Methods, apparatuses, devices, and media for rapid frailty screening and assessment in the elderly

Abstract

The application discloses a kind of old frailty rapid screening and evaluation method, device, equipment and medium, comprising: generating virtual digital doctor image, and control the virtual digital doctor is through voice and / or video guide user participates in frailty screening;Through voice acquisition module, the voice answer made by user to screening question is collected, and user voice data is obtained;The user voice data is identified by voice recognition module, and voice content is converted into text content;Wherein, the voice recognition module is combined with the language model that is fine-tuned by frailty knowledge;Based on the text content, in combination with the preset frailty screening scale and / or multidimensional prediction model.Main technical scheme and effect: the present application is automatically completed questionnaire inquiry, guide and result feedback by digital human virtual doctor, replaces the manual operation mode that must be participated by professional medical staff throughout.

Description

Technical Field

[0001] This invention relates to the field of health screening for the elderly, and more particularly to a method, apparatus, equipment, and medium for rapid screening and assessment of frailty in the elderly. Background Technology

[0002] With the accelerating aging of the global population, frailty, as an important geriatric syndrome, is increasingly becoming a focus of public health and clinical medicine. Frailty refers to a non-specific state in older adults caused by declining physiological reserves and multi-system functional decline, leading to reduced stress resistance and significantly increasing the risk of falls, disability, hospitalization, and death. Therefore, early, rapid, and accurate screening and assessment of frailty in the elderly population is crucial for implementing targeted interventions, delaying the progression of disability, improving quality of life, and optimizing the allocation of medical resources.

[0003] Currently, frailty screening for the elderly mainly relies on manual methods, where professional doctors directly ask screening questions to the elderly through face-to-face interviews or electronic questionnaires, and manually calculate scores based on their answers to complete the assessment. The core drawback of this existing technology is that its screening process is highly dependent on the full participation of professional doctors and manual operation, resulting in high labor costs and low screening efficiency, making it difficult to meet the needs of large-scale and rapid application in scenarios such as communities and elderly care institutions. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to solve the technical problems in the prior art and provide a method, device, equipment and medium for rapid screening and assessment of frailty in the elderly.

[0005] This specification relates to one or more embodiments of a rapid screening and assessment device for frailty in the elderly, an electronic device, a computer-readable storage medium, and a computer program product, in order to address the technical deficiencies existing in the prior art.

[0006] Technical solution:

[0007] Firstly, this application proposes a rapid screening and assessment method for frailty in the elderly, including:

[0008] Generate a virtual digital doctor avatar and control the virtual doctor to guide the user to participate in frailty screening through voice and / or video;

[0009] The voice acquisition module collects the user's voice responses to screening questions to obtain user voice data;

[0010] The user's voice data is recognized by a speech recognition module, and the voice content is converted into text content; wherein, the speech recognition module incorporates a language model that has been fine-tuned with weakening knowledge.

[0011] Based on the text content, combined with a preset frailty screening scale and / or a multi-dimensional prediction model, the user's frailty index is calculated.

[0012] The calculated decay index and / or the corresponding assessment results are fed back to the user through the digital human interaction system, and the relevant data is stored in the storage unit.

[0013] Preferably, generating a virtual digital doctor avatar and controlling the virtual doctor to guide the user in participating in frailty screening via voice and / or video includes:

[0014] Using deep learning generative models, highly realistic digital human appearance, expressions, and lip-syncing animations are generated based on real-person video data.

[0015] Using neural speech synthesis technology, speech with emotional and prosodic variations is generated based on text;

[0016] The generated speech is synchronized in real time with the digital human's lip-sync animation, facial expressions, and body movements using a synchronization controller.

[0017] Preferably, the method involves acquiring the user's voice responses to screening questions via a voice acquisition module to obtain user voice data. The method also includes a step of performing real-time noise reduction processing on the user voice data using a noise reduction model based on a deep neural network after acquiring the user's voice.

[0018] 4. The method according to claim 1, wherein the speech recognition module incorporates a language model fine-tuned with weakened knowledge, comprising:

[0019] S31: Obtain multiple candidate texts and their initial confidence scores output by the speech recognition model;

[0020] S32: Input each candidate text into the language model that has been fine-tuned by weakened knowledge, and obtain the language model’s score for the fluency and semantic rationality of each candidate text;

[0021] S33: The initial confidence score of each candidate text is weighted and fused with the corresponding language fluency and semantic reasonableness scores to obtain a comprehensive confidence score;

[0022] S34: Sort each candidate text according to the comprehensive confidence score, and select the one with the highest score as the final recognized text output.

[0023] Preferably, calculating a user's decay index involves a two-stage process:

[0024] Phase 1: Based on the user's responses to the standardized frailty screening scale, preliminary scoring and classification are performed to obtain preliminary frailty classification results;

[0025] The second stage: Based on the preliminary frailty classification results and multidimensional feature data, a continuous frailty risk score is calculated using a machine learning prediction model, which serves as the frailty index; the multidimensional feature data includes at least one of physiological indicators, functional performance data, nutritional and metabolic indicators, psychosocial indicators, and cognitive state indicators.

[0026] Preferably, the standardized frailty screening scale includes the Chinese Frailty Screening Scale for the Aged (CFSS-10) and / or the Chinese modified Frail Scale (CMFP).

[0027] Preferably, the feedback process is controlled by a dialogue management agent, which, based on the user's weakness index, historical interaction records, and user emotions, decides to generate feedback content, tone, and level of detail that are appropriate to the user's risk level.

[0028] Secondly, embodiments of the present invention provide a rapid screening and assessment device for frailty in the elderly, comprising:

[0029] A virtual task unit is used to generate a virtual digital doctor avatar and control the virtual digital doctor to guide users to participate in frailty screening through voice and / or video.

[0030] The acquisition unit is used to collect the user's voice answers to screening questions through the voice acquisition module to obtain user voice data;

[0031] The conversion unit is used to recognize the user's voice data through the speech recognition module and convert the voice content into text content; wherein, the speech recognition module incorporates a language model that has been fine-tuned with attenuation knowledge.

[0032] The calculation unit is used to calculate the user's frailty index based on the text content, combined with a preset frailty screening scale and / or a multi-dimensional prediction model.

[0033] The feedback unit is used to feed back the calculated decay index and / or the corresponding evaluation results to the user through the digital human interaction system, and to store the relevant data in the storage unit.

[0034] Thirdly, embodiments of the present invention provide an electronic device, including a processor and a memory. The memory stores one or more computer programs; when the one or more computer programs stored in the memory are executed by the processor, the electronic device is able to implement any of the possible design methods described in the first aspect.

[0035] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method as described in any of the above embodiments.

[0036] Fifthly, embodiments of the present invention also provide a computer program product that, when run on an electronic device, causes the electronic device to perform any possible design method of any of the above aspects.

[0037] Beneficial effects: This invention uses a digital human virtual doctor to automatically complete questionnaire inquiries, guidance, and result feedback, replacing the manual operation mode that requires the full participation of professional medical staff. This not only frees medical staff from repetitive tasks, allowing them to focus on higher-value clinical decisions and interventions, but also makes it possible to conduct large-scale, high-frequency rapid screening in communities, nursing homes, and other locations, significantly improving the coverage and efficiency of public health services.

[0038] Based on realistic digital human figures generated by deep learning and emotionally resonant neurospeech synthesis technology, a friendly, natural, and human-like interactive interface was created, effectively alleviating the tension and boredom experienced by the elderly when facing traditional questionnaires. The interactive dialogue process increased the participation and cooperation of elderly users, thereby collecting more authentic and complete response data, laying a reliable foundation for accurate subsequent assessments.

[0039] The system strictly follows the procedures and scoring rules of standard scales (such as CFSS-10 and CMFP), completely eliminating operational biases caused by different evaluators. More importantly, by introducing a language model fine-tuned with weakened knowledge and combining it with the speech recognition system, it can deeply understand the specific semantic scenarios of screening questions and answers, significantly improving the accuracy of recognizing the user's answer intent in complex linguistic environments (such as accents and ambiguous expressions), and reducing evaluation errors caused by mishearing and misjudgment.

[0040] This invention creatively employs a two-stage computational framework of "standardized questionnaire scoring + multi-dimensional prediction model". The first stage quickly derives easily understandable classification conclusions (healthy / pre-frail / frail) based on the scale, meeting the rapid screening needs of community initial screening; the second stage integrates multi-dimensional objective data such as physiological, functional, and psychological data, and outputs a continuous frailty index (such as a risk probability of 0-1) through a machine learning model, realizing the fine quantification of frailty status and dynamic risk prediction, providing more accurate data support for personalized health management. Attached Figure Description

[0041] Figure 1 A schematic diagram of the method framework for this invention is provided;

[0042] Figure 2This is a block diagram of a device structure provided in one embodiment of this application;

[0043] Figure 3 This is a block diagram of an electronic device structure provided in one embodiment of this application. Detailed Implementation

[0044] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] Example 1

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed following the word and its equivalents, but do not exclude other elements or objects.

[0047] In response to the problems existing in the current technology, such as Figure 1 A rapid screening and assessment method for frailty in the elderly is proposed, including:

[0048] S101. Generate a virtual digital doctor avatar and control the virtual digital doctor to guide users in participating in frailty screening through voice and / or video. Generating a virtual doctor avatar using digital human technology makes the screening process more humane and approachable, increasing user participation and acceptance. Traditional frailty screening for the elderly relies on face-to-face interviews or paper questionnaires, resulting in low user participation and a tedious process. Virtual doctors, through voice and video guidance, can make the screening process more vivid and interesting, increasing the elderly's enthusiasm and cooperation. The voice and / or video guidance not only improves the elderly's experience but also ensures the standardization and consistency of the entire screening process.

[0049] S102. The voice acquisition module collects the user's voice responses to screening questions, obtaining user voice data. Compared with traditional paper questionnaires or click-based electronic scales, voice acquisition allows for more natural interaction with the user, avoiding potential misunderstandings or reluctance to participate that may occur with traditional methods. The collection of voice data provides a real-time interactive method with the user, offering data support for subsequent voice recognition and frailty assessment. It also accommodates the elderly's difficulty in understanding written information, improving the accessibility of screening.

[0050] S103. The user's voice data is recognized by the speech recognition module, and the speech content is converted into text content. The speech recognition module incorporates a language model fine-tuned with attenuation knowledge. The function of the speech recognition module is to convert speech data into text content for subsequent processing. Speech recognition technology recognizes the user's voice input and converts it into text information that can be further analyzed. This eliminates errors in written or spoken expression in traditional screening, enabling the system to automatically process the user's answers. Compared with traditional speech recognition models, this invention uses a language model fine-tuned with attenuation knowledge, improving the accuracy and professionalism of recognition and ensuring the understanding of special vocabulary and sentence structures related to attenuation.

[0051] S104. Based on the text content, and combined with a preset frailty screening scale and / or a multi-dimensional prediction model, calculate the user's frailty index. According to the identified text content, combined with standardized frailty screening scales (such as the CFSS-10 scale and CMFP scale) and multi-dimensional prediction models, the system can calculate the user's frailty index. This step quantitatively assesses the user's frailty status through scales and prediction models, providing a scientific basis for subsequent health interventions. Through the fusion of multi-dimensional data (such as physiological, functional, and psychological data), this step can generate a comprehensive frailty risk score, helping doctors or health managers make more accurate health management decisions.

[0052] S105. The calculated frailty index and / or corresponding assessment results are fed back to the user through the digital human interaction system, and the relevant data is stored in the storage unit. The feedback mechanism is one of the core components of this invention. Through the digital human interaction system, the frailty index and assessment results are promptly fed back to the user. The virtual doctor provides personalized health advice and interventions based on the assessment results, helping the elderly understand their health status and guiding them to take effective health management measures. This feedback method not only enhances user engagement and acceptance but also increases information transparency, enabling the elderly to understand their potential health risks in a timely manner, further promoting their emphasis on health management.

[0053] In some specific embodiments, generating a virtual doctor avatar and controlling the virtual doctor to guide the user in participating in frailty screening via voice and / or video includes:

[0054] Using deep learning generative models, highly realistic digital human appearance, expressions, and lip-syncing animations are generated based on real-person video data.

[0055] Generating Digital Human Appearance: Through deep learning technology and training a generative model using a large amount of real-person video data, the digital human can accurately simulate the appearance features of a real person. This process ensures that the generated virtual doctor has a high degree of realism, presenting delicate details such as skin texture, hair, and clothing texture, enhancing the user's sense of trust and immersion. Generating Facial Expressions and Lip-Sync Animation: By learning from the facial expressions and lip movements of real people, the deep learning model can accurately simulate the facial expressions and lip movements of the virtual doctor. The digital human's facial expressions and lip movements can be adjusted in real time according to different situations (such as the type of question and changes in tone), making the interaction of the virtual doctor more vivid and natural, and improving the user's communication experience with the virtual doctor.

[0056] Using neural speech synthesis technology, speech with emotional and prosodic variations is generated based on text;

[0057] Generating Natural Speech: Employing neural speech synthesis technology (such as end-to-end deep learning models), the system can transform input text into speech with fluent rhythm, intonation variations, and emotional nuance. Compared to traditional concatenated speech synthesis techniques, neural speech synthesis generates more natural and emotional speech, making the virtual doctor's speech more vivid and closer to human tone. Emotion and Prosody Control: Through training the speech synthesis model, the system can adjust the tone, speed, and pauses of the speech according to different communication scenarios. For example, in a comforting context, the speech may be softer and slower, while when providing feedback, the speed and tone may be clearer and firmer. This emotion control effectively enhances the elderly's participation and comfort during the screening process.

[0058] Through a synchronization controller, the generated speech is synchronized in real time with the lip-sync animation, facial expressions, and body movements of the digital human. This achieves speech-visual synchronization: the speech generation module achieves strict time alignment with the lip-sync animation, facial expressions, and body movements of the digital human. This ensures that the lip movements and speech content precisely match during the virtual doctor's speech, avoiding common lip-sync lag or disconnection. It also enhances interactivity: the digital human's facial expressions and body movements (such as nodding and gestures) are integrated with the speech content and emotional context, making the entire communication process more coordinated and natural. For example, when the virtual doctor is answering questions from a user, it can enhance the emotional expression through nodding or gestures, making it easier for the user to understand and participate in the interaction. Finally, it enhances immersion and credibility: the synchronization controller ensures the overall performance of the digital human is smooth and consistent, eliminating inconsistencies between lip movements and speech, and between facial expressions and content, improving the realism of the virtual doctor and thus increasing the user's trust and emotional investment in the virtual doctor.

[0059] In some specific embodiments, the user's voice response to the screening questions is collected by the voice acquisition module to obtain user voice data. The method also includes a step of performing real-time noise reduction processing on the user voice data after the user voice is collected, using a noise reduction model based on a deep neural network.

[0060] Improve the quality of voice data, enhance the system's adaptability to low signal-to-noise ratio environments, reduce the false recognition rate, enhance real-time performance and interactivity, and improve overall screening accuracy.

[0061] In some specific embodiments, the speech recognition module incorporates a language model fine-tuned with weakened knowledge, including:

[0062] S31: Obtain multiple candidate texts and their initial confidence scores output by the speech recognition model;

[0063] In this step, the speech recognition model first processes the user's speech input and generates multiple candidate texts for recognition. Each candidate text has an initial confidence score, representing the probability or credibility of its recognition. The existence of these candidate texts takes into account the ambiguity or different speech expressions that may exist when the speech recognition model processes natural language. By generating multiple candidate texts, the system can take into account the ambiguity in speech, avoid directly outputting potentially incorrect or inaccurate results, and improve the flexibility of subsequent processing.

[0064] S32: Input each candidate text into the language model that has been fine-tuned by weakened knowledge, and obtain the language model’s score for the fluency and semantic rationality of each candidate text;

[0065] In this step, the system evaluates candidate texts using a language model fine-tuned with weakening knowledge. This language model is specially trained to include weakening-related knowledge, terminology, and contextual information, enabling it to understand contexts and expressions closely related to weakening screening. By scoring the fluency and semantic plausibility of each candidate text, the system can determine whether the text grammatically and semantically conforms to the context of weakening screening. For example, weakening screening may involve many medical terms or proper nouns; the language model fine-tuned with weakening knowledge can accurately assess whether candidate texts conform to the knowledge background and practical application requirements of weakening screening. This process improves the accuracy of the speech recognition system in professional fields, ensures grammatical correctness and semantic coherence, and avoids semantic bias in recognition.

[0066] S33: The initial confidence score of each candidate text is weighted and fused with the corresponding language fluency and semantic reasonableness scores to obtain a comprehensive confidence score;

[0067] In this step, the system combines the initial confidence score with the language fluency and semantic reasonableness scores to obtain a comprehensive confidence score. The goal of the weighted fusion process is to comprehensively evaluate the final accuracy of each candidate text based on its credibility and language quality. The initial confidence score reflects the degree of trust the speech recognition model has in the text, while the language fluency and semantic reasonableness scores indicate the text's reasonableness within the context of weakness screening. Through weighted fusion, this step ensures that the final selected recognition text is not only highly credible in speech but also most appropriate in semantics and context. This weighting mechanism improves the overall reliability of recognition, reduces inaccurate recognition due to ambiguity or errors in speech recognition, and ensures that the final output text better meets the actual needs of weakness screening.

[0068] S34: The system sorts all candidate texts according to the comprehensive confidence score and selects the text with the highest score as the final recognition text output. In this step, the system sorts all candidate texts according to the comprehensive confidence score and selects the text with the highest score as the final recognition result output. This process ensures that the text selected by the system is the best overall choice in terms of speech recognition accuracy, language fluency, and semantic reasonableness. By selecting the candidate text with the highest score, the system can output the most suitable recognition result. Even if there are multiple candidate texts in the speech recognition, the final output will be the correct answer that best meets the weakening screening requirements. This enables the system to process the user's voice input efficiently and accurately, and reduces misrecognition caused by differences among multiple candidate texts.

[0069] In some specific embodiments, calculating a user's attrition index involves a two-stage process:

[0070] Phase 1: Based on users' responses to standardized frailty screening scales, preliminary scoring and classification are performed to obtain preliminary frailty classification results. Phase 1 relies on users' responses to standardized frailty screening scales (e.g., CFSS-10 scale, CMFP scale). These scales have been validated and are widely used in frailty screening among older adults. Standardized question design ensures consistency and comparability in the screening process, resulting in high reliability for each user's assessment. Users provide answers to questions on the scales, and the scoring system converts these answers into numerical scores for subsequent analysis.

[0071] By categorizing the scale scores, users are initially classified as "robust," "pre-frail," or "frail." This classification process is simple and efficient, quickly identifying a user's frailty status and providing a foundation for further assessment. This initial frailty classification facilitates rapid screening of a large number of users, especially during initial screening, helping to identify which groups require more in-depth health management.

[0072] This phase, as part of a rapid screening process, simplifies the complexity of frailty assessment. Because it relies on standardized questionnaires, it is suitable for initial screening of large populations, and is particularly effective in settings such as communities or nursing homes.

[0073] The second stage: Based on the preliminary frailty classification results and multidimensional feature data, a continuous frailty risk score is calculated using a machine learning prediction model, which serves as the frailty index; the multidimensional feature data includes at least one of physiological indicators, functional performance data, nutritional and metabolic indicators, psychosocial indicators, and cognitive state indicators.

[0074] In the second stage, in addition to the initial attenuation classification results, the system also incorporates the user's multidimensional feature data, including but not limited to:

[0075] Physiological indicators, such as age, weight, BMI, and blood pressure, reflect the user's basic health status.

[0076] Functional performance data, such as walking speed, grip strength, and sit-up time, reflect the user's physical functional abilities.

[0077] Nutritional and metabolic indicators, such as serum albumin and hemoglobin, reflect the nutritional status of the elderly.

[0078] Psychosocial indicators, such as the Geriatric Depression Scale (GDS) score and social support score, reflect psychological and social support status.

[0079] Cognitive state indicators, such as Mini-Mental State Test (MMSE) scores, are used to assess cognitive function.

[0080] These multidimensional feature data can comprehensively assess a user's health status, relying not only on preliminary classification results, thus ensuring the accuracy and detail of the assessment.

[0081] The second stage combines the aforementioned multidimensional feature data with the preliminary classification results, and utilizes machine learning algorithms (such as regression analysis, decision trees, support vector machines, and random forests) to make more accurate predictions of users' frailty risk. Through learning from historical data, this model can discover complex nonlinear relationships between various health indicators and frailty states.

[0082] The final output, calculated through a machine learning model, is a continuous frailty risk score. Unlike traditional classification models, continuous scoring can reflect the severity of a user's frailty more precisely. For example, the frailty index can be a value between 0 and 1, where closer to 0 indicates good health and closer to 1 indicates severe frailty.

[0083] Machine learning models can generate a specific frailty index for each user. This index is more granular than simple categorization labels (such as "frail" or "robust"), reflecting subtle differences in frailty status among individuals. The continuity of the frailty index enables more personalized health management, facilitating precise health interventions based on specific score results. Based on the frailty index, the system can provide personalized health recommendations to users or health management personnel, such as improving nutrition, increasing exercise, and providing psychological interventions. Different risk levels (such as low, moderate, and high) can trigger different intervention strategies.

[0084] The frailty risk score generated at this stage not only supports immediate screening but also serves as the basis for long-term health management. The frailty index is a dynamic value; as health data is updated, the system can conduct regular assessments, track changes in the user's health status, and help doctors or caregivers adjust health intervention plans in a timely manner.

[0085] In some specific embodiments, the standardized frailty screening scale includes the Chinese Frailty Screening Scale for the Aged (CFSS-10) and / or the Chinese Modified Frail Scale (CMFP).

[0086] The Chinese Frailty Screening Scale-10 (CFSS-10) is a widely used frailty screening tool for the elderly population. It contains 10 concise questions covering core dimensions of frailty, such as physical activity, energy levels, functional performance, weight change, and gait. Because the questions are simple and easy to understand, it is suitable for use by the elderly population, and is especially suitable for large-scale screening.

[0087] The Chinese-modified Frail Scale (CMFP) is a localized improvement of the classic Frail scale, adapted to the characteristics and health status of the elderly population in China. The CMFP scale includes multiple objective measurements (such as grip strength and walking speed) and self-report scales (such as energy levels and physical fatigue), which can comprehensively assess multiple dimensions of frailty.

[0088] In some specific embodiments, the feedback process is controlled by a dialogue management agent, which, based on the user's weakness index, historical interaction records, and user emotions, decides to generate feedback content, tone, and level of detail that are appropriate to the user's risk level.

[0089] 1. Frailty Index:

[0090] Customized Feedback Based on Frailty Index: The frailty index is a continuous value that measures a user's degree of frailty, reflecting the current health status and frailty risk of the elderly. The dialogue management agent determines the level of detail in the feedback content based on the frailty index. For example, for users with a high frailty index, the agent will provide more detailed health assessment results and offer practical suggestions and interventions; while for users with a low frailty index, the feedback may be more concise, offering encouragement and suggestions for maintaining health.

[0091] Feedback is adjusted based on risk level: The decay index value is typically converted into a risk level (e.g., low risk, medium risk, high risk). The agent provides different levels of feedback based on the different risk levels:

[0092] High-risk users: For users with a high frailty index, the AI ​​will provide more detailed and urgent feedback, such as the need for immediate further medical evaluation or lifestyle changes. The tone may be more concerned, emphasizing the urgency of taking action.

[0093] For users with medium risk: For users with higher risk but not critical risk, the feedback may include warnings and suggestions, with a tone that appropriately balances warning and encouragement, guiding users to start paying attention to their health and take necessary intervention measures.

[0094] Low-risk users: For users with a low frailty index, the agent's feedback is more relaxed, and the tone is usually more relaxed and gentle, focusing on encouraging them to continue to maintain their health while providing simple daily health care advice.

[0095] 2. Historical interaction records:

[0096] Optimizing User Experience: Historical interaction records help dialogue management agents understand users' long-term health trends, feedback preferences, and the effects of previous health interventions. Based on historical records, agents can customize personalized feedback content and avoid repetitive or irrelevant suggestions. For example, if a user has previously undergone health interventions for a period of time and has shown some improvement, the agent will highlight these positive developments in its feedback and encourage continued efforts.

[0097] Providing more targeted recommendations: Historical records can also help the agent identify potential personalized needs or preferences of certain users. For example, some users may prefer brief feedback, while others may need detailed health guidance. Based on this information, the agent can adjust the level of detail in the feedback, the way the information is presented, and the frequency of interaction.

[0098] 3. User sentiment:

[0099] Emotion Recognition and Tone Adjustment: By analyzing the emotion of a user's voice or text input, the agent can recognize the user's emotional state (such as anxiety, confusion, worry, etc.). For example, if a user exhibits anxiety, the agent can alleviate the user's emotions through gentle, reassuring language, making them feel cared for and understood. If the user's emotions are more positive, the agent's tone can be more relaxed and encouraging.

[0100] Personalized Emotional Support: Sentiment analysis enables agents to provide more targeted emotional support, especially important for high- or medium-risk users. For these users, agents need to offer not only health intervention suggestions but also enhance user trust and compliance through appropriate tone and emotional communication.

[0101] 4. Generation of feedback content:

[0102] Content generated based on frailty index and risk level: Based on the frailty index and the user's health status, the agent can accurately generate feedback content. For high-risk users, feedback may include urgent health warnings, necessary interventions, and medical advice; for low-risk users, feedback may focus more on encouraging health maintenance and frailty prevention.

[0103] Control of details and information: The level of detail in the feedback is adjusted by the agent based on the user's needs and risk level. For example, for high-risk users who require a deeper understanding, the feedback can be more detailed, including specific physiological and functional manifestations of weakness, intervention suggestions, etc.; while low-risk users may only receive simple health maintenance suggestions.

[0104] 5. Feedback tone and communication strategies:

[0105] Emotion-driven tone adjustment: Tone is a crucial element of effective communication with users. By analyzing user emotions, the intelligent agent appropriately adjusts its tone to align with the user's emotional needs.

[0106] High-risk users: The tone is more concerned and urgent, and may contain elements of encouragement and warning to ensure that users recognize the health risks and take action.

[0107] For users with medium risk: Use a moderate tone, which should both warn and encourage them to take action. The tone may be suggestive and reminding.

[0108] For low-risk users: the tone is relaxed and encouraging, aiming to maintain health rather than urgently needing intervention.

[0109] Tone Variation and Adaptability: By analyzing the effects of user emotions and feedback, the intelligent agent can gradually adapt to the user's emotional needs through multiple interactions, establishing better communication patterns. For example, if a user exhibits high levels of anxiety during an interaction, the intelligent agent will consider emotional support more in subsequent responses, avoiding cold or overly formal feedback.

[0110] In other embodiments of the invention, combined with Figure 2 This invention discloses a rapid screening and assessment device for frailty in the elderly, comprising:

[0111] Virtual task unit 201 is used to generate a virtual digital doctor avatar and control the virtual digital doctor to guide the user to participate in frailty screening through voice and / or video.

[0112] The acquisition unit 202 is used to acquire the user's voice answers to screening questions through the voice acquisition module to obtain user voice data;

[0113] The conversion unit 203 is used to recognize the user's voice data through the speech recognition module and convert the voice content into text content; wherein, the speech recognition module incorporates a language model that has been fine-tuned with attenuation knowledge.

[0114] The calculation unit 204 is used to calculate the user's frailty index based on the text content, combined with a preset frailty screening scale and / or a multi-dimensional prediction model.

[0115] Feedback unit 205 is used to feed back the calculated decay index and / or the corresponding evaluation result to the user through the digital human interaction system, and to store the relevant data in the storage unit.

[0116] All relevant content of each step involved in the above method embodiments can be referenced from the functional description of the corresponding functional module, and will not be repeated here.

[0117] In other embodiments of the present invention, an electronic device 400 is disclosed, such as... Figure 3 As shown, the electronic device may include: one or more processors 401; a memory 402; a display 403; one or more application programs (not shown); and one or more computer programs 404. These devices can be connected via one or more communication buses 405. The one or more computer programs 404 are stored in the memory 402 and configured to be executed by the one or more processors 401. The one or more computer programs 404 include instructions that can be used to perform actions such as... Figures 1 to 2 And the various steps in the corresponding embodiments.

[0118] Through the above description of the embodiments, those skilled in the art will clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0119] In the various embodiments of this invention, the functional units can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0120] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as flash memory, portable hard disk, read-only memory, random access memory, magnetic disk, or optical disk.

[0121] The above description is merely a specific implementation of the embodiments of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the embodiments of the present invention should be covered within the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention should be determined by the protection scope of the claims.

Claims

1. A method for rapid screening and assessment of frailty in the elderly, characterized in that, include: Generate a virtual digital doctor avatar and control the virtual digital doctor to guide the user to participate in frailty screening through voice and / or video; The voice acquisition module collects the user's voice responses to screening questions to obtain user voice data; The user's voice data is recognized by a speech recognition module, and the speech content is converted into text content; wherein, the speech recognition module incorporates a language model that has been fine-tuned with weakening knowledge. Based on the text content, combined with a preset frailty screening scale and / or a multi-dimensional prediction model, the user's frailty index is calculated. The calculated decay index and / or the corresponding assessment results are fed back to the user through the digital human interaction system, and the relevant data is stored in the storage unit.

2. The method for rapid screening and assessment of frailty in the elderly according to claim 1, characterized in that, Generating a virtual digital doctor avatar and controlling the virtual doctor to guide users in frailty screening via voice and / or video includes: Using deep learning generative models, highly realistic digital human appearances, expressions, and lip-syncing animations are generated based on real-person video data. Using neural speech synthesis technology, speech with emotional and prosodic variations is generated based on text; The generated speech is synchronized in real time with the digital human's lip-sync animation, facial expressions, and body movements using a synchronization controller.

3. The method for rapid screening and assessment of frailty in the elderly according to claim 1, characterized in that, The system collects the user's voice responses to screening questions through a voice acquisition module to obtain user voice data. It also includes a step of performing real-time noise reduction processing on the user voice data using a noise reduction model based on a deep neural network after the user's voice is collected.

4. The method according to claim 1, characterized in that, The speech recognition module incorporates a language model fine-tuned with weakened knowledge, including: Obtain multiple candidate texts and their initial confidence scores output by the speech recognition model; Each candidate text is input into the language model that has been fine-tuned by weakened knowledge, and the language model scores the fluency and semantic rationality of each candidate text. The initial confidence score of each candidate text is weighted and fused with the corresponding language fluency and semantic reasonableness scores to obtain the comprehensive confidence score. Based on the comprehensive confidence score, each candidate text is sorted, and the text with the highest score is selected as the final text output.

5. The method for rapid screening and assessment of frailty in the elderly according to claim 1, characterized in that, Calculating a user's decay index involves a two-stage process: Phase 1: Based on the user's responses to the standardized frailty screening scale, preliminary scoring and classification are performed to obtain preliminary frailty classification results; The second stage: Based on the preliminary frailty classification results and multidimensional feature data, a continuous frailty risk score is calculated using a machine learning prediction model, which serves as the frailty index; the multidimensional feature data includes at least one of physiological indicators, functional performance data, nutritional and metabolic indicators, psychosocial indicators, and cognitive state indicators.

6. The method for rapid screening and assessment of frailty in the elderly according to claim 5, characterized in that, The standardized frailty screening scales include the Chinese Frailty Screening Scale for the Aged (CFSS-10) and / or the Chinese modified Frail Scale (CMFP).

7. The method for rapid screening and assessment of frailty in the elderly according to claim 1, characterized in that, The feedback process is controlled by a dialogue management agent, which, based on the user's weakness index, historical interaction records, and user emotions, decides to generate feedback content, tone, and level of detail that are appropriate to the user's risk level.

8. A rapid screening and assessment device for frailty in the elderly, characterized in that, include: A virtual task unit is used to generate a virtual digital doctor avatar and control the virtual digital doctor to guide users to participate in frailty screening through voice and / or video. The acquisition unit is used to collect the user's voice answers to screening questions through the voice acquisition module to obtain user voice data; The conversion unit is used to recognize the user's voice data through the speech recognition module and convert the voice content into text content; wherein, the speech recognition module incorporates a language model that has been fine-tuned with attenuation knowledge. The calculation unit is used to calculate the user's frailty index based on the text content, combined with a preset frailty screening scale and / or a multi-dimensional prediction model. The feedback unit is used to feed back the calculated decay index and / or the corresponding evaluation results to the user through the digital human interaction system, and to store the relevant data in the storage unit.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the computer program is executed by the processor, causes the processor to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images