An interactive intelligent learning device for the elderly and a system thereof

Abstract

This invention relates to the field of smart elderly care education technology, specifically to an interactive smart learning device and system for the elderly. The device includes a teaching operation platform, an adapter interface, and a processing unit. The system integrates modules for data acquisition, processing, learning and intervention, memory recognition, and output. It generates a structured teaching environment and practical information through multi-source data acquisition, identifies learning abilities based on machine learning, breaks down operational sub-steps, and allocates teaching intensity. It dynamically adjusts teaching strategies by combining the basal forgetting cycle and personalized forgetting cycles, introduces relevant personnel to assist in optimizing intervention intensity, and outputs adaptive operation prompts, teaching tips, and learning time suggestions. This invention breaks through the limitations of traditional standardized teaching, achieving personalized and precise teaching and maintaining capabilities throughout the entire life cycle, improving the sustainability and practicality of digital skills acquisition for the elderly, and adapting to multiple teaching scenarios and age-appropriate needs.

Description

Technical Field

[0001] This invention relates to the field of smart elderly care education technology, specifically to an interactive smart learning device and system for the elderly. Background Technology

[0002] With the deepening of population aging and the popularization of digital technology, smart technology education for the elderly has become an important part of actively addressing population aging. Currently, smart hardware (such as window cleaning robots and robot vacuum cleaners), mobile applications (such as instant messaging, navigation, and shopping platforms), and anti-fraud knowledge constitute the three core modules of digital literacy education for the elderly.

[0003] Existing training programs for seniors in smart technologies primarily employ a standardized, centralized teaching model. Training institutions or community organizations develop unified teaching outlines and conduct face-to-face instruction according to fixed class hours and schedules, supplemented by standardized operation manuals for demonstration and practical training. Some programs utilize a video course self-study model, where seniors watch and learn as needed through pre-recorded instructional videos, and some platforms include simple post-course quizzes.

[0004] Existing technologies generally employ uniform teaching requirements, standardized teaching processes, and homogenized content delivery, failing to adequately consider the significant individual differences within the elderly population. This neglect of individual differences directly results in some elderly individuals being unable to effectively master core operational skills after instruction, leading to inconsistent knowledge retention rates. Furthermore, existing technologies only focus on immediate learning effectiveness assessments during the teaching process (such as immediate post-class practice and classroom interaction), constructing a short-term closed loop of teaching-learning-testing. However, this creates a serious blind spot in follow-up tracking. Due to the lack of continuous usage data tracking and forgetting warnings, the teaching content fails to effectively translate into practical life skills for the elderly in the long term.

[0005] To address the aforementioned shortcomings, there is an urgent need for a new teaching technology that can deeply integrate elderly individual characteristic recognition, adaptive adjustment of the learning process, dynamic modeling of forgetting curves, and long-term intervention mechanisms. This technology would enable a shift from standardized teaching to personalized and precise teaching, and from immediate effect assessment to lifelong capacity maintenance, thereby improving the long-term effectiveness and practical application rate of smart technology education for the elderly. Summary of the Invention

[0006] To address the aforementioned issues, this invention provides an interactive smart learning device and system for the elderly, which provides personalized and precise smart technology instruction for seniors and enhances the long-term effectiveness and practical application rate of smart technology education for seniors.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows: On the one hand, an interactive smart learning device for the elderly is provided, comprising: The teaching operation platform is used to provide a teaching platform and communicate with teaching equipment through an adapter interface to collect video and audio data of the operation process; the adapter interface is used to obtain the operation instructions and execution instructions of the teaching equipment in real time. The teaching operation platform integrates a processing unit, which is used to collect users' personal information and teaching content information; and generate teaching environment information and practical operation process information based on operation process video, voice data, and operation and execution instructions of teaching equipment; identify users' learning abilities based on users' personal information, teaching content information, teaching environment information, and practical operation process information; and break down operation steps according to the smallest operation unit based on teaching content information and output operation sub-steps. The system assigns instruction intensity to each sub-step based on the user's learning ability; calculates the base forgetting period for each sub-step and the user's personalized forgetting period for each sub-step; increases instruction intensity when personalized forgetting period ≤ base forgetting period and decreases instruction intensity when personalized forgetting period > base forgetting period; outputs operation prompts and instruction prompts for each sub-step based on the instruction intensity of each sub-step; and outputs a suggested time for the next instruction.

[0008] Furthermore, the processing unit is used to extract first teaching environment information association features and first practical operation information association features based on the operation process video; extract second teaching environment information association features and second practical operation information association features based on voice data; and extract third practical operation information association features based on the operation instructions and execution instructions of the teaching equipment. The information association features of the first teaching environment, the second teaching environment, the first practical process, the second practical process, and the third practical process are integrated and then transformed into structured teaching environment information and structured practical process information.

[0009] Furthermore, the information-related characteristics of the first teaching environment include light intensity, distractions, and ease of operation space; the information-related characteristics of the first practical operation process include the sequence of operation actions, operation waiting time, facial emotional state, and incorrect operation actions. The second set of teaching environment information features includes environmental noise decibels and types of interfering sounds; the second set of practical process information features includes user help requests, voice emotional state, and accompanying voice during operation. The third practical process information association features include the sequence of operation steps, details of operation parameters, types of operation errors, and operation completion status.

[0010] Furthermore, personal information includes: demographic characteristics, cognitive and ability characteristics, physical health characteristics, and learning preferences; Teaching content information includes: teaching content type, teaching content attributes, and teaching objectives; Information about the teaching environment includes: network status, venue conditions, and interference factors; The practical process information includes: operation data, status data, and completion status; The learning ability rating model, which is based on training, receives users' personal information, teaching content information, teaching environment information, and practical process information, and outputs a learning ability rating.

[0011] Furthermore, based on the trained teaching intensity allocation model, the model receives the attribute features of operation sub-steps, user learning ability features, and user historical practice features, and outputs the teaching intensity of each operation sub-step.

[0012] Furthermore, the basic forgetting period is calculated as follows:

[0013] in, The basal forgetting cycle; This is the operation complexity coefficient of the sub-step, used to measure the difficulty of the sub-step. This is the correlation coefficient between operation sub-steps, used to measure the correlation between operation sub-steps; The personalized forgetting cycle is calculated as follows:

[0014] in, For personalized forgetting cycles; The weighting is based on individual learning ability and is inversely proportional to the user's learning ability; The weighting is based on physiological condition and is directly proportional to the user's physiological limitations; The weighting is based on the degree of urgency, and is inversely proportional to the degree of urgency of the teaching content.

[0015] Furthermore, the processing unit is also used to define the user currently participating in learning as the first user, collect the information of the first user's associated personnel, including associated objects and associated attributes, the associated objects being the second users participating in the same teaching content, and the number of second users is at least one; predict the meeting cycle between the first user and the second user based on the associated attributes between the first user and the second user, and define it as the assistance cycle; update the teaching intensity based on the comparison between the assistance cycle and the basic forgetting cycle and the personalized forgetting cycle.

[0016] Furthermore, the processing unit is used to define the teaching intensity assigned to each operation sub-step as the first teaching intensity; When the personalized forgetting period is less than or equal to the basal forgetting period, the first teaching intensity is increased to the second teaching intensity; if the assistance period is less than or equal to the basal forgetting period, the second teaching intensity is decreased to the third teaching intensity, and the first teaching intensity is less than the third teaching intensity and the second teaching intensity is less than the second teaching intensity; if the assistance period is greater than the basal forgetting period, the second teaching intensity is maintained. When the personalized forgetting period is greater than the basal forgetting period, the first teaching intensity is reduced to the fourth teaching intensity; if the assisted period is less than or equal to the basal forgetting period, the fourth teaching intensity is reduced to the fifth teaching intensity, and the fifth teaching intensity is less than the fourth teaching intensity; if the assisted period is greater than the basal forgetting period, the fourth teaching intensity is maintained.

[0017] Furthermore, the reduction in the intensity of instruction from the second to the third level is directly proportional to the number of assisted cycles within the basal forgetting cycle; The reduction in the intensity of instruction from the fourth to the fifth level is directly proportional to the number of assisted cycles within the basal forgetting cycle.

[0018] On the other hand, an interactive smart learning system for the elderly is provided, which is integrated into a processing unit and includes a data acquisition module, a data processing module, a learning and intervention module, a memory recognition module, and an output module. The data acquisition module is used to collect users' personal information and teaching content information; and to generate teaching environment information and practical operation process information based on operation process video, voice data, and operation and execution instructions of teaching equipment. The data processing module is used to identify a user's learning ability based on the user's personal information, teaching content information, teaching environment information, and practical process information; it is also used to break down operation steps according to the smallest operation unit based on the teaching content information and output the operation sub-steps. The learning and intervention module is used to assign instructional intensity to each operational sub-step based on the user's learning ability; The memory recognition module is used to calculate the basic forgetting period for each operation sub-step and to calculate the user's personalized forgetting period for each operation sub-step. When the personalized forgetting period is less than or equal to the basic forgetting period, the teaching intensity is increased; when the personalized forgetting period is greater than the basic forgetting period, the teaching intensity is decreased. The output module is used to output operation prompts and teaching tips for each operation sub-step based on the teaching intensity of each operation sub-step; and to output a suggestion for the next teaching time.

[0019] The above approach has the following beneficial effects: Compared to existing technologies, the interactive smart learning device and system for the elderly proposed in this application collects personal information, teaching environment, and practical data from multiple dimensions. Combined with machine learning models, it accurately identifies learning abilities, breaks down steps into the smallest operational units, and dynamically allocates teaching intensity, addressing the problem of uneven skill mastery caused by neglecting individual differences among the elderly in existing technologies. It innovatively constructs a dual model of basal and personalized forgetting cycles, dynamically adjusting teaching strategies in conjunction with the assistance of relevant personnel, achieving precise early warning and intervention for forgetting. Compared to existing technologies that only focus on immediate effects, the retention rate of skills and knowledge is significantly improved. Precise teaching avoids repetition and redundancy, and the collaborative support mechanism with relevant personnel reduces waste of teaching resources and the burden on teaching staff. It supports deployment in multiple scenarios such as community, home, and institutional elderly care, eliminating the need for separate adaptations for different teaching content, significantly improving deployment flexibility and resource utilization compared to existing technologies. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the functional modules of an interactive smart learning device and system embodiment for the elderly according to the present invention. Figure 2 This is a schematic diagram of the system architecture of an interactive smart learning device for the elderly and its system embodiment of the present invention. The reference numerals in the accompanying drawings include: 100, interactive smart learning device for the elderly; 101, teaching operation platform; 102, adapter interface; 103, processing unit; 200. Interactive smart learning system for the elderly; 201. Data acquisition module; 202. Data processing module; 203. Learning and intervention module; 204. Memory recognition module; 205. Output module. Detailed Implementation

[0021] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The following detailed description illustrates the specific implementation method: Example 1: An interactive smart learning device for the elderly can be applied to community-based elderly education, home-based aging-friendly teaching, and smart elderly care service platforms, thereby promoting the improvement of digital literacy among the elderly and the construction of a lifelong learning system. (See also...) Figure 1The diagram shown is a functional block diagram of this embodiment of the interactive smart learning device for the elderly according to the present invention. In this embodiment, the interactive smart learning device 100 for the elderly consists of a teaching operation platform 101, an adapter interface 102, and a processing unit 103 integrated inside the teaching operation platform 101.

[0023] In this embodiment, the teaching operation platform 101 is used to provide a teaching platform and communicate with teaching equipment through the adapter interface 102 to collect video and audio data of the operation process.

[0024] Specifically, the teaching operation platform 101 is the core hardware carrier of this device, and its structure includes a main control unit, a display unit, a data acquisition unit, a communication unit, an audio unit, and a power supply unit. The main control unit uses an ARM Cortex-A72 quad-core processor, equipped with ≥4GB of RAM and ≥64GB of non-volatile memory, and is used to run the operating system, process data, and run teaching application software. The display unit is a ≥10-inch touch LCD screen that supports three levels of font magnification, making it easier for elderly people with declining vision to clearly identify interface elements. The data acquisition unit includes a 1080P high-definition camera and a dual microphone array, used to acquire video and audio data during the operation process, respectively. The camera is installed at the front of the platform, covering the user's operating area, and the microphone array has directional sound pickup and environmental noise suppression capabilities. The communication unit supports Wi-Fi. 6. The system supports Bluetooth 5.0 and Ethernet for low-latency data interaction with teaching equipment, cloud servers, and associated terminals. The audio unit includes a ≥2W stereo speaker and a voice speed adjustment module, dynamically adjusting output volume and playback speed according to the user's hearing. The power unit connects to a conventional power source; in some other embodiments, the power unit is equipped with a ≥5000mAh lithium battery and a Type-C fast charging interface, supporting continuous operation for ≥6 hours, meeting the needs of single-session lessons in community teaching points and home-based self-study. The physical form of the teaching operation platform 101 can be a single desktop platform, a portable tablet, or an embedded desktop terminal. Its dimensions, casing material, and human-computer interaction layout can be flexibly adjusted according to the actual deployment scenario; this application embodiment does not impose any special limitations on this.

[0025] The adapter interface 102 is used to obtain the operation instructions and execution instructions of the teaching equipment in real time.

[0026] Specifically, the adapter interface 102 is a standardized connection channel between the teaching operation platform 101 and the device to be taught, including a USB-A interface, a USB-C interface, an HDMI interface, a Bluetooth communication module, a Wi-Fi direct connection module, and a standardized API interface at the software level. The USB-A / C interface is used to connect smart hardware teaching devices (such as window cleaning robots and sweeping robots; window cleaning robots and sweeping robots can be placed directly on the surface of the teaching operation platform 101) to realize the issuance of operation commands and the feedback of execution status. When using a mobile terminal APP, the adapter interface 102 is used to mirror the mobile terminal APP operation interface and synchronously collect the screen operation flow and the user's finger touch trajectory. The Bluetooth and Wi-Fi direct connection modules are used to establish a low-power, high-reliability connection with the mobile terminal to obtain the operation commands (such as clicking the payment button and sliding to the order page) and execution commands (such as the payment success page jump completion) of the teaching device in real time. The standardized API interface defines a unified data format and calling protocol, is compatible with mainstream APP development frameworks, and ensures the integrity and timeliness of command collection. The physical interface type, communication protocol version, and driver of the adapter interface 102 can be dynamically loaded according to the target teaching equipment model, and the support range can also be expanded through firmware upgrades. This application embodiment does not impose any special limitations on this.

[0027] The processing unit 103 is used to collect the user's personal information and teaching content information; and generate teaching environment information and practical operation process information based on the operation process video, voice data, and operation instructions and execution instructions of the teaching equipment; identify the user's learning ability based on the user's personal information, teaching content information, teaching environment information, and practical operation process information; break down the operation steps according to the smallest operation unit based on the teaching content information and output the operation sub-steps; assign teaching intensity to each operation sub-step based on the user's learning ability; calculate the basic forgetting period of each operation sub-step, and calculate the user's personalized forgetting period for each operation sub-step. When the personalized forgetting period is less than or equal to the basic forgetting period, the teaching intensity is increased; when the personalized forgetting period is greater than the basic forgetting period, the teaching intensity is decreased. Based on the teaching intensity of each operation sub-step, the unit outputs operation prompts and teaching prompts for the operation sub-steps; and outputs a suggestion for the next teaching time.

[0028] Specifically, the processing unit 103 is integrated inside the teaching operation platform 101. It is a software and hardware co-operated module that runs all intelligent analysis logic. Its software part includes a data acquisition submodule, an information fusion submodule, a learning ability recognition submodule, an operation step decomposition submodule, a teaching intensity allocation submodule, a forgetting cycle calculation submodule, and a prompt output submodule. Its hardware part relies on the CPU and GPU resources of the main control unit and works with a dedicated AI acceleration unit (such as an NPU or DSP) to complete lightweight model inference.

[0029] The data acquisition submodule synchronously acquires video frame sequences, audio streams, operation instruction logs, and execution feedback logs of the teaching equipment by calling the driver program of the acquisition unit and the adapter interface 102. The information fusion submodule maps multi-source heterogeneous data to a unified semantic space. For example, it performs spatiotemporal alignment and feature splicing on the action temporal features extracted from the video, the emotion keywords identified in the speech, and the operation parameter details parsed from the device instructions to form a structured intermediate representation.

[0030] The learning ability recognition submodule uses a pre-trained CNN+LSTM hybrid neural network model (learning ability rating model). The input vector includes dimensions such as demographic features, cognitive test scores, initial practical error rate, and environmental interference intensity. It outputs a three-level learning ability rating of excellent / average / poor and a core weakness label (such as weak short-term memory and insufficient hand-eye coordination).

[0031] The operation steps are broken down into sub-modules based on the rules defined in the teaching content information for the smallest operation unit. For example, completing an order using a shopping app is broken down into seven sub-steps: opening the app → logging into the account → searching for products → adding to the cart → filling in the shipping address → selecting the payment method → ​​confirming payment. Each sub-step is independently executable, observable, and evaluable.

[0032] The teaching intensity allocation submodule sets an initial intensity coefficient based on the learning ability rating results. For example, the intensity coefficient is 1.0 for excellent level, 1.3 for intermediate level, and 1.8 for poor level. This coefficient affects the subsequent prompt density, number of practice sessions, and content presentation granularity.

[0033] The forgetting cycle calculation submodule executes the basic forgetting cycle respectively. With personalized forgetting cycles The operation, where This is the complexity coefficient of the operation sub-step (ranging from 0.5 to 3.0, determined by the interaction level involved in the step, the number of parameters, and the fault tolerance threshold). The correlation coefficient of the operation sub-step (range 0.3–1.0, calculated based on the dependency graph of the preceding and following steps). The weights for individual learning abilities are inversely proportional to the learning ability rating, with 0.6 / 0.8 / 1.2 corresponding to excellent / average / poor. The physiological condition is weighted (mapped to 0.8–1.5 based on the results of vision / hearing / hand dexterity tests). The weighting is based on the degree of necessity (0.7 for high-frequency essential content and 1.2 for low-frequency elective content).

[0034] The prompt output submodule generates differentiated prompt strategies based on the final teaching intensity value. For example, high intensity corresponds to superimposed graphic annotations, voice repetition, and operation countdown for each step, while low intensity corresponds to pop-up prompts only for key nodes. The algorithm parameters, model weights, and scheduling strategies of all submodules are stored in non-volatile storage media, supporting remote OTA updates and local grayscale verification.

[0035] User personal information includes demographic characteristics, cognitive and ability characteristics, physical health characteristics, and learning preferences. Demographic characteristics include age, gender, education level, and residential status, collected through a structured questionnaire on a touchscreen, supporting voice input verification. Cognitive and ability characteristics include attention test scores (using a digit span task), memory test scores (using a word recognition task), logical thinking test scores (using analogy reasoning questions), and basic digital literacy scores (based on self-assessment of prior app usage proficiency), all automatically scored using a built-in standardized scale. Physical health characteristics include vision level (obtained through screen interaction). The data collection includes: comparison adaptation test (inverse inference), hearing threshold (obtained through pure-tone audiometry), hand dexterity (derived from touch trajectory smoothness and click accuracy analysis), and declaration of underlying diseases (such as Parkinson's disease and arthritis). Objective indicators are automatically analyzed by the data collection unit, while subjective declarations are selected by the user. Learning preferences include teaching method preferences (video self-study / face-to-face instruction / group peer support), pace preferences (fast / standard / slow), content format preferences (primarily text / primarily voice / primarily video), and reminder method preferences (platform pop-ups / SMS / voice calls). These are collected and dynamically corrected through an initial guidance process. The collection order, presentation method, and verification logic of the above personal information fields can be optimized according to the operating habits of the elderly. For example, multiple-choice questions can be changed to a single-button progressive expansion. This application embodiment does not impose any special limitations on this.

[0036] The teaching content information includes the teaching content type, teaching content attributes, and teaching objectives. The teaching content type can be one of three categories: smart hardware operation, mobile app usage, or anti-fraud knowledge. These can be preset by the teaching administrator in the platform backend or actively selected by the user when starting the course. Teaching content attributes include the number of steps (integer, ≥1), logical correlation (0–1 continuous values, calculated based on data dependencies and state propagation relationships between steps), age-friendliness rating (1–5, output by the age-friendliness assessment model), and necessity level (high-frequency / medium-frequency / low-frequency, set based on community survey data). Teaching objectives are the core skills or knowledge points that the content must achieve, such as independently completing a WeChat video call and accurately identifying phishing SMS characteristics. These objectives are stored in natural language phrases to guide the breakdown of sub-steps and the generation of prompts. The metadata structure of the teaching content information uses JSON Schema, supports the import and version management of third-party teaching resource packages, and the meaning and value range of its fields are fixed during system initialization to avoid runtime ambiguity.

[0037] The teaching environment information includes network status, site conditions, and interference factors. Network status is detected in real-time by the communication unit, including network type (Wi-Fi / 4G / 5G / Ethernet), downlink speed (Mbps), packet loss rate (%), and connection stability (signal strength standard deviation over 10 consecutive seconds). Site conditions are acquired collaboratively by environmental sensors and video analysis, including light intensity (lux), ease of operation (calculated based on the percentage of the effective operating area within the camera's field of view), and background noise in decibels (dB). Interference factors include the number of external disturbances (counted by voice keywords) and ease of equipment placement (based on image recognition to determine whether the stand angle and height conform to ergonomics). This teaching environment information serves as a dynamic context for learning ability recognition and forgetting curve correction. For example, when network latency is >500ms, the video stream resolution is automatically reduced to ensure continuous instruction acquisition; when light intensity is <50lux, screen brightness is increased and a high-contrast UI mode is enabled.

[0038] The practical process information includes operation data, status data, and completion status. Operation data is jointly generated by the adapter interface 102 and the data acquisition unit, including the sequence of operation steps (a list of step IDs sorted by timestamp), operation time (the time difference between the start and end of each step), error types (accidental touch, missed steps, incorrect parameter input, timeout failure, etc.), number of errors, and assistance methods (voice call for help, platform button click, or call from associated personnel). Status data is derived from multimodal analysis, including facial emotional state (classified as focused / confused / anxious / relaxed using the ViT model), hesitation duration (the time interval from the start of a step to the first effective operation), and heart rate-related focus (an attention fluctuation curve derived from heart rate variability (HRV) analysis using a smart bracelet). Completion status is a Boolean field, indicating whether it was completed independently (without external assistance), whether it was completed completely (all sub-steps were executed), and completion quality (a weighted score based on operation path deviation and error rate, with a maximum score of 100). Practical process information is continuously written to the local database in the form of an event stream, supporting millisecond-level timestamp alignment and cross-modal backtracking analysis.

[0039] The generation of teaching environment information and practical process information based on operation process video, voice data, and operation and execution instructions of teaching equipment refers to: the processing unit 103 starting a multi-threaded data processing flow—the video stream passes through the YOLOv5s model to detect operation gestures and interface element state changes in real time, generating operation action time sequence labels; the voice stream passes through the Whisper-small model to transcribe text and extract emotional keywords and help intentions; the operation instruction log is parsed to obtain instruction type, parameter value, and execution status code; after the three are aligned according to a unified time axis (UTC microsecond level precision), they are input into a graph neural network (GNN) to perform cross-modal relationship modeling, and output structured teaching environment information (such as network status: Wi-Fi, packet loss rate 2.1%; interference factors: external interference once, occurring during step 3 execution) and structured practical process information (such as operation data: step 1 time 8.2s, step 2 error type is accidental touch, step 3 help method is voice call for help; status data: facial emotion during step 2 is confusion, hesitation duration 4.7s). The information generation process does not rely on manual annotation; it is all completed automatically by a pre-built model. The model parameters can be trained incrementally locally or optimized collaboratively with the cloud through a federated learning mechanism.

[0040] Identifying a user's learning ability based on their personal information, teaching content information, teaching environment information, and practical process information involves the processing unit 103 encoding these four types of information into feature vectors: the personal information vector has a dimension of 32, containing normalized values ​​and one-hot categorical encoding; the teaching content information vector has a dimension of 16, containing scalars such as the number of steps, relevance, and urgency; the teaching environment information vector has a dimension of 12, containing real-time measurements such as network latency, illumination, and noise; and the practical process information vector has a dimension of 24, containing statistical features such as error rate, hesitation duration, and emotional distribution entropy. These four vectors are concatenated and input into a CNN+LSTM hybrid neural network model. The CNN layer extracts local patterns (such as combinations of high error rates, prolonged hesitation, and confused expressions), the LSTM layer models temporal evolution (such as error types gradually shifting from accidental touches to missed steps reflecting fatigue accumulation), and finally, the fully connected layer outputs a probability distribution for a three-level learning ability rating. Before formal use, the model is trained on a dataset with a 7:2:1 ratio, achieving an accuracy ≥85%, and supports local small-sample fine-tuning (parameter updates are triggered each time ≥5 valid samples are added).

[0041] Based on the teaching content information, the operation steps are broken down into the smallest operation units. The output operation sub-steps refer to: Processing unit 103 calls the rule engine and semantic parser to jointly perform the breakdown—the rule engine performs rigid segmentation according to the preset definition of the smallest operation unit, such as ensuring that each sub-step contains only a single interactive action (click / swipe / input), can be independently verified (with clear success / failure feedback), and cannot be further decomposed (otherwise it loses its teaching significance); the semantic parser performs dependency parsing and action entity recognition on the teaching content description text (such as initiating a video call in WeChat), extracts the core verb (initiate), object (video call), and constraints (in WeChat), and then maps them to standard operation graph nodes; after the results of the two are merged, a list of operation sub-steps with serial numbers, names, preconditions, and post-feedback is generated, and teaching objectives and evaluation indicators are automatically associated. For newly imported teaching content, the system supports a manual review intervention mode. After the review is approved, it is stored in the knowledge graph for subsequent reuse.

[0042] Assigning teaching intensity to each sub-step based on the user's learning ability means that the processing unit 103 uses the learning ability rating as the baseline intensity and dynamically weights it according to the sub-step attributes—baseline intensity: excellent = 1.0, medium = 1.3, poor = 1.8; the sub-step attributes include the complexity coefficient C and the relevance coefficient R, and their product constitutes the basic difficulty value of the step; the final teaching intensity = baseline intensity × basic difficulty value × environmental correction factor (×1.2 when the network is poor, ×0.9 when the lighting is sufficient); the intensity value is mapped to a specific teaching strategy: when the intensity is ≥1.5, triple prompts are enabled (voice announcement before the step starts + floating text and image annotation during execution + voice repetition after completion); when the intensity is 1.0–1.4, double prompts are enabled (announcement before the start + repetition after completion); when the intensity is <1.0, only single prompts (repetition after completion) are enabled. The teaching intensity value is stored in floating-point form, supports continuous adjustment, and does not force discrete grading.

[0043] Calculate the base forgetting period for each operation sub-step, and calculate the user's personalized forgetting period for each operation sub-step. When the personalized forgetting period is less than or equal to the base forgetting period, increase the teaching intensity; when the personalized forgetting period is greater than the base forgetting period, decrease the teaching intensity. This means that processing unit 103 performs batch calculations—first, according to the formula... Calculate the base forgetting period for each sub-step, where The value is obtained by looking up the step attribute in a table (e.g., filling in the shipping address). Click the back button ), The calculation is based on the in-degree and out-degree of each step in the operation flowchart; secondly, the user's real-time physiological data is substituted (e.g., if the daily average heart rate variability (HRV) decreases by 15%). Increase by 0.1), and the day's practical performance (e.g., if the error rate increases by 20% compared to the average). (Upward adjustment of 0.05) and the essential nature of teaching content ,according to Determine personalized forgetting cycles; finally, compare... and :like If so, the current teaching intensity of this sub-step will be increased by 20% (maximum not exceeding 2.5); if If the teaching intensity is reduced by 15% (the lower limit is not lower than 0.5), the adjustment will only affect the prompting strategy in the next teaching cycle and will not change the structure of the generated operation sub-steps.

[0044] Based on the teaching intensity of each operation sub-step, the system outputs operation prompts and teaching prompts for each operation sub-step; and outputs a suggestion for the next teaching time. This means that, similar to the method of the learning ability rating model outputting the learning ability rating, the training-based teaching intensity allocation model receives the attribute features of the operation sub-step, the user's learning ability features, and the user's historical practice features, and outputs the teaching intensity of each operation sub-step. The processing unit 103 then calls the prompt generation engine and the time planning engine. Operation prompts focus on the action execution level. For example, a high-intensity sub-step prompt might display "Please press and hold the 'Pay' button for 3 seconds until a green checkmark appears on the screen," a medium-intensity prompt might display "Click the 'Pay' button," and a low-intensity prompt might simply announce "Payment successful" after the step is completed. Teaching prompts focus on the cognitive understanding level. For example, a high-intensity prompt might add "This is the last step to complete the order," and the system will automatically redirect to the payment success page. A medium-intensity prompt might add "The order is effective after payment confirmation," and a low-intensity prompt might not add any explanation. In addition, sub-step teaching prompts also include prompts for teaching administrators, such as "A certain step for this user needs detailed instruction to enhance understanding and memory." The next teaching time suggestion is generated by the time planning engine by comprehensively considering the personalized forgetting cycle T, the user's learning pattern (learning periods are derived through clustering of historical learning times), and the urgency of the teaching objectives (essential content is suggested to be moved to T / 2 time, and elective content is postponed to T×0.8 time). Finally, the prompts are output in the format of "[Date] [Time Period]" and can be manually adjusted by the user for confirmation.

[0045] The interactive smart learning device 100 for the elderly proposed in this embodiment collects video and audio data of the operation process through the teaching operation platform 101 and obtains operation instructions and execution instructions of the teaching equipment through the adapter interface 102. Then, the processing unit 103 generates teaching environment information and practical operation information based on the operation process video and audio data, and the operation instructions and execution instructions of the teaching equipment. The processing unit 103 also identifies the user's learning ability based on the user's personal information, teaching content information, teaching environment information, and practical operation information. Based on the teaching content information, the operation steps are broken down into the smallest operation units, and operation sub-steps are output. The teaching intensity is assigned to each operation sub-step based on the user's learning ability. The basic forgetting period of each operation sub-step is calculated, and the user's personalized forgetting period for each operation sub-step is calculated. When the personalized forgetting period is less than or equal to the basic forgetting period, the teaching intensity is increased. When the personalized forgetting period is greater than or equal to the basic forgetting period, the teaching intensity is decreased. Based on the teaching intensity of each operation sub-step, operation prompts and teaching prompts for each operation sub-step are output. The next teaching time suggestion is also output, comprehensively and proactively ensuring the long-term acquisition of digital skills and their application in life for the elderly.

[0046] Based on the first embodiment of the interactive smart learning device 100 for the elderly of the present invention described above, a second embodiment of the present invention is proposed (e.g. Figure 2 (As shown). In this embodiment, the processing unit 103 is further configured to extract first teaching environment information association features and first practical operation information association features based on the operation process video; extract second teaching environment information association features and second practical operation information association features based on voice data; extract third practical operation information association features based on the operation instructions and execution instructions of the teaching equipment; fuse the first teaching environment information association features, second teaching environment information association features, first practical operation information association features, second practical operation information association features, and third practical operation information association features, and then transform them into structured teaching environment information and structured practical operation information.

[0047] Specifically, extracting the first teaching environment information association features and the first practical operation information association features based on the operation process video refers to: the processing unit 103 calling a pre-trained lightweight video analysis model (e.g., a MobileNetV3+Temporal Shift Module combined architecture) to perform frame-by-frame decoding and spatiotemporal feature extraction on the 1080P operation process video stream collected by the teaching operation platform 101; among which, the first teaching environment information association features include light intensity, interference objects, and the convenience of the operation space, specifically implemented as follows: using the YOLOv5s model to detect the light source distribution area and the peak position of the brightness histogram in the video frame, calculating the global average illuminance value (unit: lux), and combining image contrast and shadow ratio to determine whether the lighting is sufficient; identifying non-teaching equipment moving objects (such as pedestrians, pets, and floating objects) in the video as interference objects through target detection, and counting their frequency and duration; using a monocular depth estimation network (such as MiDaS) v3.1) Reconstruct the 3D point cloud of the operating surface and calculate the overlap between the effective operating area and the user's hand range of motion to characterize the convenience of the operating space; all the above features are output in the form of numerical vectors, corresponding to the quantized values ​​of the three indicators respectively. This part of the features can select different precision model versions according to the actual deployment environment. For example, the INT8 quantization model is used in edge computing power-limited scenarios, and the FP16 precision enhancement module is enabled on high-configuration terminals in community centers. This application embodiment does not make any special limitations on this.

[0048] The first practical process information association features include the timing of operation actions, operation waiting time, facial emotional state, and erroneous operation actions. Specifically, the timing of operation actions is achieved by extracting the user's hand key point trajectory using OpenPose and combining it with the coordinate mapping of the teaching equipment's UI controls to generate an operation event sequence (format: ...). ,in For timestamps, (The ID of the manipulated control); the operation waiting time is defined as the cumulative time when the time interval between two adjacent valid operation events is greater than 2 seconds, which is obtained by event sequence difference statistics; facial emotion state is extracted by ArcFace feature encoder to extract facial ROI embedding vector, and the Softmax classifier after fine-tuning outputs four probability distributions: calm / confused / anxious / frustrated; erroneous operation actions are identified by comparing the standard operation sequence template with the actual event sequence, such as skipping steps, reversing order, and accidentally touching non-target controls, and the corresponding step number is marked. All the above features are sliced ​​and aggregated according to the granularity of teaching steps to form the corresponding features for each step. The dimensional behavioral feature vector; the threshold values ​​of each parameter (such as the waiting time determination threshold and the lower limit of confidence in emotion classification) can be calibrated according to the physiological response characteristics of the elderly population. For example, the waiting time determination threshold can be set to be adjustable within the range of 1.5–3.0 seconds. This application embodiment does not impose any special limitations on this.

[0049] Among them, the extraction of the second teaching environment information association features and the second practical process information association features based on speech data refers to: the speech data stream collected by the processing unit 103 through the dual microphone array, after noise reduction (using the RNNoise algorithm) and endpoint detection (based on energy and zero-crossing rate dual thresholds), is sent to the acoustic feature extraction module; the second teaching environment information association features include environmental noise decibels and interference sound types, specifically implemented by: calculating the audio frame-level RMS energy and converting it into the equivalent A-weighted sound pressure level (dB(A)), and taking the average value through a sliding window to obtain the environmental noise decibel value; the interference sound type is identified through a pre-trained Sound Event Detection model (CNN-BiLSTM architecture), outputting five probability distributions: human conversation / TV sound / kitchen noise / traffic noise / other; the second practical process information association features include user help-seeking voices, voice emotional state, and operation-accompanying voices, specifically implemented by: the help-seeking voice is jointly judged by keyword triggering and semantic intent recognition (such as detecting phrases like "how to do it" or "I don't know how to do the next step", and combining the BERT fine-tuning model to determine the help-seeking intent with a confidence level ≥0.85); the voice emotional state is determined using wav2vec. 2.0 Extracts speech embeddings, which are then processed by an LSTM classifier to output a four-class classification result consistent with the aforementioned facial emotions. Operation-accompanying speech refers to the speech content that occurs synchronously within 2 seconds before and after the operation. Its text transcription result is extracted using a speech-action temporal alignment module (calling the Whisper-tiny model) to aid in understanding the operation intent. All speech features are output as fixed-length vectors with dimensions of [missing information]. (Corresponding to noise decibels + probability of 4 types of interference sounds) and (Corresponding to the existence of requests for help, 3 types of emotions + accompanying voice text identifiers), the sampling rate, frame length, hop size and other hyperparameters of each model can be adjusted according to the hardware performance configuration, and this application embodiment does not make special limitations on this.

[0050] Among them, the extraction of the third practical process information association features based on the operation instructions and execution instructions of the teaching equipment refers to: the processing unit 103 listening to the event bus of the teaching equipment's operating system layer (AccessibilityService event stream for Android, AXObserver callback for iOS, and MQTT Topic subscription for smart hardware) through the adapter interface 102, and capturing in real time the operation instructions issued by the user (such as clicking the 'Confirm Order' button) and the execution instructions fed back by the device (such as order submission success or payment failure: insufficient balance); the third practical process information association features include the operation step sequence, operation parameter details, operation error type, and operation completion status. Specifically, the operation step sequence is generated by sorting the instruction event timestamps, and the format is as follows: The operation parameter details record the context parameters carried by each operation (such as input box text, selection item ID, and sliding distance); operation error types are identified by matching a preset exception pattern library, including illegal input, denied permissions, network timeout, and invisible controls; the operation completion status is determined by a combination of the status code in the execution instruction and semantic feedback, and outputs a three-level label of success / partial success / failure. This part of the feature is organized in the form of a structured JSON object, with each operation event corresponding to an object instance, and the fields include timestamp, step_id, params, error_type, and completion_status; the adapter interface 102 supports multiple physical channels such as USB / HDMI / Bluetooth / Wi-Fi and multiple protocol encapsulations such as HTTP / MQTT / RESTful. The specific communication method can be dynamically negotiated according to the model of the teaching device, and this application embodiment does not impose special limitations on this.

[0051] The process of fusing the first, second, first, second, and third practical process information association features, and then transforming them into structured teaching environment information and structured practical process information, refers to the following: Processing unit 103 adopts a feature-level early fusion strategy, firstly aligning the five types of feature vectors in time (using operation steps as the basic time unit, performing mean pooling or max pooling on each modality feature at the step granularity), and then concatenating them into a unified feature matrix. (in The total number of steps. The input is then fed into a multilayer perceptron (MLP, containing two hidden layers, each with 128 nodes, ReLU activation) for nonlinear mapping, outputting an intermediate representation. Finally, the structured teaching environment information is decoded through two parallel fully connected branches. (Fields are: network_stability, lighting_level, noise_level, interference_frequency) and structured practical process information. (Fields are: action_sequence_order, waiting_duration, error_frequency, help_request_count, emotion_score, completion_rate, parameter_consistency); all output fields are normalized to the [0,1] range and a confidence score is added (derived from the logits before the softmax layer of the last layer of the MLP). The information fusion module can be replaced with attention mechanism fusion (such as Cross-Modal Transformer) or decision-level post-fusion (weighted voting after independent modeling of each modality). The weight coefficients can be calibrated offline according to the modality reliability in different teaching scenarios. For example, the weight of the video modality is reduced in a strong light environment, and the weight of the device instruction modality is increased in a noisy environment. This application embodiment does not make any special limitations on this.

[0052] The interactive smart learning device 100 for the elderly proposed in this embodiment extracts first teaching environment information association features and first practical operation information association features based on operation process video by the processing unit 103; extracts second teaching environment information association features and second practical operation information association features based on voice data; extracts third practical operation information association features based on operation instructions and execution instructions of the teaching equipment; integrates the first teaching environment information association features, second teaching environment information association features, first practical operation information association features, second practical operation information association features, and third practical operation information association features, and then transforms them into structured teaching environment information and structured practical operation information, thus comprehensively and proactively ensuring the long-term acquisition of digital skills and the ability to apply them to daily life for the elderly.

[0053] Example 2: The core of this example is to optimize the allocation of teaching intensity by introducing the current learner's associated users (users participating in the same teaching content) and utilizing mutual resources, thereby reducing unnecessary teaching redundancy and ensuring effective intervention at key points of forgetting. Specifically, the processing unit 103 is also used to define the user currently participating in learning as the first user and collect the first user's associated user information, which includes associated objects and associated attributes.

[0054] Among them, the associated user is a second user (other user) who participates in the same teaching content as the first user (such as a learning group on the same community shopping app, or friends and family learning "smart lock operation" together), and there must be at least one such second user. The sources of the second user include: the first user actively adding the user through the touch screen / voice of the teaching operation platform (such as "adding family / classmates to learn together"); the learning group list imported in batches by the community teaching administrator (automatically matching users with the same teaching content); and other users who are already interconnected.

[0055] Among them, the correlation attribute is the core feature of the prediction assistance cycle, which can include the following quantifiable dimensions, all of which are automatically collected by the system or supplemented by users: Associated attribute types Specific features Data types Data collection method Basic characteristics of relations Relationship type (immediate family / collateral family / classmate / neighbor), residential distance (meters), whether living together (yes / no) Classification / Numerical User input + map location assistance (distance to residence is automatically calculated by the platform) Interactive historical features Number of meetings in the past 30 days, average meeting interval (days), number of times learning together, and number of successful mutual assistance (e.g., the number of times a second user assisted a first user in completing an operation). Numerical Teaching operation platform log statistics (collaborative learning records) + user-initiated marking (confirmed after meeting) Mutual assistance ability characteristics The second user's learning ability rating (excellent / average / poor), mastery rate of the current teaching content (%), and historical mutual assistance response speed (average response time, minutes). Classification / Numerical The system automatically synchronizes the second user's learning data (learning ability, mastery rate) + mutual assistance record statistics (response speed). The meeting cycle between the first user and the second user is predicted based on the association attributes between them, and is defined as the assistance cycle. Specifically, the processing unit 103 uses an LSTM time series prediction model to predict the future meeting cycle between the first user and the second user based on the association attributes. This implementation method is as follows: 1. Feature Processing Standardize the associated attribute features: one-hot encoding is used for relationship types (e.g., direct relatives are encoded as [1,0,0,0]), numerical features such as residential distance and meeting interval are normalized to the [0,1] interval, and the number of times of joint learning and the number of times of mutual assistance are processed by logarithmic transformation to handle the long-tail distribution.

[0056] 2. Model Training and Prediction Model architecture: A 3-layer LSTM network is used (input layer dimension = 8, hidden layer node count = 64, output layer dimension = 1), the activation function is ReLU, the optimizer is Adam, and the loss function is mean squared error (MSE). Training data: The training set is constructed using the "sliding window method" with a window size of 7 days and a ratio of training set, validation set, and test set of 7:2:1. Iterative training is stopped when the validation set loss does not decrease after 3 rounds, ensuring that the prediction error is ≤10%. Predicted output: Output the average meeting cycle (i.e., assistance cycle, denoted as T_{assist}) between the first and second users in the next 30 days. For example, a prediction result of "3 days" means that they meet once every 3 days on average.

[0057] The teaching intensity is updated based on a comparison of the assisted forgetting cycle, the basal forgetting cycle, and the personalized forgetting cycle. Specifically, the processing unit defines the teaching intensity assigned to each operational sub-step as the first teaching intensity and executes the following teaching intensity adjustment logic: When the personalized forgetting period is less than or equal to the basal forgetting period, the first teaching intensity is increased to the second teaching intensity; if the assistance period is less than or equal to the basal forgetting period, the second teaching intensity is decreased to the third teaching intensity, and the first teaching intensity is less than the third teaching intensity and the second teaching intensity is less than the second teaching intensity; if the assistance period is greater than the basal forgetting period, the second teaching intensity is maintained. When the personalized forgetting period is greater than the basal forgetting period, the first teaching intensity is reduced to the fourth teaching intensity; if the assisted period is less than or equal to the basal forgetting period, the fourth teaching intensity is reduced to the fifth teaching intensity, and the fifth teaching intensity is less than the fourth teaching intensity; if the assisted period is greater than the basal forgetting period, the fourth teaching intensity is maintained.

[0058] Among them, the reduction in the second to third teaching intensity is directly proportional to the number of assisted cycles within the basal forgetting cycle; the reduction in the fourth to fifth teaching intensity is directly proportional to the number of assisted cycles within the basal forgetting cycle.

[0059] The following method can be used as a reference: Define specific teaching strategies for each level of teaching intensity: Teaching Intensity Level Explanation duration Number of demonstrations Number of practice sessions Auxiliary tool configuration Error correction feedback intensity First Intensity of Instruction (Initial) 2-3 minutes 2 times 2 times Illustrated Handbook Only indicate correct / incorrect Second Intensity of Teaching (Enhanced) 3-5 minutes 4 times 4 times Images, text, voice prompts, and operation stickers Error type and correction steps Third level of teaching intensity (appropriately increased) 2.5-4 minutes 3 times 3 times Image and voice prompts Error type and correction steps Fourth, teaching intensity (reduced) 1-2 minutes 1 time 1 time No auxiliary tools Final results feedback only Fifth, teaching intensity (appropriately reduced). 0.5-1.5 minutes 1 time 1 time No auxiliary tools Final results feedback only Define the conversion relationship between teaching intensity levels: Second teaching intensity = First teaching intensity × 1.5 Third teaching intensity = Second teaching intensity × (1 - reduction amount) Fourth teaching intensity = First teaching intensity × 0.6 Fifth teaching intensity = Fourth teaching intensity × (1 - reduction amount) Adjust the process according to different scenarios Scenario 1: Personalized Forgetting Cycle (First-time users forget things quickly) First teaching intensity (e.g., 2-3 minutes of explanation) → upgraded to second teaching intensity (3-5 minutes of explanation, 4 demonstrations); like (Assistance period ≤ basal forgetting period, mutual assistance is possible): Third teaching intensity = Second teaching intensity × (1 - reduction amount); For example, if the reduction amount is 0.4, then the explanation time for the third intensity is 3-5 minutes × 0.6 = 1.8-3 minutes, rounded to 2-3 minutes, and the number of demonstrations is 4 × 0.6 = 2.4, rounded to 3 times. like (No effective mutual assistance): Maintain the second teaching intensity.

[0060] Scenario 2: Personalized Forgetting Cycles (First-time users forget things slowly) First teaching intensity → Reduced to fourth teaching intensity (1-2 minutes of explanation, 1 demonstration); like (Mutual assistance possible): Fifth teaching intensity = Fourth teaching intensity × (1 - reduction amount); For example, if the reduction is 0.3, then the explanation time for the fifth intensity level is 1-2 minutes × 0.7 = 0.7-1.4 minutes, rounded down to 0.5-1.5 minutes. like (No effective mutual assistance): Maintain the fourth teaching intensity.

[0061] Example 3: This example provides an interactive smart learning system for the elderly. This system is integrated into the processing unit and can realize the functions of the smart learning device described in the above examples; see reference... Figure 2 The diagram shown is a system architecture diagram of this embodiment of the interactive smart learning system for the elderly of the present invention; it includes a data acquisition module 201, a data processing module 202, a learning and intervention module 203, a memory recognition module 204, and an output module 205.

[0062] The data acquisition module 201 is used to collect users' personal information and teaching content information; and to generate teaching environment information and practical operation process information based on operation process video, voice data, operation instructions and execution instructions of teaching equipment; based on the multi-source information acquisition hardware of the interactive smart learning device for the elderly 100.

[0063] The data processing module 202 is used to identify the user's learning ability based on the user's personal information, teaching content information, teaching environment information, and practical process information; it is also used to break down the operation steps according to the smallest operation unit based on the teaching content information and output the operation sub-steps; the learning and intervention module 203 is used to assign teaching intensity to each operation sub-step based on the user's learning ability; the memory recognition module 204 is used to calculate the basic forgetting period of each operation sub-step and calculate the user's personalized forgetting period for each operation sub-step. When the personalized forgetting period is less than or equal to the basic forgetting period, the teaching intensity is increased; when the personalized forgetting period is greater than or equal to the basic forgetting period, the teaching intensity is decreased; the output module 205 is used to output operation prompts and teaching prompts for each operation sub-step based on the teaching intensity of each operation sub-step; and output a suggestion for the next teaching time; all are implemented based on the processing unit 103 of the interactive smart learning device for the elderly 100.

[0064] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. An interactive smart learning device for the elderly, characterized in that, include: The teaching operation platform is used to provide a teaching platform and communicate with teaching equipment through an adapter interface to collect video and audio data of the operation process. The adapter interface is used to obtain the operation and execution instructions of the teaching equipment in real time; The teaching operation platform integrates a processing unit, which is used to collect users' personal information and teaching content information; and to generate teaching environment information and practical operation process information based on operation process video, voice data, and operation and execution instructions of teaching equipment; and to identify users' learning abilities based on users' personal information, teaching content information, teaching environment information, and practical operation process information. Based on the teaching content information, the operation steps are broken down into the smallest operation units, and the operation sub-steps are output. The teaching intensity is assigned to each sub-step of the operation based on the user's learning ability; Calculate the base forgetting period for each operation sub-step, and calculate the user's personalized forgetting period for each operation sub-step. When the personalized forgetting period is less than or equal to the base forgetting period, increase the teaching intensity; when the personalized forgetting period is greater than the base forgetting period, decrease the teaching intensity. Based on the teaching intensity of each operation sub-step, output operation prompts and teaching prompts for each operation sub-step; and output a suggested time for the next teaching session.

2. The interactive smart learning device for the elderly according to claim 1, characterized in that, The processing unit is used to extract first teaching environment information association features and first practical operation information association features based on the operation process video; extract second teaching environment information association features and second practical operation information association features based on voice data; and extract third practical operation information association features based on the operation instructions and execution instructions of the teaching equipment. The information association features of the first teaching environment, the second teaching environment, the first practical process, the second practical process, and the third practical process are integrated and then transformed into structured teaching environment information and structured practical process information.

3. The interactive smart learning device for the elderly according to claim 2, characterized in that, The first teaching environment information association features include light intensity, interference objects, and ease of operation space; the first practical operation process information association features include the sequence of operation actions, operation waiting time, facial emotional state, and incorrect operation actions. The second teaching environment information association features include ambient noise decibels and types of interfering sounds; the second practical process information association features include user help requests, emotional state of the voice, and accompanying voice during operation. The third practical process information association features include the sequence of operation steps, details of operation parameters, types of operation errors, and operation completion status.

4. The interactive smart learning device for the elderly according to claim 1, characterized in that, The personal information includes: demographic characteristics, cognitive and ability characteristics, physical health characteristics, and learning preferences; The teaching content information includes: teaching content type, teaching content attributes, and teaching objectives; The teaching environment information includes: network status, venue conditions, and interference factors; The practical operation process information includes: operation data, status data, and completion status; The learning ability rating model, which is based on training, receives users' personal information, teaching content information, teaching environment information, and practical process information, and outputs a learning ability rating.

5. The interactive smart learning device for the elderly according to claim 1, characterized in that, The training-based teaching intensity allocation model receives the attribute features of operation sub-steps, user learning ability features, and user historical practice features, and outputs the teaching intensity of each operation sub-step.

6. The interactive smart learning device for the elderly according to claim 1, characterized in that, The basic forgetting period is calculated as follows: ； in, The basal forgetting cycle; This is the operation complexity coefficient of the sub-step, used to measure the difficulty of the sub-step. This is the correlation coefficient between operation sub-steps, used to measure the correlation between operation sub-steps; The personalized forgetting cycle is calculated as follows: ； in, For personalized forgetting cycles; The weighting is based on individual learning ability and is inversely proportional to the user's learning ability; The weighting is based on physiological condition and is directly proportional to the user's physiological limitations; The weighting is based on the degree of urgency, and is inversely proportional to the degree of urgency of the teaching content.

7. The interactive smart learning device for the elderly according to claim 1, characterized in that, The processing unit is also used to define the user participating in the learning as the first user, collect the associated personnel information of the first user, the associated personnel information includes associated objects and associated attributes, the associated objects are second users participating in the same teaching content, and the number of second users is at least one; predict the meeting cycle between the first user and the second user based on the associated attributes between the first user and the second user, and define it as the assistance cycle; update the teaching intensity based on the comparison between the assistance cycle and the basic forgetting cycle and the personalized forgetting cycle.

8. The interactive smart learning device for the elderly according to claim 7, characterized in that, The processing unit is used to define the teaching intensity assigned to each operation sub-step as the first teaching intensity; When the personalized forgetting period is less than or equal to the basal forgetting period, the first teaching intensity is increased to the second teaching intensity; if the assistance period is less than or equal to the basal forgetting period, the second teaching intensity is decreased to the third teaching intensity, and the first teaching intensity is less than the third teaching intensity and the second teaching intensity is less than the second teaching intensity; if the assistance period is greater than the basal forgetting period, the second teaching intensity is maintained. When the personalized forgetting period is greater than the basal forgetting period, the first teaching intensity is reduced to the fourth teaching intensity; if the assisted period is less than or equal to the basal forgetting period, the fourth teaching intensity is reduced to the fifth teaching intensity, and the fifth teaching intensity is less than the fourth teaching intensity; if the assisted period is greater than the basal forgetting period, the fourth teaching intensity is maintained.

9. The interactive smart learning device for the elderly according to claim 7, characterized in that, The reduction in the intensity of instruction from the second to the third level is directly proportional to the number of assisted cycles within the basal forgetting cycle. The reduction in the intensity of instruction from the fourth to the fifth level is directly proportional to the number of assisted cycles within the basal forgetting cycle.

10. An interactive smart learning system for the elderly, applicable to the interactive smart learning device for the elderly as described in any one of claims 1-9, characterized in that, The system is integrated into the processing unit, which includes a data acquisition module, a data processing module, a learning and intervention module, a memory recognition module, and an output module; The data acquisition module is used to collect users' personal information and teaching content information; and to generate teaching environment information and practical operation process information based on operation process video, voice data, and operation and execution instructions of teaching equipment. The data processing module is used to identify a user's learning ability based on the user's personal information, teaching content information, teaching environment information, and practical process information; it is also used to break down operation steps according to the smallest operation unit based on the teaching content information and output the operation sub-steps. The learning and intervention module is used to assign instructional intensity to each operational sub-step based on the user's learning ability; The memory recognition module is used to calculate the basic forgetting period for each operation sub-step and to calculate the user's personalized forgetting period for each operation sub-step. When the personalized forgetting period is less than or equal to the basic forgetting period, the teaching intensity is increased; when the personalized forgetting period is greater than the basic forgetting period, the teaching intensity is decreased. The output module is used to output operation prompts and teaching prompts for each operation sub-step based on the teaching intensity of each operation sub-step. It also provides a suggested time for the next teaching session.

Images