, Star Rating);

[0200] Graphical Display Panel;

[0201] Error Review Area (Displays Input Text + Error Point Markings);

[0202] Suggestions and Practice Area (Jumps to Practice Module);

[0203] Historical Task Browser (Jumps to Previous Answer Record).

[0204] Exemplarily, the output

[0205] Total Score: 88 points (Behavior Recognition of the Elderly 26 / 30, Logic 19 / 20, Disposal Suggestions 23 / 30, Video Consistency 20 / 20)

[0206] System Comment:

[0207] You have shown strong observation skills in recognizing the behaviors of the elderly and can make relatively reasonable emergency judgments. However, the disposal suggestions are slightly conservative, lacking clear descriptions of subsequent monitoring measures.

[0208] Suggestions for Improvement:

[0209] Key links such as "blood pressure / pulse monitoring" were not mentioned in the answer. It is recommended to strengthen the study of cases of acute chest pain;

[0210] It is recommended to review Section 3 of "Emergency Disposal of Elderly Sudden Diseases".

[0211] For the part of the module structure not specifically defined in the present invention, the content recorded in the prior art shall prevail. The prior art mentioned in the foregoing background art section and the specific embodiment section of the present invention can be used as a part of the present invention to understand the meaning of some technical features or parameters.

Claims

1. An intelligent evaluation system for elderly care education, characterized in that, The system includes: A video acquisition module, which is used to acquire and store historical video data during the elderly care process; A key frame extraction module, which performs image processing on the historical video and extracts at least one frame of image containing significant expression, posture or action features of the elderly as a key frame image; An image display and answer interaction module, which is used to present the key frame image to the person being evaluated and receive the judgment information of the person being evaluated on the current state of the elderly based on the key frame image and the corresponding nursing treatment plan; A video playback module, which is used to play the complete video segment corresponding to the key frame image after receiving the judgment information and nursing treatment plan input by the person being evaluated; An intelligent evaluation module, which is used to compare and analyze the judgment information and nursing treatment plan of the person being evaluated with the preset standard answers, and generate a scoring result in combination with the complete video content; A result display and feedback module, which is used to feedback the scoring result to the person being evaluated in a graphical or text form and provide improvement suggestions.

2. The intelligent evaluation system for elderly care education according to claim 1, wherein The video acquisition module includes a video input unit, a video parsing unit and a data storage unit, and the video input unit is used to acquire video data from nursing homes, hospital nursing on-site teaching, clinical simulation drills, remote monitoring devices and historical nursing case databases.

3. The intelligent evaluation system for elderly care education according to claim 1, wherein The key frame extraction module includes a video frame decoding unit, a candidate frame detection unit, a posture mutation detection unit, a facial expression saliency detection unit, a key frame discrimination and screening unit and a key frame storage unit.

4. The intelligent evaluation system for elderly care education according to claim 3, wherein The facial expression saliency detection unit uses an expression recognition model to obtain the probability distribution of facial expressions and calculates the maximum expression confidence. If the confidence exceeds the preset threshold, it indicates that the frame image has significant expression features.

5. The intelligent evaluation system for elderly care education according to claim 1, characterized in that The image display and answer interaction module includes an image visualization unit, an answer prompt generation unit, a user interaction interface unit and an input parsing and data structuring unit.

6. The intelligent evaluation system for elderly care education according to claim 5, wherein The user interaction interface unit supports voice input and transcription functions, and can provide standardized nursing term recommendations to assist users in quickly inputting nursing suggestions.

7. The intelligent evaluation system for elderly care education according to claim 1, characterized in that, 8. The intelligent evaluation system for elderly care education according to claim 7, characterized in that The video playback module includes a key frame index positioning unit, a video segment extraction unit, a playback control unit and a visualization annotation enhancement unit.

9. The intelligent evaluation system for elderly care education according to claim 1, wherein The visualization annotation enhancement unit is used to superimpose skeleton trajectories, facial expression labels and subtitle prompts during playback to enhance the teaching evaluation effect.

10. The intelligent evaluation system for elderly care education according to claim 9, wherein The intelligent evaluation module includes an answer preprocessing unit, a standard answer matching unit, a semantic understanding and similarity calculation unit, a video comparison and analysis unit, a scoring generation unit and a feedback explanation generation unit. The result display and feedback module includes a result visualization unit, a text comment generation unit, a personalized improvement suggestion generation unit, a result storage and learning tracking unit, and a user interaction display unit.

Images