Intelligent machine old-age care robot system based on artificial intelligence

Abstract

The invention relates to the technical field of old-age service robots, in particular to an intelligent machine old-age service robot system based on artificial intelligence, which comprises the steps of comparing a standard face with feature point coordinates of a real-time image, calculating to obtain an emotion coefficient, and separating past activity items through historical data and the system. Effective interaction items are analyzed in combination with the emotion coefficients of the corresponding time periods, when it is recognized that the user is in a heavy mood state, the system selects a fixed time period according to the current time point, the occurrence frequency of the effective interaction items is counted, the high-frequency items serve as personalized service instructions to interact with the user, and then accurate perception of the user state is achieved; emotional fluctuation is responded in time, the psychological state of the user is improved through high-frequency forward activities, and emotional care is enhanced.

Description

Technical Field

[0001] This invention relates to the field of elderly care service robot technology, and in particular to an intelligent robot system for elderly care based on artificial intelligence. Background Technology

[0002] Intelligent elderly care robots, as a product integrating cutting-edge technologies from multiple fields such as artificial intelligence, mechanical engineering, and sensor technology, are gradually becoming an important means of addressing the challenges of an aging population.

[0003] The existing technology CN118691437A discloses a home-based intelligent elderly care service robot system, including a smart home elderly care environment dedicated gateway, an age-friendly smart home module, a millimeter-wave radar AI anti-fall protection module, an intelligent nursing bed module, a function expansion module, an indoor mobility module, an intelligent meal storage and processing module, an intelligent butler companion module, and a digital management platform. By combining 5G technology, IoT technology, big data and cloud platform technology with robot and AI functions, it constructs a complete new concept of home-based and institutional medical and elderly care.

[0004] However, most elderly care robots have limited capabilities in accurately monitoring and analyzing the daily activities and emotional states of the elderly. They cannot accurately grasp the true feelings and changing needs of the elderly. In terms of personalized service customization, existing robots cannot flexibly adjust service strategies according to the individual differences of the elderly and their state at different times, making it difficult to provide truly precise services that meet the actual needs of the elderly. Summary of the Invention

[0005] The purpose of this invention is to solve the problems in the background art by proposing an intelligent machine elderly care robot system based on artificial intelligence.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: An artificial intelligence-based intelligent robot system for elderly care includes: The information collection module is used to collect basic information about the target users; The dynamic data acquisition module is used to collect dynamic information of the target user, which refers to the target user's daily activities. The emotion recognition module is used to collect the facial information of the target user, obtain a real-time state image, and simultaneously acquire a standard facial image. It selects the feature points of the facial feature region of the target user, determines the coordinate positions of the feature points on the standard facial image and the real-time state image respectively, and obtains the emotion feature coordinates and real-time performance coordinates. The emotion feature coordinates and real-time performance coordinates are combined and processed to obtain the emotion coefficient of the target user at this time. The data storage module is used to store the target user's sentiment index and dynamic information; The state analysis module is used to obtain dynamic information of the target user in the historical time, separate the individual items in the dynamic information, obtain the independent time period of each individual item, calculate the state feature value of the individual item based on the sentiment coefficient in the independent time period, and determine the effective interactive items in the individual item according to the state feature value. The instruction recommendation module is used to determine the target user's mood state based on the target user's emotional coefficient at the current time. If the mood state is somber, a fixed time period is selected based on the current time point, the frequency of occurrence of each effective interaction item in the fixed time period is calculated, and the effective interaction item corresponding to the maximum frequency of occurrence is selected as the personalized service instruction. The command interaction module is used to receive personalized service commands and interact with the target user based on the command content of the personalized service commands.

[0007] As a further aspect of the present invention, the method for obtaining emotional feature coordinates and real-time performance coordinates includes: S1: Based on the image acquisition device, the image information of the target user is acquired, and then the facial position of the target user is identified in the image information, and the image of the facial position is cropped to obtain the real-time status image of the target user. Acquire standard facial images of target users and identify the locations of emotional expressions on the standard facial images; S2: Select any point in the standard facial image and mark it as the origin. Then, based on the origin position, set up a planar coordinate system. Then, select feature points in the emotional expression position and identify the coordinate positions of the feature points in the planar coordinate system and mark them as emotional feature coordinates BTi(Xai, Yai), where i represents different feature points. S3: Obtain the real-time status image, and set a planar coordinate system for the real-time status image according to the placement method of the standard face image. The origin of the planar coordinate system on the real-time status image is completely consistent with the origin and unit length of the standard face image. Identify the coordinates of feature points on the emotional expression location in the real-time state image and label them as real-time expression coordinates BSi(Xbi, Ybi).

[0008] As a further aspect of the present invention, the emotional expression location refers to the morphological changes caused by muscle movements in specific areas of the face, combined with the inherent correlation between emotions and expressions, quantitatively analyzing the location features, and finally matching the corresponding emotion type. The facial feature areas include eyebrows and mouth, and the feature points of the emotional expression location include the eyebrow head position and the mouth corner position, where i=1 and i=2, i=1 represents the eyebrow head position and i=2 represents the mouth corner position. A standard facial image refers to a facial image of a target user in a relaxed state, and the acquisition of a standard facial image requires the guidance and assistance of a professional.

[0009] As a further aspect of the present invention, the method for obtaining the emotion coefficient includes: Select a feature point i, obtain the emotional feature coordinates and real-time performance coordinates of this feature point, and then use the formula... Obtain the offset coefficient of feature point i on the left. offset coefficient of feature point i on the right ; Reuse formula The emotional coefficient W of the target user is obtained, where, This is the proportional coefficient corresponding to the location of emotional expression, and .

[0010] As a further aspect of the present invention, the method for calculating the state characteristic value of a single project includes: Based on the target user's dynamic information, identify each activity item in the dynamic information and mark each individual activity item as a single item. A single item refers to the target user's daily activities or social activities. Arbitrarily select a single item and mark it as the target item. Obtain the time interval in which the target item appears in the dynamic information and mark the single time interval as an independent time period j, where j represents the number of times the target item appears, and j∈[1,m], indicating that the target item has appeared a total of m times in the historical time. Select an independent time period and set a unit time. Use the unit time as the interval time to identify the target user's emotional coefficient within each unit time period. Set time as the X-axis variable and sentiment coefficient as the Y-axis variable, and establish a coordinate system. Mark the sentiment coefficient of each unit time in the independent time period in the coordinate system in sequence to obtain multiple data points. According to the time order, starting from the first data point, calculate the tangent value between adjacent data points based on the coordinate position of each data point. The tangent value is the slope change value between adjacent data points. Once the tangent value is calculated, all tangent values ​​in the independent time period are summed up, and the final value is marked as the state change value.

[0011] As a further aspect of the present invention, the method for determining effective interactive items includes: All individual projects are sequentially labeled as target projects. The state characteristic value of each individual project is calculated. Then, the state characteristic value is identified. Individual projects with positive state characteristic values ​​are labeled as effective interactive projects, individual projects with negative state characteristic values ​​are labeled as negative growth projects, and individual projects with a state value of 0 are labeled as general projects.

[0012] As a further aspect of the present invention, the method for determining the heavy state includes: Obtain the target user's current mood coefficient. If the mood coefficient W is greater than or equal to the mood threshold A1, mark the target user's current mood state as relaxed. Conversely, if the mood coefficient W is less than the mood threshold A1, mark the target user's current mood state as somber.

[0013] As a further aspect of the present invention, the method for obtaining personalized service instructions includes: When the target user's mood is identified as somber, the current time point is identified, and a fixed time period is selected centered on the current time point, with the duration of the fixed time period set to 2 hours. Extract all fixed time periods from the historical time, identify the individual projects that appear in each fixed time period, and count the number of all individual projects, marking them as the total number of projects Dz; Obtain valid interaction items and count the number of times each valid interaction item appears in a fixed time period, and mark it as the single item quantity Dp, where p represents different valid interaction items. Then, use the formula Dp÷Dz=fp to obtain the frequency fp of the occurrence of valid interaction item p. Compare the occurrence frequencies fp of all valid interactive items p within this fixed time period, select the maximum frequency, and use the valid interactive item corresponding to the maximum frequency as the personalized service instruction at this time.

[0014] As a further aspect of the present invention, when the target user's mood state is identified as relaxed, a blank instruction is generated and transmitted to the instruction interaction module. Subsequently, when a blank instruction is received, the instruction interaction module will maintain the current operating state.

[0015] Compared with existing technologies, the advantages of this invention are: This invention collects dynamic data such as basic user information and daily activities, while simultaneously capturing facial feature information. By comparing the feature point coordinates of a standard facial appearance with those of a real-time image, an emotion coefficient is calculated. Using historical data, the system separates past activities and analyzes the emotion coefficients for corresponding time periods to identify effective interaction items. When a user is identified as being in a somber mood, the system selects a fixed time period based on the current time point, statistically analyzes the frequency of effective interaction items, and uses these high-frequency items as personalized service commands to interact with the user. This allows for precise perception of the user's state, a comprehensive understanding of daily activity patterns and emotional changes, and accurate identification of user-preferred activities through data mining. This ensures personalized and applicable service recommendations, timely responses to emotional fluctuations, and improved user mental state through high-frequency positive activities. Ultimately, this enhances the intelligence and humanization of elderly care services, effectively improving the quality of life and well-being of the elderly. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Reference Figure 1 An intelligent robot system for elderly care based on artificial intelligence includes an information acquisition module, a dynamic acquisition module, an emotion recognition module, a data storage module, a status analysis module, an instruction recommendation module, and an instruction interaction module. The information collection module is used to collect basic information of the target user, which refers to the elderly who need to be monitored or accompanied. The basic information includes the target user's identity information, health information, and family information. Furthermore, the identity information includes gender and age, the health information includes the target user's history of diseases and current physical health status, and the family information includes the target user's current family structure, including the target user's children and whether the target user lives alone. Then, the information collection module establishes a one-way communication connection with the dynamic collection module and transmits the target user's basic information to the dynamic collection module. The dynamic acquisition module is used to collect dynamic information of the target user, where dynamic information refers to the target user's daily activities. The target user's dynamic information at this time is then transmitted to the status analysis module and the data storage module respectively. The emotion recognition module is used to collect facial information from target users and analyze the collected facial information to determine the target user's emotion coefficient at the current time. Specific methods for determining the emotion coefficient include: S1: Based on the image acquisition device, the image information of the target user is acquired, and then the facial position of the target user is identified in the image information, and the image of the facial position is cropped to obtain the real-time status image of the target user. Acquire a standard facial image of the target user and identify the location of emotional expression on the standard facial image. The location of emotional expression refers to the morphological changes caused by muscle movement in a specific area of ​​the face. Combined with the inherent relationship between emotions and expressions, the location features are quantitatively analyzed and finally matched to the corresponding emotion type. In this embodiment, the facial feature areas include eyebrows and mouth. It should be further explained that the standard facial image refers to the facial image of the target user in a relaxed state. Specifically, the standard facial image needs to be acquired under the guidance and assistance of professionals. At the same time, the identification of the location of emotional expression adopts the MTCNN model in the convolutional neural network, and then outputs the key points in the facial region. The specific processing method of the MTCNN model is existing technology and will not be elaborated here. S2: Select any point in the standard facial image and mark it as the origin. Then, based on the origin position, set up a planar coordinate system. Then, select feature points in the emotional expression position and identify the coordinate positions of the feature points in the planar coordinate system and mark them as emotional feature coordinates BTi(Xai, Yai), where i represents different feature points. Furthermore, in this embodiment, the feature points of the emotional expression location include the eyebrow head position and the mouth corner position, and i=1, 2, where i=1 represents the eyebrow head position and i=2 represents the mouth corner position. S3: Acquire a real-time status image, and set a planar coordinate system for the real-time status image according to the placement method of the standard facial image. The origin of the planar coordinate system on the real-time status image is completely consistent with the origin and unit length of the standard facial image. Furthermore, the specific value of the unit length is set by those skilled in the art based on big data experience. It should be further explained that when aligning the real-time state image with the standard face image, the real-time state image needs to be processed. The image processing includes operations such as region cropping and alignment to standardize the face region in the real-time state image, thereby reducing pose interference. The specific image processing methods are existing technologies and will not be elaborated here. Then, the coordinates of the feature points on the emotional expression location in the real-time state image are identified and marked as the real-time expression coordinates BSi(Xbi, Ybi). S4: Randomly select a feature point i, obtain the emotional feature coordinates and real-time performance coordinates of this feature point, and then use the formula... Obtain the offset coefficient of feature point i on the left. offset coefficient of feature point i on the right Among them, the larger the offset coefficient, the more relaxed the target user is; conversely, the smaller the offset coefficient, the more tense or angry the target user is. Reuse formula The emotional coefficient W of the target user is obtained, where, This is the proportional coefficient corresponding to the location of emotional expression, and , The specific values ​​were obtained by those skilled in the art through big data calculations; The emotion recognition module then transmits the target user's real-time emotion coefficient W to the data storage module and the instruction recommendation module, respectively. The data storage module is used to store the target user's emotion coefficient and dynamic information, and a one-way communication connection is established between the data storage module and the state analysis module. The state analysis module is used to obtain the target user's sentiment index and dynamic information over historical time, and to determine the target user's effective interaction items. Specifically, the methods for determining effective interaction items include: Based on the target user's dynamic information, each activity item in the dynamic information is identified and each individual activity item is marked as a single item. A single item refers to the target user's daily or social activities, such as doing housework, reading, chatting, etc. The identification of single items in the dynamic information is processed using a multimodal fusion model. The specific processing method is existing technology and will not be elaborated here. Arbitrarily select a single item and mark it as the target item. Obtain the time interval in which the target item appears in the dynamic information and mark the single time interval as an independent time period j, where j represents the number of times the target item appears, and j∈[1,m], indicating that the target item has appeared a total of m times in the historical time. Select an independent time period and set a unit time. Use the unit time as the interval time to identify the emotional coefficient of the target user within each unit time. Specifically, the specific length of the unit time is set by those skilled in the art based on big data experience. Set time as the X-axis variable and sentiment coefficient as the Y-axis variable, and establish a coordinate system. Mark the sentiment coefficient of each unit time in the independent time period in the coordinate system in sequence to obtain multiple data points. According to the time order, starting from the first data point, calculate the tangent value between adjacent data points based on the coordinate position of each data point. The tangent value is the slope change value between adjacent data points. Once the tangent value is calculated, all the tangent values ​​in the independent time period are summed up, and the final value is marked as the state change value. Then, obtain all independent time periods of the target project, average the state change values ​​corresponding to all independent time periods, and mark the obtained average results as state feature values; All individual items are sequentially marked as target items and processed according to the above method to obtain the state feature value of each individual item. Then, the state feature value is identified, and individual items with positive state feature values ​​are marked as effective interaction items, individual items with negative state feature values ​​are marked as negative growth items, and individual items with a state value of 0 are marked as general items. The status analysis module then transmits the target user's valid interaction items to the instruction recommendation module; The instruction recommendation module is used to obtain the target user's sentiment index at the current time, and combined with effective interaction items, to determine the personalized service instructions for the current target user. The specific methods for determining personalized service instructions include: The target user's current mood coefficient is obtained. If the mood coefficient W is greater than or equal to the mood threshold A1, the target user's current mood state is marked as relaxed. Conversely, if the mood coefficient W is less than the mood threshold A1, the target user's current mood state is marked as somber. The specific value of the mood threshold A1 is obtained by those skilled in the art through big data calculation. When the target user's mood is identified as relaxed, a blank instruction is generated and transmitted to the instruction interaction module. When the target user's mood is identified as somber, the current time point is identified, and a fixed time period is selected with the current time point as the center. The length of the fixed time period is set by those skilled in the art based on big data experience. In this embodiment, the length of the fixed time period is set to 2 hours. For example, if the current time point is 9:20, 9:20 is set as the center of the fixed time period, and the fixed time period is from 8:20 to 10:20. Extract all fixed time periods from the historical time, identify the individual projects that appear in each fixed time period, and count the number of all individual projects, marking them as the total number of projects Dz; Obtain valid interaction items and count the number of times each valid interaction item appears in a fixed time period, and mark it as the single item quantity Dp, where p represents different valid interaction items. Then, use the formula Dp÷Dz=fp to obtain the frequency fp of the occurrence of valid interaction item p. Compare the occurrence frequencies fp of all valid interactive items p in this fixed time period, select the maximum frequency, and take the valid interactive item corresponding to the maximum frequency as the personalized service instruction at this time, and transmit it to the instruction interaction module. The instruction interaction module is used to receive blank instructions and personalized service instructions. When a blank instruction is received, the instruction interaction module will maintain its current running state. When a personalized service instruction is received, the module will interact with the target user according to the content of the personalized service instruction, thereby improving the target user's emotional state.

[0019] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An intelligent robotic elderly care system based on artificial intelligence, characterized in that, include: The information collection module is used to collect basic information about the target users; The dynamic data acquisition module is used to collect dynamic information of the target user, which refers to the target user's daily activities. The emotion recognition module is used to collect the facial information of the target user, obtain a real-time state image, and simultaneously acquire a standard facial image. It selects the feature points of the facial feature region of the target user, determines the coordinate positions of the feature points on the standard facial image and the real-time state image respectively, and obtains the emotion feature coordinates and real-time performance coordinates. The emotion feature coordinates and real-time performance coordinates are combined and processed to obtain the emotion coefficient of the target user at this time. The data storage module is used to store the target user's sentiment index and dynamic information; The state analysis module is used to obtain dynamic information of the target user in the historical time, separate the individual items in the dynamic information, obtain the independent time period of each individual item, calculate the state feature value of the individual item based on the sentiment coefficient in the independent time period, and determine the effective interactive items in the individual item according to the state feature value. The instruction recommendation module is used to determine the target user's mood state based on the target user's emotional coefficient at the current time. If the mood state is somber, a fixed time period is selected based on the current time point, the frequency of occurrence of each effective interaction item in the fixed time period is calculated, and the effective interaction item corresponding to the maximum frequency of occurrence is selected as the personalized service instruction. The command interaction module is used to receive personalized service commands and interact with the target user based on the command content of the personalized service commands.

2. The intelligent machine elderly care robot system based on artificial intelligence according to claim 1, characterized in that, Methods for obtaining emotional feature coordinates and real-time performance coordinates include: S1: Based on the image acquisition device, the image information of the target user is acquired, and then the facial position of the target user is identified in the image information, and the image of the facial position is cropped to obtain the real-time status image of the target user. Acquire standard facial images of target users and identify the locations of emotional expressions on the standard facial images; S2: Select any point in the standard facial image and mark it as the origin. Then, based on the origin position, set up a planar coordinate system. Then, select feature points in the emotional expression position and identify the coordinate positions of the feature points in the planar coordinate system and mark them as emotional feature coordinates BTi(Xai, Yai), where i represents different feature points. S3: Obtain the real-time status image, and set a planar coordinate system for the real-time status image according to the placement method of the standard face image. The origin of the planar coordinate system on the real-time status image is completely consistent with the origin and unit length of the standard face image. Identify the coordinates of feature points on the emotional expression location in the real-time state image and label them as real-time expression coordinates BSi(Xbi, Ybi).

3. The intelligent robotic elderly care system based on artificial intelligence according to claim 2, characterized in that, Emotional expression location refers to the morphological changes caused by muscle movements in specific areas of the face, combined with the inherent relationship between emotions and expressions, quantitatively analyzing the location features, and finally matching them to the corresponding emotion type. Facial feature areas include eyebrows and mouth. The feature points of emotional expression location include the inner corner of the eyebrow and the corner of the mouth, where i=1 and i=2, i=1 represents the inner corner of the eyebrow and i=2 represents the corner of the mouth. A standard facial image refers to a facial image of a target user in a relaxed state, and the acquisition of a standard facial image requires the guidance and assistance of a professional.

4. The intelligent robot system for elderly care based on artificial intelligence according to claim 1, characterized in that, Methods for obtaining the sentiment index include: Select a feature point i, obtain the emotional feature coordinates and real-time performance coordinates of this feature point, and then use the formula... Obtain the offset coefficient of feature point i on the left. offset coefficient of feature point i on the right ; Reuse formula The emotional coefficient W of the target user is obtained, where, This is the proportional coefficient corresponding to the location of emotional expression, and .

5. The intelligent robotic elderly care system based on artificial intelligence according to claim 1, characterized in that, The methods for calculating the state characteristic values ​​of a single project include: Based on the target user's dynamic information, identify each activity item in the dynamic information and mark each individual activity item as a single item. A single item refers to the target user's daily activities or social activities. Arbitrarily select a single item and mark it as the target item. Obtain the time interval in which the target item appears in the dynamic information and mark the single time interval as an independent time period j, where j represents the number of times the target item appears, and j∈[1,m], indicating that the target item has appeared a total of m times in the historical time. Select an independent time period and set a unit time. Use the unit time as the interval time to identify the target user's emotional coefficient within each unit time period. Set time as the X-axis variable and sentiment coefficient as the Y-axis variable, and establish a coordinate system. Mark the sentiment coefficient of each unit time in the independent time period in the coordinate system in sequence to obtain multiple data points. According to the time order, starting from the first data point, calculate the tangent value between adjacent data points based on the coordinate position of each data point. The tangent value is the slope change value between adjacent data points. Once the tangent value is calculated, all tangent values ​​in the independent time period are summed up, and the final value is marked as the state change value.

6. The intelligent robotic elderly care system based on artificial intelligence according to claim 5, characterized in that, Methods for identifying effective interactive items include: All individual projects are sequentially labeled as target projects. The state characteristic value of each individual project is calculated. Then, the state characteristic value is identified. Individual projects with positive state characteristic values ​​are labeled as effective interactive projects, individual projects with negative state characteristic values ​​are labeled as negative growth projects, and individual projects with a state value of 0 are labeled as general projects.

7. The intelligent robotic elderly care system based on artificial intelligence according to claim 1, characterized in that, Methods for determining a heavy state include: Obtain the target user's current mood coefficient. If the mood coefficient W is greater than or equal to the mood threshold A1, mark the target user's current mood state as relaxed. Conversely, if the mood coefficient W is less than the mood threshold A1, mark the target user's current mood state as somber.

8. The intelligent robotic elderly care system based on artificial intelligence according to claim 7, characterized in that, Methods for obtaining personalized service instructions include: When the target user's mood is identified as somber, the current time point is identified, and a fixed time period is selected centered on the current time point, with the duration of the fixed time period set to 2 hours. Extract all fixed time periods from the historical time, identify the individual projects that appear in each fixed time period, and count the number of all individual projects, marking them as the total number of projects Dz; Obtain valid interaction items and count the number of times each valid interaction item appears in a fixed time period, and mark it as the single item quantity Dp, where p represents different valid interaction items. Then, use the formula Dp÷Dz=fp to obtain the frequency fp of the occurrence of valid interaction item p. Compare the occurrence frequencies fp of all valid interactive items p within this fixed time period, select the maximum frequency, and use the valid interactive item corresponding to the maximum frequency as the personalized service instruction at this time.

9. The intelligent robotic elderly care system based on artificial intelligence according to claim 1, characterized in that, When the target user's mood is identified as relaxed, a blank instruction is generated and transmitted to the instruction interaction module. After that, when a blank instruction is received, the instruction interaction module will maintain the current running state.

Images