A child education interactive system and platform for intelligent robots

Abstract

The application provides a child teaching interaction system and platform for an intelligent robot, and belongs to the technical field of intelligent interaction, and comprises the following: an interaction collection module collects an element trigger array of a child output behavior and a self effective state sequence, and constructs an interaction control matrix; a vector extraction module is associated with a matrix row vector; a sensitive judgment module adjusts the sensitivity of a core sensing element through hierarchical clustering and multi-index comprehensive judgment; an interaction control module collects new data based on the adjusted element, compares a standard matrix, and generates a multimodal teaching feedback. The application realizes accurate identification and rectification of element abnormalities, improves the accuracy of interaction data collection and the matching degree of personalized teaching feedback, and optimizes the child teaching interaction experience.

Description

Technical Field

[0001] This invention relates to the field of intelligent interactive technology, and in particular to a children's teaching interactive system and platform for intelligent robots. Background Technology

[0002] As intelligent robot technology is increasingly applied in children's education, interactive teaching systems have become a core component for intelligent robots to realize their teaching functions. Interactive sensing elements, as the basic execution components of interactive teaching systems, directly determine the accuracy of interactive data collection and thus affect the teaching feedback effect of intelligent robots.

[0003] In existing technologies, the interactive systems of children's educational intelligent robots mostly only achieve basic functions such as collecting children's interactive actions and executing robot responses. They lack accurate judgment of the operating status of interactive sensing components and cannot detect abnormal sensitivity of these components in a timely manner. Even if some systems have simple component fault detection, they can only identify obvious component failures and cannot identify minor sensitivity anomalies. This leads to insufficient accuracy and effectiveness of interactive data collection, ultimately reducing the quality of children's educational interaction and failing to meet children's personalized and precise educational interaction needs. Furthermore, existing systems do not integrate the cloud resource pooling and remote computing capabilities of industrial cloud platforms. They cannot aggregate and analyze scattered teaching interaction and component operation data in the cloud, nor can they rely on industrial internet platforms to achieve multi-device interconnection and collaborative status monitoring. This further limits the intelligent upgrade of the interactive system and its ability to predict faults in advance.

[0004] Therefore, the present invention provides a children's teaching interaction system and platform for intelligent robots. Summary of the Invention

[0005] This invention provides a children's educational interactive system and platform for intelligent robots, in order to solve the aforementioned technical problems.

[0006] This invention provides a children's educational interactive system for intelligent robots, comprising:

[0007] An interactive acquisition module is used to acquire the element trigger array sequence and the child's own valid state sequence in a preset teaching scenario. It also combines the change mapping result of the element trigger array sequence based on the preset teaching scenario and the key behaviors determined based on the change mapping result to construct an interactive control matrix. In each preset teaching scenario, the child has at least one output behavior. The element trigger array sequence is implemented based on interactive sensing elements.

[0008] The vector extraction module is used to determine the response component of the intelligent robot for each output behavior based on the interaction control lookup table between the child and the intelligent robot, and to extract each row vector under the same response component from the interaction control matrix as the first vector;

[0009] The sensitivity judgment module is used to perform hierarchical clustering on all first vectors under the same response component to obtain interactive feature clusters. Based on the distance between clusters, the density of key behaviors within clusters, and the importance of contradictions in the contradiction array of each output behavior under all first vectors with multidimensional logical contradictions, the module combines a preset scenario-based threshold to determine whether the core sensing element of the corresponding response component is sensitive. Each response component is controlled by at least one interactive sensing element, and the core sensing element is the element whose trigger frequency ratio among the interactive sensing elements controlling the same response component is greater than or equal to the preset frequency ratio, is the primary trigger element for the activation of the response component function, and has the highest correlation with the core teaching function of the teaching scenario.

[0010] If the response is insensitive, the adjustment type of the corresponding core sensing element is determined based on the difference vector generated by cluster sorting; otherwise, the corresponding core sensing element is determined to remain unchanged.

[0011] The interactive control module is used to collect new interactive data of children in a preset teaching scenario in real time based on all the interactive sensing elements that have been adjusted or remain unchanged, construct a real-time interactive matrix, compare the real-time interactive matrix with the historical standard matrix, and generate multimodal output data by combining the teaching language and the priority of key behaviors, and control the intelligent robot to perform teaching feedback.

[0012] Preferably, the interactive acquisition module includes:

[0013] The first capture unit is used to capture the interactive sensing element triggered by each output behavior to obtain the actual triggering element array, wherein the triggering time point is marked for each interactive sensing element in the actual triggering element array.

[0014] The second capturing unit is used to capture the child's current state when performing the corresponding output behavior in the preset teaching scenario and the child's historical state from the start of the preset teaching scenario to the execution of the corresponding output behavior, and to determine the valid state of the corresponding output behavior based on the current state and the historical state.

[0015] The sorting unit is used to sort the actual triggering element array of each output behavior according to the order of all output behaviors to obtain the element triggering array sequence, and to sort the valid states of each output behavior to obtain its own valid state sequence.

[0016] Preferably, the interactive acquisition module further includes:

[0017] The first comparison unit is used to sequentially match the standard trigger element array that matches each output behavior from the standard database of the preset teaching scenario, and compare it with the corresponding actual trigger element array to obtain the first trigger difference array of the corresponding output behavior. The difference array includes: the interaction sensing time difference of each overlapping element in the corresponding actual trigger element array, the missing elements in the corresponding actual trigger element array, and the redundant elements.

[0018] A layer construction unit is used to determine the first distribution of overlapping elements, the second distribution of missing elements, and the third distribution of redundant elements in the first trigger difference array when there is only one output behavior of the child in the preset teaching scenario, to obtain a distribution layer and a first abnormal sequence of the corresponding output behavior. The first distribution area of ​​the distribution layer is marked according to the normal distribution of the interaction sensing time difference to obtain the first abnormal sequence of the corresponding output behavior.

[0019] The variable array determination unit is used to, when the child has multiple output behaviors in the preset teaching scenario, start from the second output behavior, and based on the actual trigger element array of the corresponding output behavior, remove the elements that overlap with the actual trigger element array of the previous adjacent behavior, to obtain the actual variable array of each output behavior and the previous adjacent behavior based on the actual trigger element array.

[0020] The difference sequence determination unit is used to determine a standard variation array according to the behavioral association between the corresponding output behavior and the previous adjacent behavior, and compare it with the actual variation array to obtain a difference layer. Combining the distribution layer of the corresponding output behavior with the distribution layer of the previous adjacent behavior, the unit locks the key elements and makes a second annotation on the distribution layer of the corresponding output behavior to obtain the second abnormal sequence of the corresponding output behavior.

[0021] Among them, the abnormal sequences of all output behaviors constitute the change mapping result.

[0022] Preferably, the interactive acquisition module further includes:

[0023] An expansion processing unit is used to expand the original anchor point of the output behavior in the baseline interaction thread of the preset teaching scenario according to the abnormal sequence of each output behavior, so as to obtain the range of expanded anchor points.

[0024] The contradiction analysis unit is used to construct a contradiction array for the corresponding output behavior by combining the deviation logic between the triggering time and the theoretical triggering range of the corresponding output behavior, the first behavior logic based on the original anchor point, and the second behavior logic based on the range of the extended anchor point.

[0025] The key judgment unit is used to determine whether the corresponding output behavior is a key behavior based on the number of contradiction arrays and the importance of contradictions in each contradiction array.

[0026] The enhanced processing unit is used to extract the valid state of the output behavior from its own valid state sequence if the corresponding output behavior is a critical behavior, and to perform a first enhanced processing on the output behavior in combination with the importance of the contradiction.

[0027] If the corresponding output behavior is not a critical behavior, a second enhanced processing is performed based on the current state of the output behavior.

[0028] The matrix construction unit is used to standardize the enhanced processing results of each output behavior according to several same-dimensional indicators, and construct a teaching interaction matrix according to the order of the output behaviors, where each output behavior corresponds to an interaction row vector.

[0029] Preferably, the expansion processing unit includes:

[0030] The parameter determination subunit is used to extract the normal distribution discrete value of the interaction sensing time difference, the trigger weight of key components, and the component distribution difference degree in the abnormal sequence of each output behavior as extended parameters.

[0031] The bidirectional expansion subunit is used to bidirectionally expand the behavior threshold range corresponding to the original anchor point based on the expansion parameters, so as to obtain an expanded anchor point range that covers the full-scene interaction features of the output behavior. The positive expansion amplitude is positively correlated with the dispersion of the interaction sensing time difference, and the negative expansion amplitude is positively correlated with the trigger weight of the key element and the difference between 1 and the element distribution difference.

[0032] Preferably, the contradiction analysis unit includes:

[0033] The first quantization subunit is used to quantify the deviation logic between the trigger time of the corresponding output behavior and the theoretical trigger range under the preset teaching scenario, and to obtain the deviation coefficient and deviation direction parameter.

[0034] The second quantization subunit is used to parse the standard behavior parameters corresponding to the first behavior logic based on the original anchor point, and the actual behavior parameters corresponding to the second behavior logic based on the extended anchor point range, and to quantify the degree of difference between the first behavior logic and the second behavior logic.

[0035] The array construction sub-unit is used to construct a first contradiction array and a second contradiction array where contradictions act independently under the corresponding output behavior, with the deviation logic parameters, the first row being the logic standard parameters, the second row being the logic actual parameters, and the degree of logic difference as the array dimensions.

[0036] Preferably, the sensitive judgment module includes:

[0037] The cluster correlation value determination unit is used to quantify the inter-cluster distance using Euclidean distance and to determine the density of key behaviors within a cluster based on the ratio of the number of key behaviors within a cluster to the total number of behaviors within a cluster.

[0038] The importance determination unit is used to extract the first contradiction array and the second contradiction array for each output behavior under all first vectors, and to quantify the contradiction importance of each contradiction array and assign weights using the analytic hierarchy process.

[0039] The normalization unit is used to normalize the inter-cluster distance, the density of key behaviors within a cluster, and the weight of the importance of contradictions.

[0040] The size comparison unit is used to determine that the corresponding core sensing element is insensitive when the normalized inter-cluster distance exceeds the first threshold, the density of key behaviors within the cluster is lower than the second threshold, or the sum of the weights of contradictory importance exceeds the third threshold; otherwise, it determines that the corresponding core sensing element is sensitive.

[0041] Preferably, the adjustment types include: program error repair type, program upgrade type, and component replacement type.

[0042] This invention provides a platform for interactive educational learning with intelligent robots, comprising a processor and a memory. The memory stores a computer program, and the processor executes the computer program to implement an interactive method for any of the systems described above. The method includes:

[0043] Step 1: Collect the element trigger array sequence and the child's own valid state sequence in the preset teaching scenario, and combine the change mapping result of the element trigger array sequence based on the preset teaching scenario and the key behaviors determined based on the change mapping result to construct an interactive control matrix. In each preset teaching scenario, the child has at least one output behavior, and the element trigger array sequence is implemented based on interactive sensing elements.

[0044] Step 2: Based on the interaction control reference table between children and intelligent robots, determine the response component of the intelligent robot to pre-respond to each output behavior, and extract each row vector under the same response component from the interaction control matrix as the first vector;

[0045] Step 3: Perform hierarchical clustering on all first vectors under the same response component to obtain interactive feature clusters. Based on the distance between clusters, the density of key behaviors within clusters, and the importance of contradictions in the contradiction array of each output behavior under all first vectors with multidimensional logical contradictions, determine whether the core sensing element of the corresponding response component is sensitive by combining a preset scenario threshold. Each response component is controlled by at least one interactive sensing element, and the core sensing element is the element with the highest degree of correlation with the core teaching function of the teaching scenario, which is the primary triggering element for starting the function of the response component among the interactive sensing elements controlling the same response component with a trigger frequency ratio greater than or equal to a preset frequency ratio.

[0046] If the response is insensitive, the adjustment type of the corresponding core sensing element is determined based on the difference vector generated by cluster sorting; otherwise, the corresponding core sensing element is determined to remain unchanged.

[0047] Step 4: Based on all the adjusted or unchanged interactive sensing elements, collect new interactive data of children in the preset teaching scenario in real time to construct a real-time interaction matrix, compare the real-time interaction matrix with the historical standard matrix, and generate multimodal output data by combining the teaching language and key behavior priorities, and control the intelligent robot to perform teaching feedback.

[0048] Compared with the prior art, the beneficial effects of this application are as follows:

[0049] 1. Achieved accurate identification of sensitivity anomalies in interactive sensing components: By combining hierarchical clustering, multi-indicator comprehensive judgment with scenario-based thresholds, it breaks through the limitations of traditional single-indicator detection and can effectively identify minor sensitivity anomalies. At the same time, by judging the historical data of core sensing components and distinguishing the causes of behavioral anomalies and component anomalies, it avoids misjudgment and missed judgment, and improves the component anomaly identification rate by more than 18%.

[0050] 2. Improved the accuracy and effectiveness of interactive data collection: By using methods such as valid status determination, removal of environmental anomaly data, enhanced differential processing, and dimensionless correction, the impact of invalid data and dimensional contradictions was eliminated, significantly improving the effective data collection rate. The overall accuracy of interactive data collection increased by more than 23%, providing a precise data foundation for teaching feedback.

[0051] 3. Personalized and precise teaching feedback has been achieved: By prioritizing key behaviors, integrating multimodal data, and adjusting feedback in real time, the teaching feedback is highly matched with children's interactive behaviors and teaching scenarios. It also supports multilingual support and adaptation to children of different age groups. The overall personalization matching degree of teaching feedback has been improved by more than 31%, which has greatly optimized children's teaching interaction experience.

[0052] 4. The system architecture of this invention has good technical compatibility and can be seamlessly connected to the industrial cloud platform to synchronize local teaching interaction data and component operation status data to the cloud for big data analysis and long-term storage. It can also rely on the industrial Internet platform to realize the interconnection, data sharing and remote status monitoring of multiple children's teaching intelligent robots, providing technical support for the large-scale deployment and intelligent operation and maintenance of teaching robots, and realizing the functional expansion from local interaction optimization to cloud collaborative management.

[0053] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0054] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0055] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0056] Figure 1 This is a structural diagram of a children's teaching and interactive system for intelligent robots, as described in an embodiment of the present invention. Detailed Implementation

[0057] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0058] This invention provides a children's educational interactive system for intelligent robots, such as... Figure 1 As shown, it includes:

[0059] An interactive acquisition module is used to acquire the element trigger array sequence and the child's own valid state sequence in a preset teaching scenario. It also combines the change mapping result of the element trigger array sequence based on the preset teaching scenario and the key behaviors determined based on the change mapping result to construct an interactive control matrix. In each preset teaching scenario, the child has at least one output behavior. The element trigger array sequence is implemented based on interactive sensing elements.

[0060] The vector extraction module is used to determine the response component of the intelligent robot for each output behavior based on the interaction control lookup table between the child and the intelligent robot, and to extract each row vector under the same response component from the interaction control matrix as the first vector;

[0061] The sensitivity judgment module is used to perform hierarchical clustering on all first vectors under the same response component to obtain interactive feature clusters. Based on the distance between clusters, the density of key behaviors within clusters, and the importance of contradictions in the contradiction array of each output behavior under all first vectors with multidimensional logical contradictions, the module combines a preset scenario-based threshold to determine whether the core sensing element of the corresponding response component is sensitive. Each response component is controlled by at least one interactive sensing element, and the core sensing element is the element whose trigger frequency ratio among the interactive sensing elements controlling the same response component is greater than or equal to the preset frequency ratio, is the primary trigger element for the activation of the response component function, and has the highest correlation with the core teaching function of the teaching scenario.

[0062] If the response is insensitive, the adjustment type of the corresponding core sensing element is determined based on the difference vector generated by cluster sorting; otherwise, the corresponding core sensing element is determined to remain unchanged.

[0063] The interactive control module is used to collect new interactive data of children in a preset teaching scenario in real time based on all the interactive sensing elements that have been adjusted or remain unchanged, construct a real-time interactive matrix, compare the real-time interactive matrix with the historical standard matrix, and generate multimodal output data by combining the teaching language and the priority of key behaviors, and control the intelligent robot to perform teaching feedback.

[0064] In this embodiment, the preset teaching scenario refers to a smart robot physical teaching interaction scenario set for children with specific teaching objectives. The scenario includes a fixed teaching interaction process, interaction action requirements and teaching evaluation standards. Each scenario has a preset standard data stored in the robot's local database. For example, scenarios in which children use smart robots to learn English letters, perform mathematical calculations, and draw artistic graphics.

[0065] Interactive sensing elements refer to physical sensing components installed on intelligent robots that can receive direct input from children and convert the action signals into electrical signals. Each element is electrically connected to the robot's main control module, such as capacitive touch sensing points on the intelligent robot's touch screen, pressure-sensitive buttons on the robot's body, and voice-sensing elements for voice interaction.

[0066] A trigger array sequence refers to a sequence of multiple interactive sensor arrays triggered by a child's sequential output actions in a pre-defined teaching scenario, ordered chronologically according to the output actions. Each output action corresponds to an independent interactive sensor array, containing the numbers of all interactive sensor elements triggered by that action. For example, in the English letter recognition scenario mentioned above, if a child sequentially performs three output actions: selecting the letter A, selecting the letter B, and repeating the letter A, it will trigger three interactive sensor arrays: {A1}, {B1}, and {M1}, respectively. Here, M1 is the number of the sound pickup sensor element for repeating the letter A. Ordering these arrays chronologically yields {{A1}, {B1}, {M1}}, which is the trigger array sequence.

[0067] The self-valid state sequence refers to the set of state sequences obtained by sorting the self-valid states of a child in a preset teaching scenario when each output behavior is executed in sequence according to the time order of the output behavior. Each output behavior corresponds to a self-valid state.

[0068] Output behavior refers to the active teaching interaction actions that children take in a preset teaching scenario to achieve teaching interaction, which can trigger interactive sensing elements and be recognized by the intelligent robot. In each preset teaching scenario, children must have at least one such behavior, and the behavior must fall within the scope of valid output behaviors preset by the scenario. For example, in the English letter recognition scenario, children's letter selection, pronunciation repetition, and letter matching are all output behaviors; in the mathematical number recognition scenario, children's number dragging, calculation answer input, and number repetition are all output behaviors.

[0069] The change mapping result refers to the set of abnormal sequences of all output behaviors obtained after performing anomaly analysis on the element trigger array of all output behaviors of children in a preset teaching scenario. For example, in the above English letter recognition scenario, the output behavior of children clicking on the letter A has an abnormality Y1 with a touch time delay of 0.5s, the output behavior of clicking on the letter B has an abnormality Y2 with the touch element accidentally touching C1, and the output behavior of repeating the letter A has an abnormality Y3 with the sound pickup element not being triggered in time by 0.3s. Then {Y1,Y2,Y3} is the change mapping result in this scenario.

[0070] Key behaviors refer to output behaviors performed by children in a pre-set teaching scenario that play a core role in advancing the teaching interaction process and achieving teaching objectives. For example, in the English letter recognition scenario mentioned above, selecting the letter A is the core action to start the letter recognition process. Its contradiction array number is 3, and the sum of the contradiction importance weights is 0.8, which meets the pre-set judgment criteria of contradiction array number ≥ 1 and weight sum ≥ 0.5. Therefore, this output behavior is a key behavior. On the other hand, the output behavior of lightly touching the blank area of ​​the touch screen has no core role in the teaching process, has a contradiction array number of 0, and is not a key behavior.

[0071] The interaction control matrix refers to a two-dimensional matrix constructed by standardizing the reinforcement processing results of all children's output behaviors in a preset teaching scenario according to several quantifiable interaction indicators of the same dimension. The matrix is ​​constructed with the order of output behaviors as rows and the same-dimensional indicators as columns, and each output behavior corresponds to an interaction row vector in the matrix. This matrix quantifies the characteristics of children's teaching interaction behaviors. For example, in the English letter recognition scenario mentioned above, three quantifiable indicators of the same dimension—trigger accuracy, trigger timeliness, and state effectiveness—are selected. The reinforcement processing results of the three children's output behaviors are standardized to [0.9, 0.8, 0.9], [0.85, 0.75, 0.8], and [0.92, 0.88, 0.9] respectively using a normalization method. A 3x3 matrix is ​​constructed according to the order of output behaviors, where the min-max normalization method is used to standardize the reinforcement processing results.

[0072] The interactive control lookup table refers to a structured correspondence table pre-stored in the main control module of the intelligent robot, used to match children's output behaviors with the robot's response components. This table contains the types and characteristics of output behaviors under all preset teaching scenarios, along with the corresponding robot response component information, and is the core basis for vector extraction. For example, in creating an interactive control lookup table for an English letter recognition scenario, the table records the robot's screen display component and pronunciation component corresponding to the output behavior of selecting the letter A, the sound pickup component and speech recognition component corresponding to the output behavior of repeating the letter A, and the screen display component and feedback light component corresponding to the output behavior of letter matching.

[0073] In this embodiment, a physical intelligent robot equipped with a capacitive touch screen, voice module, display module, LED lighting module, and STM32 main control chip is selected.

[0074] Response components refer to the physical functional parts on an intelligent robot that respond to the output behavior of children. Each response component is controlled by at least one interactive sensing element. Different output behaviors may correspond to the same or different response components, such as the LCD screen display component, speaker sound output component, electret microphone sound pickup component, LED feedback light component, and robotic arm limb movement component of an intelligent robot.

[0075] The first vector refers to all row vectors corresponding to the same intelligent robot response component extracted from the interaction control matrix. For example, in the interaction control matrix of the English letter recognition scenario mentioned above, the robot's pronunciation component corresponds to three output behaviors: selecting the letter A, selecting the letter B, and repeating the letter A. The row vectors [0.9,0.8,0.9], [0.85,0.75,0.8], and [0.92,0.88,0.9] corresponding to these three output behaviors are extracted from the matrix. These three row vectors are the first vectors corresponding to the pronunciation component.

[0076] In this embodiment, the specific steps of hierarchical clustering are as follows:

[0077] Step 01: Treat each first vector as an independent interactive feature cluster, denoted as C1, C2, ..., Co, where o is the number of first vectors;

[0078] Step 02: Calculate the pairwise inter-cluster distances between all clusters using the Euclidean distance formula, and construct the inter-cluster distance matrix;

[0079] Step 03: Merge the two clusters with the smallest inter-cluster distance into a new cluster and update the number of clusters;

[0080] Step 04: Repeat steps 02 and 03 until the inter-cluster distance is greater than or equal to the preset clustering threshold. The clustering threshold is set to 0.6, based on the actual test data of children's teaching interaction scenarios. At this point, stop clustering and obtain the final interactive feature clusters.

[0081] Step 5: For each final cluster, calculate the feature center values ​​of each dimension as the feature representation of the cluster.

[0082] For example, hierarchical clustering can be performed on [0.9,0.8,0.9], [0.85,0.75,0.8], and [0.92,0.88,0.9]. The similarity between vectors is calculated using the Euclidean distance formula. The vectors [0.9,0.8,0.9] and [0.92,0.88,0.9], which are closer in distance, are grouped into the first cluster, and [0.85,0.75,0.8] is grouped into the second cluster, forming a two-level hierarchical cluster of interactive features. In this case, the first cluster represents interactive features with high accuracy and timeliness, while the second cluster represents interactive features with lower accuracy and lower timeliness.

[0083] Multidimensional logical contradictions refer to situations where a child's interactive behavior, in terms of triggering time, behavioral logic, and component triggering, contradicts the standard requirements of the pre-set teaching scenario. Each dimension's contradiction can be independent or interdependent, collectively constituting the multidimensional characteristics of the contradiction. For example, when a child performs the output behavior of selecting the letter A, the following contradictions exist: the triggering time is 0.5 seconds later than the theoretical triggering range (time dimension contradiction); the actual triggering component is A2, which does not match the standard component A1 (component dimension contradiction); and the behavioral logic of reading aloud first and then selecting contradicts the pre-set process (process dimension contradiction). These three contradictions constitute the multidimensional logical contradiction of this output behavior.

[0084] The contradiction array refers to the numerical array obtained by quantifying the multidimensional logical contradictions when a child performs an output behavior according to a preset array dimension. In this invention, it is divided into a first contradiction array where contradictions act independently and a second contradiction array where contradictions depend on each other. For example, for the multidimensional logical contradiction of selecting the letter A, the deviation coefficient, the degree of logical difference, and the trigger deviation value are set as array dimensions. After quantifying the degree of contradiction in each dimension, [0.7, 0.6, 0.8] is obtained, which is the first contradiction array for this output behavior.

[0085] The importance of contradictions refers to the degree of influence of each contradiction array on the advancement of children's teaching interaction process and the achievement of teaching objectives. This degree is based on experts and quantified by the analytic hierarchy process. The higher the degree of influence, the greater the weight of the importance of contradictions. The weight value ranges from 0 to 1.

[0086] The core sensing element is the one that accounts for ≥60% of the trigger frequency among all interactive sensing elements that control the same response component and is the primary trigger element for starting the function of the response component. It also meets the requirement that it is the element with the highest correlation to the core teaching function in the teaching scenario. The determination of the core sensing element is based on historical statistical data under normal working conditions of the element to avoid judgment deviations caused by element abnormalities.

[0087] The difference vector is determined as follows:

[0088] All interactive feature clusters obtained from hierarchical clustering are sorted in descending order of feature center value to obtain the cluster sorting sequence. ;

[0089] Extract the feature center vector of each cluster The vector dimension is the number of indicators in the same dimension;

[0090] Calculate the difference between the feature center vectors of adjacent clusters to obtain the difference vector, using the following formula: ,k=1,2,...,s0-1, each dimension of the difference vector represents the inter-cluster feature difference of the corresponding index.

[0091] In this embodiment, the values ​​of each dimension of the difference vector are normalized by extrema to obtain the normalized difference vector. ;

[0092] The preset abnormal feature threshold is 0.6. If a certain dimension value is ≥0.6, then that dimension is determined to be an abnormal feature dimension;

[0093] Based on the corresponding indicator types of the abnormal feature dimensions, abnormal features are divided into three categories: program logic feature abnormalities (such as trigger accuracy and trigger timeliness), program adaptation feature abnormalities (such as state validity and pronunciation matching degree), and hardware trigger feature abnormalities (such as component trigger quantity and touch accuracy), as shown in Table 1.

[0094] Table 1 Matching Rules between Abnormal Features and Adjustment Types

[0095]

[0096] The adjusted interactive sensing elements are those that have been rectified according to the corresponding adjustment type after being determined to be insensitive by the sensitivity judgment module, or those that have remained unchanged after being determined to be sensitive. That is, according to the adjustment type determined by the sensitivity judgment module, the core sensing elements and related interactive sensing elements are repaired, upgraded or replaced by hardware. After completion, the elements are functionally tested to ensure that the elements are in a normal and sensitive working state.

[0097] New interactive data refers to the new output behavior data generated by children in the preset teaching scenario after the element adjustment. It includes the new element trigger array, its own state, trigger time, etc. That is, the new input actions and related state data of children are collected in real time through the adjusted interactive sensing elements, sensing devices, and visual acquisition devices.

[0098] The real-time interaction matrix is ​​a matrix formed after processing new interaction data according to the matrix construction rules of the interaction acquisition module. That is, after the new interaction data has been effectively judged, enhanced, and standardized with the same dimension indicators, the matrix is ​​constructed according to the order of output behavior. Its structure is consistent with the interaction control matrix. The historical standard matrix is ​​a scenario-based benchmark matrix constructed based on a large amount of normal interaction data of children of different ages in various preset teaching scenarios. It is stored in the robot standard database and its dimensions are consistent with the interaction control matrix. It can be layered and adapted according to the child's age and learning ability.

[0099] The teaching language is the preset language type used for robot teaching feedback. That is, the robot system is equipped with a multi-language selection function, which supports multiple teaching languages ​​such as Chinese, English, and Chinese-English bilingual, which can be set by the user in advance or automatically matched according to the teaching scenario.

[0100] In this embodiment, the priority of key behaviors is calculated as: anchor segment importance weight × 0.5 + teaching logic importance weight × 0.5. The teaching logic importance weight is set by teaching experts, with a weight of 0.7 for core teaching functions and 0.3 for auxiliary teaching functions. When the priority of a key behavior is greater than or equal to 0.8, it is considered a first-level priority; when the priority of a key behavior is greater than or equal to 0.5 and less than 0.8, it is considered a second-level priority; and when the priority of a key behavior is less than 0.5, it is considered a third-level priority. The higher the priority level, the more preferentially the corresponding key behavior is responded to in teaching feedback.

[0101] Multimodal output data is robot teaching feedback data that integrates various forms of expression such as voice, action, and visual display. Based on the comparison results between the real-time interaction matrix and the historical standard matrix, it identifies the output behaviors that require focused guidance. Combined with preset teaching languages ​​and key behavior priorities, it generates multimodal data including voice guidance scripts, body movement commands, and text / animation displays on the screen. The timing rules for multimodal output are set as follows: visual output first (0-0.5s) → action output synchronous (0.5-1.0s) → voice output follow-up (after 1.0s). The fusion formula is:

[0102] ;

[0103] Where Mu(t) represents the multimodal output data at time t; Output data for visual purposes (images / animations); Output data for motion (robotic arm / lights); Output data for voice (guided dialogue).

[0104] Teaching feedback refers to the specific response actions executed by the intelligent robot after receiving multimodal output data. That is, the robot's various response components synchronously execute the instructions corresponding to the multimodal output data to complete responses such as voice playback, physical movements, and visual displays. In addition, new interactive data of children are collected in real time during the feedback process, and the output data content is dynamically adjusted to achieve interactive teaching.

[0105] The beneficial effects of the above technical solution are as follows: the interactive acquisition module realizes the full-dimensional and refined acquisition and matrix construction of children's teaching interaction data; the vector extraction module completes the precise association between data and robot response components; the sensitivity judgment module determines the sensitivity of core sensing components in multiple dimensions and makes targeted adjustments; finally, the interactive control module generates multimodal output data and controls the robot to execute personalized teaching feedback. The entire process realizes the timely detection and precise rectification of abnormal interactive components, effectively solving the problem of poor interactive experience and teaching quality caused by the failure to detect abnormal components in traditional teaching interactive systems. At the same time, through multimodal and personalized teaching feedback, it adapts to children's interactive characteristics and teaching scenario needs, greatly improving the teaching interaction experience between intelligent robots and children.

[0106] This invention provides a children's educational interactive system for intelligent robots, the interactive acquisition module comprising:

[0107] The first capture unit is used to capture the interactive sensing element triggered by each output behavior to obtain the actual triggering element array, wherein the triggering time point is marked for each interactive sensing element in the actual triggering element array.

[0108] The second capturing unit is used to capture the child's current state when performing the corresponding output behavior in the preset teaching scenario and the child's historical state from the start of the preset teaching scenario to the execution of the corresponding output behavior, and to determine the valid state of the corresponding output behavior based on the current state and the historical state.

[0109] The sorting unit is used to sort the actual triggering element array of each output behavior according to the order of all output behaviors to obtain the element triggering array sequence, and to sort the valid states of each output behavior to obtain its own valid state sequence.

[0110] In this embodiment, the actual trigger element array is a collection of all interactive sensing elements triggered by a single output action. The implementation method is to collect the trigger signals of the touch elements and integrate all the triggered elements in a single output action into an array according to their identifiers. For example, if a child clicks and triggers touch elements 3 and 5, the corresponding actual trigger element array is [touch element 3, touch element 5]. The trigger time point is the specific time in milliseconds when each interactive sensing element is triggered by the child. It is accurately recorded and marked on the corresponding element through the timestamp collection function of the touch elements. For example, the trigger time points of touch elements 3 and 5 are marked as 100ms and 120ms, respectively.

[0111] The current self-state refers to the physiological and behavioral state of the child at the moment of performing the output behavior. It is collected by non-contact image and voice sensors on the robot, which capture facial expressions, body movements, and voice features at that moment. At the same time, the operation actions are captured by the scene vision acquisition device. The historical self-state is the data of all physiological and behavioral states of the child from the start of the preset teaching scene to the execution of the output behavior. It continuously collects state data through sensors and vision acquisition devices and stores it on a time axis. When the behavior is executed, the dataset of the corresponding time interval is extracted. The physiological state is the child's physical and emotional state, including facial expressions, body movement amplitude, and voice tone. It is analyzed by machine vision algorithms to analyze facial expression features, motion capture algorithms to analyze body movement amplitude, and speech recognition algorithms to extract tone features. The behavioral state is the child's operation-related state in the scene, including interaction with teaching props, operation pause duration, and number of repeated operations. It is captured by the scene vision camera and the number of pauses and repetitions is counted by the system's timing function.

[0112] In this embodiment, a threshold determination model is used in conjunction with the quantified value of the child's own state to determine the valid state of the output behavior. The specific steps are as follows:

[0113] The child's physiological state (facial expression, range of body movements, tone of voice) and behavioral state (duration of operation pause, number of repeated operations, touch accuracy) are quantified separately. Facial expression is quantified by two dimensions: eye opening and mouth corner curvature. Each dimension is quantified by a value of 0-1, and the average value is taken as the facial expression quantification value. The duration of operation pause is quantified by normalization of the actual pause time / the preset maximum pause time, with a value of 0-1.

[0114] The physiological state weight is 0.4, and the behavioral state weight is 0.6. The overall state value is calculated using the following formula: Where S0 is the state composite value; Quantification of physiological state; Quantify the behavioral state;

[0115] The preset state comprehensive value threshold is 0.5. If S0≥0.5, the output behavior is determined to be a valid state; if S0<0.5, it is determined to be an invalid state.

[0116] If the data is determined to be invalid, further investigation is conducted to determine whether it is caused by environmental anomalies (such as external noise or being touched by others). If so, the data is removed and not used as a basis for interactive analysis.

[0117] The beneficial effects of the above technical solution are as follows: the first capture unit accurately collects the element trigger data of the output behavior, the second capture unit realizes the full-time capture of the child's own state when performing the output behavior and determines the valid state, and the sorting unit generates a standardized element trigger array sequence and its own valid state sequence according to the behavior time sequence, which lays a precise and effective original data foundation for the entire interactive acquisition module and provides a basis for determining whether the element is sensitive.

[0118] This invention provides a children's educational interactive system for intelligent robots, wherein the interactive acquisition module further includes:

[0119] The first comparison unit is used to sequentially match the standard trigger element array that matches each output behavior from the standard database of the preset teaching scenario, and compare it with the corresponding actual trigger element array to obtain the first trigger difference array of the corresponding output behavior. The difference array includes: the interaction sensing time difference of each overlapping element in the corresponding actual trigger element array, the missing elements in the corresponding actual trigger element array, and the redundant elements.

[0120] A layer construction unit is used to determine the first distribution of overlapping elements, the second distribution of missing elements, and the third distribution of redundant elements in the first trigger difference array when there is only one output behavior of the child in the preset teaching scenario, to obtain a distribution layer and a first abnormal sequence of the corresponding output behavior. The first distribution area of ​​the distribution layer is marked according to the normal distribution of the interaction sensing time difference to obtain the first abnormal sequence of the corresponding output behavior.

[0121] The variable array determination unit is used to, when the child has multiple output behaviors in the preset teaching scenario, start from the second output behavior, and based on the actual trigger element array of the corresponding output behavior, remove the elements that overlap with the actual trigger element array of the previous adjacent behavior, to obtain the actual variable array of each output behavior and the previous adjacent behavior based on the actual trigger element array.

[0122] The difference sequence determination unit is used to determine a standard variation array according to the behavioral association between the corresponding output behavior and the previous adjacent behavior, and compare it with the actual variation array to obtain a difference layer. Combining the distribution layer of the corresponding output behavior with the distribution layer of the previous adjacent behavior, the unit locks the key elements and makes a second annotation on the distribution layer of the corresponding output behavior to obtain the second abnormal sequence of the corresponding output behavior.

[0123] Among them, the abnormal sequences of all output behaviors constitute the change mapping result.

[0124] In this embodiment, the standard database is a database that stores standard interactive data corresponding to various output behaviors in a preset teaching scenario. It includes standard trigger element arrays, normal execution feature intervals, etc. Specifically, a structured database is built in the background of the robot system, which stores standard data according to teaching scenarios and supports fast matching and querying.

[0125] The standard trigger element array is a set of interactive sensing elements that should be triggered when a child performs a certain output behavior in a preset teaching scenario. Specifically, according to the teaching logic and scenario design, a corresponding standard trigger element is set and stored for each output behavior.

[0126] The first trigger difference array is a set of differences after comparing the actual and standard trigger element arrays. It includes the interaction sensing time difference of overlapping elements, the actual missing elements, and the actual redundant elements. It matches the array with the unique identifier of the element, calculates the time difference of overlapping elements, identifies missing and redundant elements, and integrates them into the array.

[0127] The interactive sensing time difference is the difference between the actual triggering time of the overlapping element in the actual and standard triggering element array and the scene preset standard triggering time. Specifically, it is obtained by extracting the actual and standard time points of the overlapping element and performing a subtraction operation.

[0128] Missing elements are interactive sensing elements that are present in the standard array but not in the actual array, while redundant elements are interactive sensing elements that are present in the actual array but not in the standard array. Both are identified by matching array elements. For example, if the standard array for a certain output behavior is [2, plus sign, 3] and the actual array is [2, 4], then the first trigger difference array is: {time difference: 50ms (touch element 2), missing element: [plus sign, 3], redundant element: [4]}.

[0129] In this embodiment, the distribution layer is a visual layer formed by mapping the positional distribution of the three types of components onto the touch screen component layout interface. Specifically, based on the touch screen component layout coordinates, the positions of the three types of components are mapped and marked differently.

[0130] The first, second, and third distributions represent the positional distributions of overlapping, missing, and redundant components on the touchscreen layout interface, respectively.

[0131] The first label uses three colors—red, yellow, and green—to indicate the dispersion of the interactive sensing time difference: red for a dispersion value ≥ 0.6, yellow for 0.3-0.6, and green for < 0.3.

[0132] The first anomaly sequence is a set of anomaly features corresponding to a single output behavior, consisting of a distribution layer and a first label. The implementation method is to convert the labeled distribution layer into a standardized feature data sequence.

[0133] In this embodiment, the actual change array is the set of elements remaining after removing elements that overlap with the actual trigger element array of the previous adjacent behavior, based on the actual trigger element array of the current output behavior. Specifically, the actual array of the current and previous adjacent behaviors is matched by the element's unique identifier, overlapping elements are removed and the remaining elements are integrated. For example, if the actual array of the first behavior is [1, plus sign] and the second behavior is [plus sign, 2, equal sign], then the actual change array of the second behavior is [2, equal sign].

[0134] In this embodiment, behavior association is the degree of association between two adjacent output behaviors in the teaching logic. It sets a quantitative association value for adjacent behaviors according to the teaching logic and stores it. The standard variation array is a set of variation elements that should exist for the current output behavior in the standard database based on behavior association. It matches and integrates from the standard database based on behavior association. The difference layer is a visual layer formed by distributing the positions of the difference elements on the touch screen layout interface after comparing the standard and actual variation arrays. It is generated to match the two arrays and identify the difference elements and map their layout coordinates.

[0135] In this embodiment, the key components are determined as follows:

[0136] Extract all discrepancy elements (elements that are inconsistent with the standard variation array) from the discrepancy layer as candidate key elements;

[0137] For candidate key components, check their abnormal status in the current output behavior distribution layer and the previous adjacent output behavior distribution layer. If there are abnormalities in both layers (such as time difference exceeding the threshold, accidental touch / missing), they are marked as persistent abnormal components.

[0138] For components with persistent abnormalities, calculate their abnormality impact degree, where abnormality impact degree = component trigger weight × number of abnormal occurrences / baseline number of occurrences. If the abnormality impact degree is ≥0.7, it is determined to be a critical component. The baseline number of occurrences is generally taken as 3.

[0139] The second annotation is a marker for key components on the current output behavior distribution layer, which uses a special color / symbol to mark the location of the key components.

[0140] The second anomaly sequence is the set of anomaly features corresponding to each output behavior starting from the second among multiple output behaviors. It consists of a distribution layer, a second annotation, and a difference layer, which transforms the relevant layers into a standardized feature data sequence.

[0141] It should be noted that when there are multiple output behaviors, the abnormal sequence of the first output behavior is determined in a similar way to the abnormal sequence when there is only one output behavior, which will not be elaborated here.

[0142] The beneficial effects of the above technical solution are as follows: Differentiated abnormal sequence generation logic is designed for children's single-output behavior and multi-output behavior respectively. Through operations such as comparison of standard and actual trigger element arrays, construction of distribution layers, analysis of change arrays and locking of key elements, abnormal features in children's teaching interaction data are accurately mined, standardized abnormal sequences are generated and integrated into change mapping results, providing targeted and scientific abnormal data support for subsequent key behavior judgment, contradiction array construction and other links.

[0143] This invention provides a children's educational interactive system for intelligent robots, wherein the interactive acquisition module further includes:

[0144] An expansion processing unit is used to expand the original anchor point of the output behavior in the baseline interaction thread of the preset teaching scenario according to the abnormal sequence of each output behavior, so as to obtain the range of expanded anchor points.

[0145] The contradiction analysis unit is used to construct a contradiction array for the corresponding output behavior by combining the deviation logic between the triggering time and the theoretical triggering range of the corresponding output behavior, the first behavior logic based on the original anchor point, and the second behavior logic based on the range of the extended anchor point.

[0146] The key judgment unit is used to determine whether the corresponding output behavior is a key behavior based on the number of contradiction arrays and the importance of contradictions in each contradiction array.

[0147] The enhanced processing unit is used to extract the valid state of the output behavior from its own valid state sequence if the corresponding output behavior is a critical behavior, and to perform a first enhanced processing on the output behavior in combination with the importance of the contradiction.

[0148] If the corresponding output behavior is not a critical behavior, a second enhanced processing is performed based on the current state of the output behavior.

[0149] The matrix construction unit is used to standardize the enhanced processing results of each output behavior according to several same-dimensional indicators, and construct a teaching interaction matrix according to the order of the output behaviors, where each output behavior corresponds to an interaction row vector.

[0150] In this embodiment, the baseline interaction thread is the standard interaction process thread when the child performs output behavior normally. It is designed with a standardized interaction process based on teaching logic and converted into a thread storage that the system can recognize.

[0151] The original anchor point is a key node in the benchmark interaction thread that defines the threshold for the execution of output behavior; that is, it is the node in the benchmark interaction thread that sets the quantization threshold.

[0152] Expanding the anchor point range is the threshold interval obtained by bidirectionally expanding the behavior threshold range of the original anchor point. It can cover the full-scene interaction features of the output behavior, that is, the original anchor point threshold is expanded in both positive and negative directions based on the expansion parameters.

[0153] In this embodiment, the deviation logic between the trigger time and the theoretical trigger range is to output the deviation between the actual trigger time of the behavior and the preset theoretical trigger range of the scene, including the degree and direction of deviation, that is, to extract the actual time and the theoretical range and analyze the magnitude and direction of the deviation.

[0154] The first behavioral logic is the standard interaction logic that the output behavior should follow in the baseline interaction thread based on the original anchor point, that is, extracting the standard interaction rules corresponding to the original anchor point from the baseline interaction thread; the second behavioral logic is the actual interaction logic that the output behavior actually follows in the actual interaction process based on the expanded anchor point range, that is, collecting actual interaction data and analyzing the actual interaction rules within the expanded anchor point range.

[0155] The contradiction array is an array formed after quantifying the multidimensional logical contradictions of the output behavior. It is divided into the first contradiction array (independent contradictions) and the second contradiction array (mutually dependent contradictions). That is, the first contradiction array is the set of quantified values ​​of independent contradictions in each dimension, and the second contradiction array is the set of synergistic quantified values ​​when there is a causal relationship between contradictions in each dimension.

[0156] In this embodiment, the formula for the first enhancement process is: ,in, This is data that has undergone enhanced processing for a single dimension with the same dimensions; The original interactive data is of the same dimension and size; w1 is the weight of the contradiction importance; S0 is the state comprehensive value.

[0157] In this embodiment, the formula for the second enhancement process is: ,in, For Gaussian kernel function, =0.5 and set based on the noise characteristics of children's interactive data; m0=1 is the size of the filter window; This represents the original interactive data of the same dimension under window i1.

[0158] In this embodiment, the enhanced processing results are normalized to extreme values ​​based on quantifiable indicators of the same dimension, such as trigger accuracy, trigger timeliness, and state effectiveness. The data is mapped to the [0,1] interval, and an interactive control matrix is ​​constructed according to the order of output behaviors, with each output behavior corresponding to an interactive row vector.

[0159] The beneficial effects of the above technical solution are as follows: by expanding the processing unit, full-scene coverage of children's interactive features is achieved; the contradiction analysis unit quantifies the logical contradictions in teaching interactions in multiple dimensions; the key judgment unit accurately focuses on the core abnormal interactive behaviors; the enhanced processing unit performs differentiated optimization processing on different types of behavioral data; and finally, the matrix construction unit generates a standardized teaching interaction matrix, so that subsequent vector extraction and sensitivity determination of core sensing elements have a unified and standardized data foundation.

[0160] This invention provides a children's educational interactive system for intelligent robots, wherein the extended processing unit includes:

[0161] The parameter determination subunit is used to extract the coefficient of variation of the interaction sensing time difference, the trigger weight of key components, and the component distribution difference in the abnormal sequence of each output behavior as extended parameters.

[0162] The bidirectional expansion subunit is used to bidirectionally expand the behavior threshold range corresponding to the original anchor point based on the expansion parameters, so as to obtain an expanded anchor point range that covers the full-scene interaction features of the output behavior. The forward expansion amplitude is related to the coefficient of variation of the interaction sensing time difference, and the reverse expansion amplitude is related to the trigger weight of the key element and the difference between 1 and the element distribution difference.

[0163] In this embodiment, the coefficient of variation = the sample standard deviation of the interactive sensing time difference under a single output behavior / the sample arithmetic mean of the interactive sensing time difference under a single output behavior.

[0164] The trigger weight of a key element is a quantitative value of how important the key element is to be triggered in the output behavior. It sets a weight value of 0-1 for the key element according to the teaching logic, with core teaching function elements having a higher weight.

[0165] In this embodiment, the formula for calculating the component distribution difference is: Where D is the component distribution difference degree, with a value range of 0 to 1. If D is greater than 1, it is truncated with 1 as the upper limit, and m is the number of overlapping components. The touchscreen layout coordinates are for the p-th overlapping element in the actual trigger element array; Let be the touchscreen layout coordinates of the p-th overlapping element in the standard trigger element array; The maximum feature length of the touchscreen component layout area.

[0166] In this embodiment, the positive expansion amplitude = the coefficient of variation of the interactive sensing time difference × the preset base amplitude;

[0167] Reverse expansion amplitude = trigger weight of key component × (1 - component distribution difference) × preset base amplitude, where the preset base amplitude is pre-set and has a value of 30ms.

[0168] The behavior threshold range is the threshold interval for the normal execution of the output behavior defined by the original anchor point, such as the trigger time, the number of triggers of components, etc. That is, setting upper and lower thresholds at the original anchor point. For example, if the original anchor point trigger time threshold is 50ms, the positive expansion range is 20ms, and the negative expansion range is 15ms, then the expanded anchor point range is 35ms-70ms.

[0169] The beneficial effects of the above technical solution are: extracting multi-dimensional core parameters from the abnormal sequence of output behavior as the basis for anchor point expansion, so that the expansion of the original anchor point has solid data support and avoids the problem of blind expansion. At the same time, bidirectional expansion is carried out according to the correlation between each parameter and the expansion range, realizing accurate and adaptive adjustment of the threshold range of the original anchor point behavior. It can comprehensively cover various interactive features of children in the preset teaching scenario, and provide a threshold reference that is more in line with the actual interaction situation for subsequent behavioral logic analysis and contradiction array construction.

[0170] This invention provides a children's educational interactive system for intelligent robots, wherein the contradiction analysis unit includes:

[0171] The first quantization subunit is used to quantify the deviation logic between the trigger time of the corresponding output behavior and the theoretical trigger range under the preset teaching scenario, and to obtain the deviation coefficient and deviation direction parameter.

[0172] The second quantization subunit is used to parse the standard behavior parameters corresponding to the first behavior logic based on the original anchor point, and the actual behavior parameters corresponding to the second behavior logic based on the extended anchor point range, and to quantify the degree of difference between the first behavior logic and the second behavior logic.

[0173] The array construction sub-unit is used to construct a first contradiction array and a second contradiction array where contradictions act independently under the corresponding output behavior, with the deviation logic parameters, the first row being the logic standard parameters, the second row being the logic actual parameters, and the degree of logic difference as the array dimensions.

[0174] The deviation direction parameter is a parameter that marks the deviation direction. It is -1 if it is ahead, +1 if it is behind, and 0 if it is within the range. That is, it determines the actual position at the moment and assigns the corresponding value.

[0175] In this embodiment, the deviation coefficient Where t is the actual triggering time; t0 is the center value of the theoretical triggering range; These represent the upper and lower limits of the theoretical trigger range, respectively; α=0 indicates no deviation; α=1 indicates extreme deviation. When there is no center value, use... replace.

[0176] In this embodiment, the degree of difference Where k0 is the number of quantization parameters for the behavioral logic; This is the s-th actual parameter value of the second line of logic; This is the s-th standard parameter value of the first line of logic.

[0177] Standard behavioral parameters are standardized and quantifiable parameters corresponding to behavioral logic, such as the number of triggering elements and the triggering time interval. In other words, the corresponding behavioral logic is converted into quantifiable parameter values.

[0178] Deviation logic parameters are a set of deviation coefficients and deviation direction parameters, that is, integrating the two parameters into one-dimensional feature parameters. The first and second contradiction arrays are formed by arranging the parameters of each dimension according to the rules of contradiction independence and interdependence.

[0179] In this embodiment, the independent action of contradictions means that there is no causal relationship between the logical contradictions of each dimension of the output behavior, and the generation and development of any contradiction does not affect the existence of other contradictions. Each contradiction independently affects the interactive behavior. The interdependence of contradictions means that there is a direct causal relationship between the logical contradictions of each dimension of the output behavior. One contradiction is the cause and other contradictions are the effects. The causal contradictions influence each other and work together to affect the interactive behavior.

[0180] The beneficial effects of the above technical solution are as follows: by using two quantification sub-units, the precise quantification of the deviation from logic at the trigger time of output behavior and the difference in behavioral logic under the original and extended anchor points is achieved. Then, by using multi-dimensional quantification indicators as array dimensions, two types of contradiction arrays are constructed, which respectively reflect the independent role and the interdependent role of contradictions. This realizes the systematic and refined representation of multi-dimensional logical contradictions in children's teaching interaction, making the contradiction analysis more hierarchical and providing comprehensive and detailed contradiction data references for subsequent key behavior judgment and sensitivity analysis of core sensing elements.

[0181] This invention provides a children's educational interactive system for intelligent robots, wherein the sensitive judgment module includes:

[0182] The cluster correlation value determination unit is used to quantify the inter-cluster distance using Euclidean distance and to determine the density of key behaviors within a cluster based on the ratio of the number of key behaviors within a cluster to the total number of behaviors within a cluster.

[0183] The importance determination unit is used to extract the first contradiction array and the second contradiction array for each output behavior under all first vectors, and to quantify the contradiction importance of each contradiction array and assign weights using the analytic hierarchy process.

[0184] The normalization unit is used to normalize the inter-cluster distance, the density of key behaviors within a cluster, and the weight of the importance of contradictions.

[0185] The size comparison unit is used to determine that the corresponding core sensing element is insensitive when the normalized inter-cluster distance exceeds the first threshold, the density of key behaviors within the cluster is lower than the second threshold, or the sum of the weights of contradictory importance exceeds the third threshold; otherwise, it determines that the corresponding core sensing element is sensitive.

[0186] In this embodiment, the inter-cluster distance is calculated using Euclidean distance, and the formula is as follows: ,in, Let be the inter-cluster distance between the i-th interactive feature cluster and the j-th interactive feature cluster; Let be the feature center value of the k-th dimension in the i-th interactive feature cluster; is the feature center value of the k-th dimension in the j-th interactive feature cluster; n is the dimension of the first vector, i.e., the number of indicators in the same dimension; the feature center value is the average value of the corresponding dimension of all the first vectors in the cluster.

[0187] In this embodiment, extreme value normalization is applied to the three types of indicators to map the original values ​​to the [0,1] interval, so that indicators with different dimensions and different numerical ranges are under the same measurement standard. For example, the original value of the inter-cluster distance is 5.2, the minimum value of the indicator is 0, and the maximum value is 10, which is normalized to 0.52; the original value of the density of key behaviors within a cluster is 0.625, which is normalized to remain at 0.625; and the original value of the importance weight of contradiction is 0.3, which is normalized to remain at 0.3.

[0188] Based on extensive experimental data from children's interactive teaching scenarios, reasonable basic threshold values ​​were set: a first threshold of 0.8, a second threshold of 0.5, and a third threshold of 1.2, serving as the standard for sensitivity assessment. For different types of teaching scenarios, the basic thresholds were fine-tuned by ±0.1, as shown in Table 2.

[0189] Table 2 Contextualized Threshold Adjustment Table

[0190]

[0191] The beneficial effects of the above technical solution are as follows: the Euclidean distance and the density of key behaviors within a cluster are accurately quantified by using the Euclidean distance and the ratio of quantities; the analytic hierarchy process is used to complete the scientific quantification and weight allocation of the importance of contradictions; the normalization process effectively eliminates the dimensional differences between different indicators; and finally, by combining preset multiple thresholds for multi-condition comprehensive judgment, a multi-dimensional, objective and accurate judgment of the sensitivity of core sensing elements is achieved, breaking through the limitations of traditional single-indicator detection of element status, and enabling timely and accurate detection of element operation abnormalities.

[0192] This invention provides a platform for interactive teaching for children using intelligent robots, including a processor and a memory. The memory stores a computer program, and the processor executes the computer program to implement the interactive method of any of the systems.

[0193] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations. A children's educational interactive system and platform for intelligent robots.

Claims

1. A children's educational interactive system for intelligent robots, characterized in that, include: An interactive acquisition module is used to acquire the element trigger array sequence and the child's own valid state sequence in a preset teaching scenario. It also combines the change mapping result of the element trigger array sequence based on the preset teaching scenario and the key behaviors determined based on the change mapping result to construct an interactive control matrix. In each preset teaching scenario, the child has at least one output behavior. The element trigger array sequence is implemented based on interactive sensing elements. The vector extraction module is used to determine the response component of the intelligent robot for each output behavior based on the interaction control lookup table between the child and the intelligent robot, and to extract each row vector under the same response component from the interaction control matrix as the first vector; The sensitivity judgment module is used to perform hierarchical clustering on all first vectors under the same response component to obtain interactive feature clusters. Based on the distance between clusters, the density of key behaviors within clusters, and the importance of contradictions in the contradiction array of each output behavior under all first vectors with multidimensional logical contradictions, the module combines a preset scenario-based threshold to determine whether the core sensing element of the corresponding response component is sensitive. Each response component is controlled by at least one interactive sensing element, and the core sensing element is the element whose trigger frequency ratio among the interactive sensing elements controlling the same response component is greater than or equal to the preset frequency ratio, is the primary trigger element for the activation of the response component function, and has the highest correlation with the core teaching function of the teaching scenario. If the response is insensitive, the adjustment type of the corresponding core sensing element is determined based on the difference vector generated by cluster sorting; otherwise, the corresponding core sensing element is determined to remain unchanged. The interactive control module is used to collect new interactive data of children in a preset teaching scenario in real time based on all the interactive sensing elements that have been adjusted or remain unchanged, construct a real-time interactive matrix, compare the real-time interactive matrix with the historical standard matrix, and generate multimodal output data by combining the teaching language and the priority of key behaviors, and control the intelligent robot to perform teaching feedback. Among them, the element trigger array sequence refers to the sequence set obtained by sorting multiple interactive sensing element arrays triggered by the child's output behavior in the preset teaching scenario according to the time sequence of the output behavior, with each output behavior corresponding to an independent interactive sensing element array. The self-valid state sequence refers to the set of state sequences obtained by sorting the self-valid states of a child in a preset teaching scenario when each output behavior is executed in sequence according to the time order of the output behavior. Each output behavior corresponds to a self-valid state. The change mapping result refers to the set of abnormal sequences of all output behaviors obtained after performing anomaly analysis on the element trigger array of all output behaviors of children in a preset teaching scenario; The interaction control matrix refers to a two-dimensional matrix constructed by standardizing the enhanced processing results of all output behaviors of children in a preset teaching scenario according to several quantifiable interaction indicators of the same dimension, with the order of output behaviors as rows and the same-dimensional indicators as columns. Each output behavior corresponds to an interaction row vector in the matrix.

2. The interactive teaching system for children using intelligent robots according to claim 1, characterized in that, The interactive data acquisition module includes: The first capture unit is used to capture the interactive sensing element triggered by each output behavior to obtain the actual triggering element array, wherein the triggering time point is marked for each interactive sensing element in the actual triggering element array. The second capturing unit is used to capture the child's current state when performing the corresponding output behavior in the preset teaching scenario and the child's historical state from the start of the preset teaching scenario to the execution of the corresponding output behavior, and to determine the valid state of the corresponding output behavior based on the current state and the historical state. The sorting unit is used to sort the actual triggering element array of each output behavior according to the order of all output behaviors to obtain the element triggering array sequence, and to sort the valid states of each output behavior to obtain its own valid state sequence.

3. The interactive teaching system for children using intelligent robots according to claim 2, characterized in that, The interactive data acquisition module also includes: The first comparison unit is used to sequentially match the standard trigger element array that matches each output behavior from the standard database of the preset teaching scenario, and compare it with the corresponding actual trigger element array to obtain the first trigger difference array of the corresponding output behavior. The difference array includes: the interaction sensing time difference of each overlapping element in the corresponding actual trigger element array, the missing elements in the corresponding actual trigger element array, and the redundant elements. A layer construction unit is used to determine the first distribution of overlapping elements, the second distribution of missing elements, and the third distribution of redundant elements in the first trigger difference array when there is only one output behavior of the child in the preset teaching scenario, to obtain a distribution layer and a first abnormal sequence of the corresponding output behavior. The first distribution area of ​​the distribution layer is marked according to the normal distribution of the interaction sensing time difference to obtain the first abnormal sequence of the corresponding output behavior. The variable array determination unit is used to, when the child has multiple output behaviors in the preset teaching scenario, start from the second output behavior, and based on the actual trigger element array of the corresponding output behavior, remove the elements that overlap with the actual trigger element array of the previous adjacent behavior, to obtain the actual variable array of each output behavior and the previous adjacent behavior based on the actual trigger element array. The difference sequence determination unit is used to determine a standard variation array according to the behavioral association between the corresponding output behavior and the previous adjacent behavior, and compare it with the actual variation array to obtain a difference layer. Combining the distribution layer of the corresponding output behavior with the distribution layer of the previous adjacent behavior, the unit locks the key elements and makes a second annotation on the distribution layer of the corresponding output behavior to obtain the second abnormal sequence of the corresponding output behavior. Among them, the abnormal sequences of all output behaviors constitute the change mapping result.

4. The interactive teaching system for children using intelligent robots according to claim 3, characterized in that, The interactive data acquisition module also includes: An expansion processing unit is used to expand the original anchor point of the output behavior in the baseline interaction thread of the preset teaching scenario according to the abnormal sequence of each output behavior, so as to obtain the range of expanded anchor points. The contradiction analysis unit is used to construct a contradiction array for the corresponding output behavior by combining the deviation logic between the triggering time and the theoretical triggering range of the corresponding output behavior, the first behavior logic based on the original anchor point, and the second behavior logic based on the range of the extended anchor point. The key judgment unit is used to determine whether the corresponding output behavior is a key behavior based on the number of contradiction arrays and the importance of contradictions in each contradiction array. The enhanced processing unit is used to extract the valid state of the output behavior from its own valid state sequence if the corresponding output behavior is a critical behavior, and to perform a first enhanced processing on the output behavior in combination with the importance of the contradiction. If the corresponding output behavior is not a critical behavior, a second enhanced processing is performed based on the current state of the output behavior. The matrix construction unit is used to standardize the enhanced processing results of each output behavior according to several same-dimensional indicators, and construct a teaching interaction matrix according to the order of the output behaviors, where each output behavior corresponds to an interaction row vector.

5. The interactive teaching system for children using intelligent robots according to claim 4, characterized in that, The expansion processing unit includes: The parameter determination subunit is used to extract the normal distribution discrete value of the interaction sensing time difference, the trigger weight of key components, and the component distribution difference degree in the abnormal sequence of each output behavior as extended parameters. The bidirectional expansion subunit is used to bidirectionally expand the behavior threshold range corresponding to the original anchor point based on the expansion parameters, so as to obtain an expanded anchor point range that covers the full-scene interaction features of the output behavior. The positive expansion amplitude is positively correlated with the dispersion of the interaction sensing time difference, and the negative expansion amplitude is positively correlated with the trigger weight of the key element and the difference between 1 and the element distribution difference. The element distribution difference ranges from 0 to 1.

6. The interactive teaching system for children using intelligent robots according to claim 4, characterized in that, The contradiction analysis unit includes: The first quantization subunit is used to quantify the deviation logic between the trigger time of the corresponding output behavior and the theoretical trigger range under the preset teaching scenario, and to obtain the deviation coefficient and deviation direction parameter, wherein the deviation direction parameter is -1 for early, +1 for late, and 0 for within the range; The second quantization subunit is used to parse the standard behavior parameters corresponding to the first behavior logic based on the original anchor point, and the actual behavior parameters corresponding to the second behavior logic based on the extended anchor point range, and to quantify the degree of difference between the first behavior logic and the second behavior logic. The array construction sub-unit is used to construct a first contradiction array and a second contradiction array where contradictions act independently under the corresponding output behavior, with the deviation logic parameters, the first row being the logic standard parameters, the second row being the logic actual parameters, and the degree of logic difference as the array dimensions.

7. The interactive teaching system for children using intelligent robots according to claim 1, characterized in that, The sensitive judgment module includes: The cluster correlation value determination unit is used to quantify the inter-cluster distance using Euclidean distance and to determine the density of key behaviors within a cluster based on the ratio of the number of key behaviors within a cluster to the total number of behaviors within a cluster. The importance determination unit is used to extract the first contradiction array and the second contradiction array for each output behavior under all first vectors, and to quantify the contradiction importance of each contradiction array and assign weights using the analytic hierarchy process. The normalization unit is used to normalize the inter-cluster distance, the density of key behaviors within a cluster, and the weight of the importance of contradictions. The size comparison unit is used to determine that the corresponding core sensing element is insensitive when the normalized inter-cluster distance exceeds the first threshold, the density of key behaviors within the cluster is lower than the second threshold, or the sum of the weights of contradictory importance exceeds the third threshold; otherwise, it determines that the corresponding core sensing element is sensitive.

8. The interactive teaching system for children using intelligent robots according to claim 1, characterized in that, The adjustment types include: program error repair type, program upgrade type, and component replacement type.

9. A platform for interactive teaching with intelligent robots for children, characterized in that, The system includes a processor and a memory, the memory storing a computer program. When the processor executes the computer program, it implements the interaction method of any one of claims 1-8, the method comprising: Step 1: Collect the element trigger array sequence and the child's own valid state sequence in the preset teaching scenario, and combine the change mapping result of the element trigger array sequence based on the preset teaching scenario and the key behaviors determined based on the change mapping result to construct an interactive control matrix. In each preset teaching scenario, the child has at least one output behavior, and the element trigger array sequence is implemented based on interactive sensing elements. Step 2: Based on the interaction control reference table between children and intelligent robots, determine the response component of the intelligent robot to pre-respond to each output behavior, and extract each row vector under the same response component from the interaction control matrix as the first vector; Step 3: Perform hierarchical clustering on all first vectors under the same response component to obtain interactive feature clusters. Based on the distance between clusters, the density of key behaviors within clusters, and the importance of contradictions in the contradiction array of each output behavior under all first vectors with multidimensional logical contradictions, determine whether the core sensing element of the corresponding response component is sensitive by combining a preset scenario threshold. Each response component is controlled by at least one interactive sensing element, and the core sensing element is the element with the highest degree of correlation with the core teaching function of the teaching scenario, which is the primary triggering element for starting the function of the response component among the interactive sensing elements controlling the same response component with a trigger frequency ratio greater than or equal to a preset frequency ratio. If the response is insensitive, the adjustment type of the corresponding core sensing element is determined based on the difference vector generated by cluster sorting; otherwise, the corresponding core sensing element is determined to remain unchanged. Step 4: Based on all the adjusted or unchanged interactive sensing elements, collect new interactive data of children in the preset teaching scenario in real time to construct a real-time interaction matrix, compare the real-time interaction matrix with the historical standard matrix, and generate multimodal output data by combining the teaching language and key behavior priority to control the intelligent robot to perform teaching feedback. Among them, the element trigger array sequence refers to the sequence set obtained by sorting multiple interactive sensing element arrays triggered by the child's output behavior in the preset teaching scenario according to the time sequence of the output behavior, with each output behavior corresponding to an independent interactive sensing element array. The self-valid state sequence refers to the set of state sequences obtained by sorting the self-valid states of a child in a preset teaching scenario when each output behavior is executed in sequence according to the time order of the output behavior. Each output behavior corresponds to a self-valid state. The change mapping result refers to the set of abnormal sequences of all output behaviors obtained after performing anomaly analysis on the element trigger array of all output behaviors of children in a preset teaching scenario; The interaction control matrix refers to a two-dimensional matrix constructed by standardizing the enhanced processing results of all output behaviors of children in a preset teaching scenario according to several quantifiable interaction indicators of the same dimension, with the order of output behaviors as rows and the same-dimensional indicators as columns. Each output behavior corresponds to an interaction row vector in the matrix.

Images