An interactive control method, system, device and medium of a robot

Abstract

The application relates to an interactive control method, system, device and medium of a robot. The method comprises the following steps: acquiring a physical environment signal and a user interaction signal; performing standardization processing on the physical environment signal and the user interaction signal to obtain a multi-modal scene data package; identifying a current emotional state of a user based on the user interaction signal in the multi-modal scene data package to obtain an emotional type judgment result; calculating a scene parameter based on the multi-modal scene data package; calculating a core state parameter of the robot based on the scene parameter and the emotional type judgment result to obtain a double-axis parameter value; and matching the double-axis parameter value and a personality type of the robot with a preset emotional behavior to obtain a behavior instruction set. The method can accurately locate discrete emotions, so that the emotional feedback of the robot has the delicate feeling of a real living being, the user emotions are perceived, and accurate emotional accompaniment is achieved.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence interaction, and in particular relates to an interactive control method, system, device and medium for robots. Background Technology

[0002] With the in-depth development of artificial intelligence interaction technology, AI pet interaction technology has emerged, focusing on simulating biological emotions and companionship behaviors. This technology aims to provide users with a realistic digital companion experience by simulating emotional responses and personality traits. Its core characteristics are the pursuit of diverse emotional expressions and unique personality development, leading to the emergence of traditional interaction methods based on single-valued driving forces and static personality models. In traditional technologies, emotional processing typically relies on linear correlations using single-dimensional values ​​such as affection or intimacy, simply mapping the pet's emotional state to a limited range of categories like happy or unhappy. Personality processing often uses factory-preset fixed type labels, such as lively or quiet, which remain constant throughout the pet's lifespan. The interaction logic is designed as a one-way output mode, meaning the system determines the pet's behavior based on its internal state, unable to perceive or adapt to changes in the user's emotions. However, current traditional interaction methods have significant limitations. The singular dimension of emotion driving makes it impossible to express complex emotions such as fatigue yet comfort, resulting in flat emotional feedback lacking the subtlety of a real animal. The static personality development logic leads to highly similar behavioral patterns across all similar products, making it difficult to create a sense of novelty and user stickiness for long-term companionship. More importantly, the lack of user emotional awareness means that pets may still be active when users are in a low mood, creating a strong sense of incongruity in the interaction and thus weakening the core value of emotional companionship. Summary of the Invention

[0003] Therefore, it is necessary to provide an interactive control method, system, device, and medium for robots that can accurately locate discrete emotions, thereby enabling robots to provide emotional feedback with the subtlety of real living beings, perceive user emotions, and achieve emotional companionship, in order to address the aforementioned technical problems.

[0004] In a first aspect, this application provides an interactive control method for a robot, comprising:

[0005] Acquire physical environment signals and user interaction signals; and standardize the physical environment signals and user interaction signals to obtain multimodal scene data packets;

[0006] Based on user interaction signals in the multimodal scene data package, the user's current emotional state is identified to obtain the emotion type determination result; and scene parameters are calculated based on the multimodal scene data package; the user interaction signals include voice signals and touch signals to the robot;

[0007] Based on the scene parameters and emotion type determination results, the robot's core state parameters are calculated to obtain dual-axis parameter values; the dual-axis parameter values ​​include comfort and liveliness values.

[0008] The dual-axis parameter values ​​and the robot's personality type are matched with preset emotional behaviors to obtain a set of behavioral instructions; the set of behavioral instructions is used to instruct the robot to perform actions.

[0009] Furthermore, the robot's personality type is updated through the following methods:

[0010] Based on a fixed time window, user interaction signals are acquired; if the cumulative duration of user interaction signals exceeds a preset threshold, a personality adjustment identifier is generated; the personality adjustment identifier indicates that a personality type update is required.

[0011] Based on personality adjustment identifiers and user interaction signals, the adjustment amount of personality parameters is calculated through interaction ratio analysis.

[0012] The personality parameters are updated based on the adjustment amount of the personality parameters to obtain the updated personality parameters; then the updated personality parameters are matched with the four temperament personality judgment threshold table to obtain the updated personality type.

[0013] Furthermore, based on personality adjustment identifiers and user interaction signals, the adjustment amount of personality parameters is calculated through interaction ratio analysis, including:

[0014] Based on personality adjustment identifiers, voice and touch signals are extracted from user interaction signals to obtain a detailed list of interaction records;

[0015] Based on a predefined interactive action rule library, each interactive action in the interaction record details is judged for both intimacy value interaction and extroversion value interaction, resulting in a dual interaction record list.

[0016] Based on the dual interaction record list, calculate the intimacy value adjustment amount; and based on the dual interaction record list, calculate the extroversion value adjustment amount;

[0017] By integrating the intimacy value adjustment and the extroversion value adjustment, the personality parameter adjustment is obtained.

[0018] Furthermore, based on the dual interaction record list, the intimacy value adjustment amount is calculated, including:

[0019] Iterate through the list of dual interaction records, count the number of interaction actions marked as positive for intimacy value, and get the positive count; and count the number of interaction actions marked as negative for intimacy value, and get the negative count;

[0020] Based on the number of positive and negative interactions, the intimate interaction ratio is calculated using the following formula:

[0021]

[0022] Where R is the ratio of intimate interactions, P is the number of positive interactions, and N is the number of negative interactions;

[0023] Based on the intimacy interaction ratio, the basic adjustment range value is obtained by consulting the personality adjustment level table;

[0024] Based on personality type, the corresponding intimacy value adjustment rule is invoked to adjust the base adjustment value and obtain the intimacy value adjustment amount.

[0025] Furthermore, based on the scene parameters and emotion type determination results, the robot's core state parameters are calculated to obtain dual-axis parameter values, including:

[0026] Based on scene parameters and personality type, calculate basic dual-axis parameter values; scene parameters include time scene, interaction scene, environmental scene and device scene;

[0027] If the emotion type determination result matches the preset forced overriding emotion, then based on the emotion type determination result, the corresponding fixed dual-axis parameter value is called to overwrite the basic dual-axis parameter value to obtain the dual-axis parameter value;

[0028] If the emotion type determination result does not match the forced overriding emotion, then the basic dual-axis parameter value will be determined as the dual-axis parameter value.

[0029] Furthermore, based on user interaction signals in the multimodal scene data packet, the user's current emotional state is identified to obtain the emotion type determination result, including:

[0030] Extract voice and touch signals from user interaction signals; and preprocess the voice and touch signals to obtain a voice data frame sequence and a touch event sequence.

[0031] The speech data frame sequence is subjected to feature extraction to obtain a speech feature vector; the speech feature vector is then input into a speech emotion classifier to obtain a speech emotion determination result; the speech feature vector includes at least one of fundamental frequency, energy, speech rate and spectral features.

[0032] Feature extraction is performed on the touch event sequence to obtain touch feature vectors; and the touch feature vectors are input into a touch behavior classifier to obtain the touch emotion judgment result.

[0033] Based on the voice emotion determination results and the touch emotion determination results, an emotion type determination result is generated.

[0034] Furthermore, the dual-axis parameter values ​​and the robot's personality type are matched with preset emotional behaviors to obtain a set of behavioral instructions, including:

[0035] Based on the dual-axis parameter values, the basic emotional state is obtained by querying the unique discrete emotion label in the emotion mapping table.

[0036] Using personality type and basic emotional state as the joint key, the corresponding basic behavioral templates are retrieved from the behavioral rule base;

[0037] Based on the emotion type determination results, the amplitude and frequency of the basic behavior template are fine-tuned to obtain the fine-tuned behavior;

[0038] By mapping fine-tuning behaviors to the corresponding robot execution parameters, a set of behavior instructions is obtained.

[0039] Secondly, this application also provides an interactive control system for a robot, comprising:

[0040] The standardization module is used to acquire physical environment signals and user interaction signals; and to standardize the physical environment signals and user interaction signals to obtain multimodal scene data packets.

[0041] The scene emotion module is used to identify the user's current emotional state based on user interaction signals in the multimodal scene data package and obtain the emotion type determination result; and to calculate scene parameters based on the multimodal scene data package; the user interaction signals include voice signals and touch signals to the robot;

[0042] The dual-axis parameter module is used to calculate the robot's core state parameters based on scene parameters and emotion type determination results, and obtain dual-axis parameter values; the dual-axis parameter values ​​include comfort and liveliness values.

[0043] The behavior instruction module is used to match the dual-axis parameter values ​​and the robot's personality type with preset emotional behaviors to obtain a behavior instruction set; the behavior instruction set is used to instruct the robot to perform actions.

[0044] Thirdly, this application also provides a computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any step of the method provided in the first aspect of this application.

[0045] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of the method provided in the first aspect of this application.

[0046] The aforementioned robot interaction control method, system, device, and medium acquire physical environment signals and user interaction signals; standardize the physical environment signals and user interaction signals to obtain a multimodal scene data package; based on the user interaction signals in the multimodal scene data package, identify the user's current emotional state to obtain an emotion type determination result; and calculate scene parameters based on the multimodal scene data package; the user interaction signals include voice signals and touch signals to the robot; based on the scene parameters and the emotion type determination result, calculate the robot's core state parameters to obtain dual-axis parameter values; the dual-axis parameter values ​​include comfort and liveliness values; match the dual-axis parameter values ​​and the robot's personality type with preset emotional behaviors to obtain a behavior instruction set; the behavior instruction set is used to instruct the robot to perform actions, and through the dual-axis emotion engine, overcome the limitations of single-value-driven emotions, achieving accurate positioning and dynamic switching of multiple discrete and complex emotions, covering the emotional expression dimensions of real pets; and perform behavior control based on user emotion recognition to achieve two-way emotional empathy and solve the problem of incongruity in one-way emotion output. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram of the flow of an interactive control method for a robot according to an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the structure of an interactive control system for a robot provided in an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0051] In one embodiment, such as Figure 1 As shown, an interactive control method for a robot is provided. This embodiment illustrates the method applied to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0052] Step 101: Acquire physical environment signals and user interaction signals; and standardize the physical environment signals and user interaction signals to obtain multimodal scene data packets.

[0053] Among them, physical environment signals refer to objective environmental data collected by sensors, including information reflecting the robot's physical state such as temperature, light intensity, ambient noise levels (decibels), and device battery percentage, used to determine the impact of the external environment on the robot's basic state. User interaction signals refer to the active behavioral data generated when users interact with the robot, mainly including voice and touch signals, which are direct evidence for determining user intentions and emotions. The multimodal scene data package is a standardized, structured data set that integrates physical environment signals and user interaction signals, and adds metadata such as timestamps and signal quality identifiers. It serves as a unified input source for all subsequent analyses, ensuring consistency in data processing across different modules. The terminal synchronously collects multi-channel signals through a hardware sensor network. Physical environment signals are acquired in real time by devices such as temperature sensors, noise sensors, and power monitoring chips. User interaction signals are collected through a microphone array for voice and through a piezoelectric thin film sensor grid covering the robot's surface for touch data. These signals are then standardized, and the physical environment signals are subjected to dimensional unification and outlier filtering. The user interaction signals are then subjected to noise reduction, normalization, and time alignment. Specifically, the voice signals undergo pre-emphasis and frame segmentation, while the touch signals undergo force calibration and touch point clustering. Finally, all signals are packaged according to time windows to generate a multimodal scene data package.

[0054] Step 102: Based on the user interaction signals in the multimodal scene data package, identify the user's current emotional state and obtain the emotion type determination result; and calculate the scene parameters based on the multimodal scene data package; the user interaction signals include voice signals and touch signals to the robot.

[0055] Specifically, the emotion type determination result is a classification conclusion of the user's current emotional state, which can be one of three discrete labels: joy, anger, and sorrow. It may include a confidence score and directly reflects the user's emotional state. Scene parameters are a quantitative description of the current interaction context, including time scene, interaction scene, environmental scene, and device scene, providing contextual basis for the robot's state adjustment. The terminal separates the user interaction signal from the multimodal data packet and processes speech and touch information in parallel. For speech signals, acoustic features such as fundamental frequency contour, energy envelope, speech rate, and spectral centroid are extracted and input into a pre-trained speech emotion classifier for pattern matching. For touch signals, features such as the mean pressure, interval variance, and duration of the touch sequence are analyzed. The emotional tendency is determined by a touch behavior classifier. A decision-level fusion strategy is adopted to combine the determination results of the two modalities to generate the final emotion type determination result.

[0056] Step 103: Based on the scene parameters and the emotion type determination results, calculate the robot's core state parameters to obtain dual-axis parameter values; the dual-axis parameter values ​​include comfort and liveliness values.

[0057] Specifically, the dual-axis parameter values ​​are two dimensions describing the robot's core state: comfort, reflecting its physical comfort in the current environment and interactions, and liveliness, reflecting its energy level and willingness to be active. Both values ​​range from 0 to 100. These two parameters jointly determine the emotional state of the AI ​​(Artificial Intelligence) pet robot. The terminal performs basic dual-axis parameter value calculations. Based on scene parameters and the robot's personality type, a weighted calculation is performed to obtain the base value. The system then checks whether the emotion type determination result is a mandatory overriding emotion requiring priority response. If so, a preset fixed parameter value is used to overwrite the base calculated value; otherwise, the base value is directly used as the final dual-axis parameter value. The entire calculation process must follow priority rules, with user emotion intervention having the highest priority to ensure timely response to user emotions.

[0058] Step 104: Match the dual-axis parameter values ​​and the robot's personality type with the preset emotional behaviors to obtain a set of behavioral instructions; the set of behavioral instructions is used to instruct the robot to perform actions.

[0059] Specifically, the behavior instruction set is a collection of specific commands controlling the robot's actuators, detailing parameters such as action type, amplitude, duration, sound frequency, light color, and brightness. It serves as the final instruction driving the robot to produce external behaviors. Personality type is a category label describing the robot's stable behavioral tendencies, such as choleric or sanguine, influencing the selection of different behavior templates under the same conditions. Preset emotional behaviors are a set of mapping relationships stored in the behavior rule base. They define recommended behavior templates corresponding to different combinations of dual-axis parameter values ​​and personality types, as well as adjustable parameter ranges. The terminal performs emotion mapping, mapping continuous dual-axis parameter values ​​to a discrete basic emotional state label by querying the emotion mapping table. Using personality type and basic emotional state as a joint key, the most matching basic behavior template is retrieved from the behavior rule base. Behavior fine-tuning is then performed, adjusting dynamic parameters such as amplitude and frequency of the basic behavior template based on the emotion type determination result. The fine-tuned behavior description is then transformed into executable low-level control instructions specific to the robot's hardware platform, forming the final behavior instruction set.

[0060] This embodiment provides an interactive control method for a robot, which acquires physical environment signals and user interaction signals; standardizes these signals to obtain a multimodal scene data package; identifies the user's current emotional state based on the user interaction signals in the multimodal scene data package, obtaining an emotion type determination result; and calculates scene parameters based on the multimodal scene data package. The user interaction signals include voice signals and touch signals to the robot. Based on the scene parameters and the emotion type determination result, the robot's core state parameters are calculated to obtain dual-axis parameter values, including comfort and liveliness values. The dual-axis parameter values ​​and the robot's personality type are matched with preset emotional behaviors to obtain a behavior instruction set. This behavior instruction set is used to instruct the robot to perform actions. Through these methods, a dual-axis emotion engine is introduced to overcome the limitations of single-value-driven emotions, achieving precise positioning and dynamic switching of multiple discrete and complex emotions, covering the emotional expression dimensions of real pets. Behavior control based on user emotion recognition enables two-way emotional empathy, resolving the incongruity of one-way emotion output.

[0061] In one embodiment, the robot's personality type is updated using the following method:

[0062] Step 201: Based on a fixed time window, acquire user interaction signals; if the cumulative duration of user interaction signals exceeds a preset threshold, generate a personality adjustment identifier; the personality adjustment identifier indicates that a personality type update is required.

[0063] The fixed time window refers to a fixed-length time period set by the system, which can be an hour or a day as an independent statistical unit. The time window is used to divide and limit the statistical range of user interaction data, ensuring that personality assessments are based on data at the same time scale and avoiding analytical biases caused by different time periods. User interaction signals are behavioral data generated when users interact with the robot, mainly including voice signals and touch signals, and are the basic raw data for analyzing user interaction patterns. Cumulative duration is the sum of the durations of all valid user interaction behaviors within the fixed time window; only interactions that meet preset conditions are included in the duration statistics. The preset threshold is a pre-set time threshold used to determine whether user interaction is sufficient to trigger personality adjustment, ensuring that the update process is initiated only when a certain interaction intensity is reached. The personality adjustment identifier is a Boolean value or flag; when set to true, it indicates that the personality adjustment conditions have been detected, and the subsequent personality parameter update process needs to be executed. It is an internal status signal and is not directly presented to the user. The terminal periodically collects user interaction signals according to a preset fixed time window, filters and identifies valid interaction segments, eliminates invalid data, and calculates the total duration of these valid segments. The accumulated duration is compared with a preset threshold: if the accumulated duration exceeds the threshold, a personality adjustment identifier is generated and its state is set to true; if it does not exceed the threshold, the identifier remains false, and the system waits for the next time window detection cycle. This achieves conditional triggering of personality adjustment, avoiding frequent personality fluctuations caused by brief or accidental interactions. Through the dual control of time windows and thresholds, it ensures that personality adjustment only occurs after the user has engaged in substantial interaction for a sufficient duration, making the personality development process more stable and reasonable, consistent with the logic of long-term development.

[0064] Step 202: Based on personality adjustment identifiers and user interaction signals, calculate the adjustment amount of personality parameters through interaction ratio analysis.

[0065] Specifically, the personality adjustment identifier is a signal indicating whether personality parameter adjustments are needed. Interaction ratio analysis is a calculation method that quantifies the impact of user interactions on personality by statistically analyzing the ratio of positive to negative interactions; its core is analyzing the tendencies of user behavior. The personality parameter adjustment amount is a calculated numerical result containing adjustment values ​​in two dimensions: an intimacy value adjustment reflecting changes in closeness and an extroversion value adjustment reflecting changes in social inclination, used to update the current personality parameters. When the personality adjustment identifier is true, the terminal initiates this step, extracts the detailed list of user interaction signals within the time window, lists all valid interaction actions, and performs a dual judgment on each interaction action based on a predefined interaction action rule base: first, it determines its impact on intimacy value, and second, it determines its impact on extroversion value. For example, a gentle touch may be marked as positive intimacy value, while an enthusiastic conversation may be marked as positive extroversion value. The number of positive and negative intimacy value actions is counted separately, the ratio is calculated, the preset adjustment level table is queried, the basic adjustment range is determined, and the basic range is adjusted in a personalized manner based on the current personality type to obtain the final adjustment amount for intimacy value and extroversion value.

[0066] Step 203: Update the personality parameters based on the personality parameter adjustment amount to obtain the updated personality parameters; and match the updated personality parameters with the four temperament personality judgment threshold table to obtain the updated personality type.

[0067] Specifically, the personality parameter adjustment amount includes the specific numerical change that needs to be applied to the current personality parameter. Personality parameters are two core internal variables describing the robot's personality traits, jointly defining the personality foundation. The updated personality parameter is the new value obtained after applying the personality parameter adjustment amount to the current personality parameter; it represents the latest state of the personality parameter. The four-temperament personality judgment threshold table is a preset mapping table that defines the correspondence between different combinations of intimacy and extraversion value ranges and four personality types: choleric, sanguine, phlegmatic, and melancholic. For example, an intimacy value greater than 50 and an extraversion value greater than 50 correspond to choleric. The terminal obtains the current personality parameter and the personality parameter adjustment amount, performs a parameter update operation, adds the adjustment amount to the corresponding current parameter value, and performs boundary checks on the updated value to limit it within a preset range. After obtaining the updated personality parameter, it is matched with the four-temperament personality judgment threshold table. The updated value is checked sequentially to see which personality type's definition range it falls within in the threshold table, and the matched personality type is taken as the final result. If the numerical change causes a change in personality type, this personality conversion event is recorded.

[0068] This embodiment maps quantitative parameter changes to qualitative personality types by updating personality types, making personality development no longer an abstract concept, but a process with clear rules and definite results. The updating of personality types will directly affect the robot's tendency in emotional expression and behavioral feedback, thereby achieving the core goal of dynamic personality evolution and significantly improving the realism and personalization of the companionship experience.

[0069] In one embodiment, based on personality adjustment identifiers and user interaction signals, the adjustment amount of personality parameters is calculated through interaction ratio analysis, including:

[0070] Step 301: Based on the personality adjustment identifier, extract the voice signal and touch signal from the user interaction signal to obtain the interaction record details.

[0071] The personality adjustment identifier is a Boolean trigger signal. When this identifier is true, it indicates that the cumulative duration of user interaction signals within a fixed time window has exceeded a preset threshold, requiring the initiation of the personality parameter adjustment process. This is the execution condition for this step. User interaction signals include all raw data generated during the interaction between the user and the robot, mainly divided into audio waveform data collected by the microphone and force, position, and duration data collected by the pressure sensor. The interaction record details are a structured list of extracted and organized interaction data, recording each independent interaction action in chronological order. It includes metadata such as action type, timestamp, duration, and signal strength, providing clear input for subsequent interaction determination. The terminal extracts all data within the corresponding time window from the stored user interaction signal buffer, segments and classifies these raw signals, divides the continuous voice stream into independent voice segments, clusters touch signals into discrete touch events, assigns a unique identifier to each identified interaction action, and, together with its feature parameters, organizes them into a structured interaction record in chronological order. The process also includes data cleaning to remove invalid or erroneous signal segments, transforming the raw, continuous user interaction signals into structured, analyzable data records.

[0072] Step 302: Based on the predefined interaction action rule library, perform intimacy value interaction judgment and extroversion value interaction judgment on each interaction action in the interaction record details to obtain a dual interaction record list.

[0073] Specifically, the predefined interaction action rule base is a knowledge base that stores the mapping relationship between various interaction actions and their influence on personality parameters. It clearly defines how different types of voice content, tone, touch methods, and other actions affect intimacy and extraversion values. Intimacy value interaction determination is performed on each action in the interaction record details, judging its direction and degree of influence on the intimacy value parameter based on the rule base. The determination result is usually positive, negative, or neutral. Extraversion value interaction determination is performed on the same interaction action, judging its direction and degree of influence on the extraversion value parameter based on the rule base. The determination result is also divided into positive, negative, or neutral. The dual interaction record list is based on the interaction record details, adding a new list of intimacy value and extraversion value influences for each interaction action, clearly indicating the specific impact of each historical interaction action on the two personality dimensions. The terminal loads the predefined interaction action rule base, traverses each interaction action in the interaction record details, matches its features with the conditions in the rule base for each action, and performs intimacy value and extraversion value interaction determinations for the same action in parallel. All original interaction records and their corresponding dual determination results are integrated to generate a dual interaction record list. For example, a gentle, prolonged stroking is considered to have a positive impact on intimacy; a high-pitched, fast-paced voice interaction is considered to have a positive impact on extroversion.

[0074] Step 303: Calculate the intimacy value adjustment amount based on the dual interaction record list; and calculate the outward value adjustment amount based on the dual interaction record list.

[0075] Specifically, the dual interaction record list is a complete record containing each interaction action and its impact on intimacy and extraversion values. The intimacy adjustment amount is a numerical result representing the total recommended adjustment to the current intimacy parameter based on all interaction records within the current time window; its calculation relies on the statistical analysis of all intimacy interaction judgments in the list. The extraversion adjustment amount is a numerical result representing the total recommended adjustment to the current extraversion parameter based on all interaction records within the current time window; its calculation relies on the statistical analysis of all extraversion interaction judgments in the list. The terminal iterates through the dual interaction record list, counting the number of all interactions marked as positive intimacy values ​​and the number of all interactions marked as negative intimacy values. It calculates the intimacy interaction ratio according to the formula, queries the personality adjustment level table based on the ratio to obtain a basic adjustment range value, and then, combined with the robot's current personality type, calls the corresponding fine-tuning rules to correct this basic range value, obtaining the final intimacy adjustment amount. The process of calculating the extraversion value adjustment is similar to that of calculating the intimacy value adjustment, but the statistics and rules are based on the extraversion value dimension. The number of positive and negative extraversion actions is counted, the ratio is calculated, the level table is consulted to obtain the base value, and finally, fine-tuning is performed according to personality type to obtain the extraversion value adjustment.

[0076] Step 304: Integrate the intimacy value adjustment amount and the extroversion value adjustment amount to obtain the personality parameter adjustment amount.

[0077] The personality parameter adjustment amount is a data structure containing two components, explicitly indicating how much the intimacy and extraversion values ​​should change within the current adjustment period. The terminal integrates the two independently calculated adjustment amounts, encapsulating them into a unified data structure or object for subsequent steps to read and process in one go.

[0078] This embodiment aggregates a large number of discrete interaction judgment results into two concise, numerical adjustment quantities. Through statistical analysis and rule conversion, it ensures that the calculation of the adjustment quantities reflects both the overall tendency of interactive behavior and takes into account individual sensitivity differences, making the updating of personality parameters more scientific and reasonable.

[0079] In one embodiment, the intimacy value adjustment amount is calculated based on the dual interaction record list, including:

[0080] Step 401: Traverse the dual interaction record list, count the number of interaction actions marked as positive intimacy value, and obtain the positive count; and count the number of interaction actions marked as negative intimacy value, and obtain the negative count.

[0081] The Dual Interaction Record List is a structured data list that records all user interactions within a fixed time window. Each action is labeled with its impact on intimacy and extraversion values. Each entry includes the action type, timestamp, and intimacy interaction result. The positive count refers to the cumulative number of interactions marked as positive for intimacy in the Dual Interaction Record List, representing interactions that increase intimacy, such as a gentle touch or friendly verbal encouragement. The negative count refers to the cumulative number of interactions marked as negative for intimacy, representing interactions that decrease intimacy, optionally including pats or indifferent verbal responses. The terminal iterates through each interaction action entry in the dual interaction record list. During the iteration, it checks the intimacy value interaction judgment field of each action. If the judgment result is positive, the positive count counter is incremented by 1; if the judgment result is negative, the negative count counter is incremented by 1; neutral judgments are ignored and do not participate in the counting. After the iteration is completed, two independent integer values ​​are output: the positive count and the negative count. This ensures the comprehensiveness and accuracy of the statistics, based solely on predefined judgment rules, avoiding subjective bias.

[0082] Step 402: Based on the number of positive and negative interactions, calculate the intimacy interaction ratio using the following formula:

[0083]

[0084] Where R is the ratio of intimate interactions, P is the number of positive interactions, and N is the number of negative interactions.

[0085] Specifically, the intimacy interaction ratio is a numerical result representing the proportion of positive to negative interactions. It quantifies the net impact of user interactions on intimacy value. An R greater than 1 indicates that positive interactions dominate, while an R less than 1 indicates that negative interactions dominate. The ratio is used to assess interaction quality and is a key intermediate variable in calculating the adjustment magnitude. The terminal calculates the ratio based on the number of positive and negative interactions using a formula, handling boundary cases. If there are no negative interactions, R is considered to be infinity or a preset maximum value to avoid division by zero errors. If both P and N are 0, R is set to 1, representing a neutral state. This is a purely mathematical operation, emphasizing the objectivity of the ratio and avoiding subjective judgment. The frequency statistics are transformed into a standardized indicator, eliminating the influence of absolute quantities and making the window periods of different interaction intensities comparable.

[0086] Step 403: Based on the intimacy interaction ratio, query the personality adjustment level table to obtain the basic adjustment range value.

[0087] Specifically, the personality adjustment level table is a pre-defined mapping table that defines the correspondence between the intimacy interaction ratio and the basic adjustment range value. It contains multiple ratio ranges, each mapping to a specific adjustment range value. The table is based on historical data or psychological models to ensure that the adjustment range is in a reasonable proportion to the interaction ratio. The basic adjustment range value is a value obtained by querying the personality adjustment level table. It represents the basic adjustment amount recommended for the intimacy value parameter without considering personality differences. It is an intermediate result and will be fine-tuned later based on personality type. The terminal obtains the intimacy interaction ratio and matches it with intervals in the personality adjustment level table. The matching process is a sequential query, traversing each interval condition in the table and checking whether the R value falls within that interval. Once a matching interval is found, the corresponding basic adjustment range value is read. If the R value is an extreme case where no interval is matched, the default range value is used. The query operation is based on precise numerical comparison to ensure the repeatability of the results, transforming the abstract ratio into specific adjustment suggestions. The adjustment range is standardized through a preset level table, avoiding hard-coded rules and enabling flexible adaptation to different interaction scenarios. At the same time, it provides a consistent and configurable benchmark for basic adjustment. The basic adjustment range value is the basis for subsequent fine-tuning, ensuring the modularity and maintainability of the adjustment process.

[0088] Step 404: Based on personality type, call the corresponding intimacy value adjustment rule for personality type, adjust the base adjustment range value, and obtain the intimacy value adjustment amount.

[0089] The personality type is a category label describing the current personality traits of the AI ​​pet, including choleric, sanguine, phlegmatic, and melancholic. Personality type affects the sensitivity of parameter adjustments; for example, melancholic pets are more sensitive to negative interactions. The intimacy value adjustment rules are a set of predefined fine-tuning rules that specify how to adjust the base adjustment range for different personality types. Optionally, for melancholic personalities, the adjustment range for negative interactions needs to be increased, while for choleric personalities, it may be decreased. The intimacy value adjustment amount represents the total recommended adjustment to the intimacy value parameter after personality type fine-tuning, and is directly used to update the personality parameter. Illustratively, the terminal obtains the base adjustment range value and the current personality type, calls the intimacy value adjustment rule corresponding to the personality type. If the personality type is melancholic, the rule may require multiplying the base range value by a coefficient greater than 1 to increase sensitivity; if the personality type is choleric, it may multiply by a coefficient less than 1 to reduce the impact. The fine-tuning operation involves arithmetic operations. The rules are based on personality traits to ensure personalization, and the final intimacy value adjustment amount is output, which is limited within a certain range to prevent excessive fluctuations.

[0090] This embodiment fine-tunes the personality type, making the intimacy value adjustment more closely match the robot's personality traits, thus enhancing the realism of personality development. The intimacy value adjustment provides precise input for updating the overall personality parameters, ensuring that personality evolution is based on interaction data and adapts to individual differences, improving the realism of the user experience. This enables the AI ​​pet robot's personality development to dynamically respond to user interactions, thereby achieving dynamic personality evolution and personalized companionship. The entire process emphasizes quantitative analysis and rule-based processing, avoiding subjective arbitrariness and providing a stable and reliable personality adjustment mechanism.

[0091] In one embodiment, based on scene parameters and emotion type determination results, the robot's core state parameters are calculated to obtain dual-axis parameter values, including:

[0092] Step 501: Calculate the basic dual-axis parameter values ​​based on the scene parameters and personality type; the scene parameters include time scene, interaction scene, environmental scene and device scene.

[0093] The scenario parameters are a set of quantitative indicators describing the environment and interaction scenarios of the AI ​​pet robot, including: time scenario (specific time of day affecting the AI ​​pet robot's circadian rhythm); interaction scenario (type of interaction between the user and the pet); environmental scenario (physical environmental parameters); and device scenario (the AI ​​pet's own battery level, hardware operating status, etc.). Personality type is the current personality trait classification of the AI ​​pet; each personality type has different sensitivities and reaction patterns to the same scenario. The basic dual-axis parameter values ​​are initial state values ​​calculated based on the scenario parameters and personality type, containing two dimensions: basic comfort, which is an initial score reflecting the pet's adaptability to the environment; and basic activity, which is an initial score reflecting the pet's energy level. The terminal loads all current scenario parameters and performs weighted calculations according to a pre-set rule base. Each scenario parameter generates a basic influence value according to preset rules, and the basic influence value is personalized based on the current personality type. For example, the personality sensitivity rule indicates that melancholic temperament is greater than phlegmatic temperament, which is greater than sanguine temperament, which is greater than choleric temperament. That is, under the same negative scenario, the parameter attenuation of melancholic temperament is the greatest. By comprehensively considering the characteristics of the scenario and personality differences, the basic dual-axis parameter values ​​establish a state basis that conforms to the real biological logic, ensuring the rationality and individual differences of the AI ​​pet's response to the environment.

[0094] Step 502: If the emotion type determination result matches the preset forced overriding emotion, then based on the emotion type determination result, call the corresponding fixed dual-axis parameter value to overwrite the basic dual-axis parameter value to obtain the dual-axis parameter value.

[0095] Specifically, the emotion type determination result is the user's current emotional state obtained through multimodal recognition, including forced coverage emotions and no emotion. The preset forced coverage emotions are user emotion types that must be prioritized in advance, including joy, anger, and sorrow. Fixed dual-axis parameter values ​​are preset parameter values ​​for each forced coverage emotion. For example, when a user is joyful, the comfort level is forcibly set to 80, and when a user is angry, the activity level is forcibly set to 30. The dual-axis parameter values ​​are the final parameter values ​​after forced coverage processing, directly determining the AI ​​pet's emotional state. The terminal matches the emotion type determination result with the forced coverage emotion list. If it detects that the user is in a strong emotional state, and that emotion belongs to the forced coverage type, it triggers the coverage process, calling the corresponding fixed dual-axis parameter values ​​to completely replace the base calculated values. Optionally, if the user is angry, regardless of the base comfort level, it is forcibly increased to 50, and the activity level is forcibly decreased to 30; if the user is joyful, the comfort level is forcibly increased to 80, and the activity level is increased to 60 if it is below 60. User emotion adjustment has the highest priority, and the coverage rules ignore the influence of scene parameters.

[0096] Step 503: If the emotion type determination result does not match the forced overriding emotion, then the basic dual-axis parameter value is determined as the dual-axis parameter value.

[0097] Specifically, the dual-axis parameter values ​​are the final output values ​​under non-forced coverage conditions, and in this case, they remain consistent with the basic dual-axis parameter values. When the emotion type determination result does not belong to the forced coverage emotion, the terminal performs simplified processing, confirms that the user's emotion is a non-priority response type, and directly determines the calculated basic dual-axis parameter values ​​as the final output values. At this time, the AI ​​pet's state is entirely determined by the scene parameters and personality type, maintaining normal scene response logic.

[0098] This embodiment ensures environmental adaptability through a basic computing layer and guarantees emotional empathy through an overlay processing layer, forming a dual guarantee from environmental baseline to emotional priority. The forced overlay mechanism enables rapid response to strong user emotions, greatly improving the realism of the interaction. It ensures that when users are emotionally strong, the AI ​​pet's state can make timely and appropriate responses, avoiding any sense of interaction incongruity. This is a key technical feature for achieving two-way emotional linkage.

[0099] In one embodiment, based on user interaction signals in a multimodal scene data packet, the user's current emotional state is identified to obtain an emotion type determination result, including:

[0100] Step 601: Extract voice signals and touch signals from user interaction signals; and preprocess the voice signals and touch signals to obtain a voice data frame sequence and a touch event sequence.

[0101] Specifically, the user interaction signal is the raw data stream generated during the interaction between the user and the AI ​​pet. It includes two independent channels: the voice signal is audio waveform data collected by a microphone array, containing information such as the user's tone, volume, and speech rate; the touch signal is physical contact data collected by a piezoelectric sensor grid, containing information such as touch position, pressure value, and duration. The voice data frame sequence is a sequence of discrete data units obtained by framing continuous voice signals. Each frame is 20 to 40 milliseconds long and contains audio waveform sampling points and their timestamps, providing standardized input for subsequent feature extraction. The touch event sequence is a discrete sequence of interactive actions identified from continuous touch signals. Each touch event contains attributes such as trigger time, pressure intensity, contact area, and duration, arranged in chronological order to form a sequence. The terminal separates the voice signal and touch signal from the signal buffer. The voice signal is preprocessed, pre-emphasized, and denoised. The continuous voice is segmented into overlapping short frames through frame windowing to form a voice data frame sequence. The touch signal is subjected to event detection. A pressure threshold is set to distinguish between effective touch and accidental contact. For signals exceeding the threshold, touch point clustering and boundary recognition are performed to generate a structured touch event sequence. The entire process ensures that the two signals are synchronized on the time axis, laying the foundation for multimodal fusion.

[0102] Step 602: Extract features from the speech data frame sequence to obtain speech feature vectors; and input the speech feature vectors into the speech emotion classifier to obtain the speech emotion determination result; the speech feature vectors include at least one of fundamental frequency, energy, speech rate and spectral features.

[0103] The speech feature vector is a set of numerical features extracted from a sequence of speech data frames. These features include: fundamental frequency (FFF), which reflects the pitch of the voice (increased with anger, decreased with sadness); energy, which is the intensity of the sound (increased with anger, decreased with sadness); speech rate, which is the number of syllables per unit time (faster with excitement, slower with frustration); and spectral features, which reflect frequency distribution characteristics and timbre variations. The speech emotion classifier is a trained machine learning model that maps speech feature vectors to emotion categories. The speech emotion determination result is the emotion category label and confidence score output by the classifier. The terminal extracts features frame by frame from the speech data frame sequence, calculates the fundamental frequency through autocorrelation algorithm, obtains the energy value through short-time energy calculation, calculates the speech rate through syllable boundary detection, and extracts spectral features through fast Fourier transform. The features of the entire sequence are statistically aggregated to form a fixed-dimensional speech feature vector. The vector is input into a pre-trained speech emotion classifier. The classifier calculates the most likely emotion category and the corresponding confidence level through multi-layer decision boundary calculation. The entire process adopts a sliding window mechanism to ensure real-time performance. Through quantitative analysis of acoustic features, subjective speech emotion is transformed into an objective numerical judgment.

[0104] Step 603: Extract features from the touch event sequence to obtain touch feature vectors; and input the touch feature vectors into the touch behavior classifier to obtain the touch emotion judgment result.

[0105] The touch feature vector is a set of numerical features extracted from the touch event sequence, including pressure features, temporal features, and spatial features. The touch behavior classifier is a classification model specifically trained for touch behavior, identifying emotional tendencies through pressure patterns and temporal dynamics features. The touch emotion judgment result is the emotion category judgment output by the classifier based on the touch features. The terminal analyzes the attributes of each event in the touch event sequence, calculates the statistical characteristics of the pressure value, analyzes the touch time pattern, and statistically analyzes the touch position distribution. After combining the features into a touch feature vector, it is input into the touch behavior classifier. The classifier judges the emotional state behind the touch behavior by analyzing the relationship between pressure and time patterns, paying particular attention to the dynamic changes of continuous touch events. It provides emotion evidence independent of speech, is suitable for scenarios where voice interaction is limited, and provides an important second layer of verification for emotion judgment.

[0106] Step 604: Based on the voice emotion determination result and the touch emotion determination result, generate the emotion type determination result.

[0107] The emotion type determination result is the final emotion assessment generated by integrating both voice and touch evidence. It includes the emotion category and overall confidence level, serving as a standardized emotion description for external output. The terminal compares the results of the two modalities. If they match, the result is directly adopted with increased confidence; otherwise, arbitration is performed based on preset priority rules and weighted confidence levels. The voice assessment result is adopted first, with the touch assessment result serving as an auxiliary factor. For example, when the voice assessment is anger and the touch assessment is joy, the voice result is adopted first, but the final confidence level is appropriately reduced. The fusion process also considers time synchronization to ensure that the comparison is made within the same time window. The final emotion type determination result includes a quality assessment flag for subsequent modules to reference.

[0108] This embodiment significantly improves the robustness and accuracy of emotion recognition through a multimodal fusion mechanism. By verifying dual evidence, it effectively overcomes the limitations of single-modal recognition, making emotion judgment more consistent with the real psychological state. This ensures that the AI ​​pet robot can accurately understand the user's emotions and make appropriate responses, thus solving the problem of incongruity in one-way emotion output.

[0109] In one embodiment, the dual-axis parameter values ​​and the robot's personality type are matched with preset emotional behaviors to obtain a set of behavioral instructions, including:

[0110] Step 701: Based on the dual-axis parameter values, query the discrete emotion label that uniquely corresponds to the emotion mapping table to obtain the basic emotion state.

[0111] The dual-axis parameter values ​​are two-dimensional parameter pairs containing comfort and activity levels, accurately describing the AI ​​pet's current internal state. The emotion map is a pre-defined two-dimensional lookup table that divides the continuous dual-axis parameter space into multiple discrete intervals, each corresponding to a specific basic emotional state label. The basic emotional state is obtained by querying the emotion map, and these discrete emotion labels serve as the initial basis for behavior selection. The terminal obtains the real-time calculated dual-axis parameter values ​​and performs interval matching in the emotion map. The query process uses a nearest neighbor matching algorithm to ensure smooth transitions at parameter boundaries. The emotion map employs a five-level discretization strategy, dividing comfort and activity levels into five intervals: 0-20, 20-40, 40-60, 60-80, and 80-100, respectively. A unique basic emotional state is determined through two-dimensional cross-location. For example, the comfort level interval of 60-80 combined with the activity level interval of 40-60 corresponds to an enjoyment state.

[0112] Step 702: Using personality type and basic emotional state as the joint key, retrieve the corresponding basic behavioral template from the behavioral rule base.

[0113] Specifically, personality type refers to the current four temperament categories of AI pets. Basic emotional state is a discrete emotional label. The composite key is a compound query condition composed of personality type and basic emotional state. The behavior rule base is a database storing behavior templates corresponding to various composite keys. The basic behavior template is a behavior blueprint containing parameters such as action type, default amplitude, and baseline frequency, but it has not yet been personalized. The terminal uses personality type and basic emotional state as composite keys to perform precise matching and retrieval in the behavior rule base. The behavior rule base adopts a hierarchical index structure, partitioned by personality type, and within each personality partition, a secondary index is built according to emotional state to match completely identical composite keys. Each basic behavior template contains a complete description of the action sequence.

[0114] Step 703: Based on the emotion type determination result, fine-tune the amplitude and frequency of the basic behavior template to obtain the fine-tuned behavior.

[0115] Specifically, the emotion type determination result comes from the user's emotion recognition module. The basic behavior template is the retrieved original behavior plan. Amplitude is the intensity parameter of the behavior action. Frequency is the time density parameter of the behavior action. Fine-tuned behavior is the personalized behavior plan after amplitude and frequency adjustments. The terminal dynamically adjusts the basic behavior template based on the emotion type determination result. The fine-tuning rules adopt a sensitivity coefficient mechanism, selecting the corresponding fine-tuning coefficient table according to the user's emotion type, and making secondary adjustments in conjunction with personality traits. Frequency adjustment mainly targets periodic actions. The fine-tuning process ensures that the basic behavior type is not changed, only its dynamic performance parameters are adjusted, achieving a deep linkage between user emotions and pet behavior, enabling the AI ​​pet to exhibit empathy.

[0116] Step 704: Map the fine-tuning behavior to the corresponding robot's execution parameters to obtain the behavior instruction set.

[0117] Among these, fine-tuning behavior refers to a complete behavior plan that has been personalized and adjusted. Execution parameters are low-level instruction parameters that directly control the robot hardware. The behavior instruction set is a time-synchronized sequence of execution parameters, containing control instructions for multiple actuators and their timestamps. The terminal maps the abstract fine-tuning behavior to specific hardware control instructions. The mapping process employs a hierarchical conversion strategy, decomposing the behavior into action units and generating corresponding execution parameters for each unit. For multimodal output, a time synchronization mechanism is established to ensure coordination between different modalities. The generated behavior instruction set uses a timestamp queue format to precisely control the start time and duration of each action.

[0118] This embodiment ensures that each internal state has a unique corresponding behavioral expression through the joint retrieval of an emotion mapping table with 25 intervals and a behavior rule base, thus solving the problem of monotonous emotion-driven behavior. The basic template guarantees the basic rationality of the behavior, while the fine-tuning mechanism adds personalized variation, giving each AI pet unique behavioral characteristics. The execution parameter mapping process adopts a hardware abstraction layer design, which allows the same set of behavior schemes to be adapted to different models of robot hardware, improving the versatility and portability of the technical solution and achieving the design goal of a deeply realistic companion experience.

[0119] To further illustrate the action interaction scheme of this application embodiment, several specific scenario examples are described below.

[0120] In this embodiment, an interactive control method for a robot is used in the form of an AI pet.

[0121] Initial state settings include:

[0122] Personality type: Sanguine (Extroversion 75, Intimacy 0, conforming to extroversion ≥ 50, intimacy < 50);

[0123] Dual-axis parameters: Comfort 50, Vitality 50 (50% battery, baseline reset at 7:00, no scene / user emotion adjustment);

[0124] Emotional state: Calm (comfort level 40-60, liveliness level 40-60, sanguine, occasional head shaking, no high-frequency movements);

[0125] Initial conditions: ambient temperature 25℃, noise level 50dB, battery level 50% (no abnormal influences).

[0126] The interaction process and module linkage include:

[0127] Scenario 1: User's positive emotions are linked (18:00-18:20, evening period, battery level 60%)

[0128] Scene awareness: The user's voice tone is pleasant (base frequency 280Hz), and the user lightly touches the head 6 times (force 1.5N, behavioral signal).

[0129] User emotion recognition: If the voice is pleasant and the behavior is gentle, it is judged as happiness;

[0130] Numerical calculation: Dual-axis parameter calculation: Comfort = 50 (initial), forced to increase to 80 (covering evening scenarios); Activity = 60 (60% battery), evening +20, then 80 (because it is not necessary to increase to 60).

[0131] Personality parameter calculation: If the interaction time is 20 minutes > 10 minutes, there are 6 positive interactions and 0 negative interactions (ratio ≥ 2), then the intimacy value is 0 + 5 = 5, and the extroversion value is 75 + 5 = 80.

[0132] Dual-axis emotion engine: Comfort level 80+ and liveliness level 80 indicate the emotion of laughter (sanguine temperament);

[0133] Feedback control: Output actions that conform to the adaptation rules of sanguine laughter and user preferences;

[0134] Personality status: Still sanguine (intimacy 5 < 50), intimacy value is accumulating positively.

[0135] Scenario 2: User anger triggered (19:00-19:10, battery 55%, noise 65dB)

[0136] Scene awareness: The user says in voice: "Working overtime again! So annoying" (base frequency 380Hz, score -0.7), and pats the pet twice (force 3.5N, behavior signal);

[0137] User emotion recognition: If the user's voice indicates anger and they also pat their face, the emotion is determined to be anger.

[0138] Numerical calculation: Dual-axis parameter calculation: If the activity value is 80 (previous scene), then it will be forcibly reduced to 30 (to cover noise, evening scene); if the comfort value is 80, then it will be maintained at 80 (no need to increase it to 50).

[0139] Personality parameter calculation: With an interaction duration of 10 minutes, 0 positive interactions and 2 negative interactions (ratio ≤ 0.5), the intimacy value is 5-5=0, and the extroversion value is 80-5=75.

[0140] Dual-axis emotion engine: Comfort level 80+ and liveliness level 30 indicate a comfortable emotion (adapting to the user's anger and turning it into a calm tendency).

[0141] Feedback control: Outputs a low-frequency humming sound, conforming to the adaptation rules of pernicious comfort and user anger;

[0142] Personality status: Still sanguine (extroversion 75≥50, intimacy 0<50).

[0143] To further illustrate the personality update scheme of this application, several specific scenario examples are provided below.

[0144] Taking the initial sanguine temperament (extroversion 75, intimacy 0) as an example, this demonstrates the personality transformation process based on user interaction:

[0145] Sanguine → Choleric: Daily 17:00-20:00 (high-frequency interaction when the user is happy), 20 minutes of interaction, 10 positive interactions (hugs, head pats), 5 additional comforting times when the user is happy, for 7 consecutive days, the intimacy value is 0+5×7=35. On the 8th day, add 15 minutes of back patting at noon, the intimacy value is 35+5=40. On the 9th day, interact with the user in the evening when the user is happy, the intimacy value is 40+5=50, the extroversion value is 75+5×9=120, which is converted to the upper limit of 100, satisfying the choleric threshold (extroversion ≥50, intimacy ≥50).

[0146] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0147] Based on the same inventive concept, this application also provides an interactive control system for implementing the interactive control method for a robot described above. The solution provided by this system is similar to the implementation described in the above method; therefore, the specific limitations of one or more embodiments of the interactive control system for robots provided below can be found in the limitations of the interactive control method for a robot described above, and will not be repeated here.

[0148] In one exemplary embodiment, such as Figure 2 As shown, an interactive control system 800 for a robot is provided, comprising:

[0149] The standardization module 801 is used to acquire physical environment signals and user interaction signals; and to standardize the physical environment signals and user interaction signals to obtain multimodal scene data packets.

[0150] The scene emotion module 802 is used to identify the user's current emotional state based on the user interaction signals in the multimodal scene data package, and obtain the emotion type determination result; and to calculate scene parameters based on the multimodal scene data package; the user interaction signals include voice signals and touch signals to the robot;

[0151] The dual-axis parameter module 803 is used to calculate the robot's core state parameters based on scene parameters and emotion type determination results, and obtain dual-axis parameter values; the dual-axis parameter values ​​include comfort and liveliness values.

[0152] The behavior instruction module 804 is used to match the dual-axis parameter values ​​and the robot's personality type with preset emotional behaviors to obtain a behavior instruction set; the behavior instruction set is used to instruct the robot to perform actions.

[0153] Furthermore, the system also includes a personality update module for:

[0154] Based on a fixed time window, user interaction signals are acquired; if the cumulative duration of user interaction signals exceeds a preset threshold, a personality adjustment identifier is generated; the personality adjustment identifier indicates that a personality type update is required.

[0155] Based on personality adjustment identifiers and user interaction signals, the adjustment amount of personality parameters is calculated through interaction ratio analysis.

[0156] The personality parameters are updated based on the adjustment amount of the personality parameters to obtain the updated personality parameters; then the updated personality parameters are matched with the four temperament personality judgment threshold table to obtain the updated personality type.

[0157] Furthermore, the personality update module is also used for:

[0158] Based on personality adjustment identifiers, voice and touch signals are extracted from user interaction signals to obtain a detailed list of interaction records;

[0159] Based on a predefined interactive action rule library, each interactive action in the interaction record details is judged for both intimacy value interaction and extroversion value interaction, resulting in a dual interaction record list.

[0160] Based on the dual interaction record list, calculate the intimacy value adjustment amount; and based on the dual interaction record list, calculate the extroversion value adjustment amount;

[0161] By integrating the intimacy value adjustment and the extroversion value adjustment, the personality parameter adjustment is obtained.

[0162] Furthermore, the personality update module is also used for:

[0163] Iterate through the list of dual interaction records, count the number of interaction actions marked as positive for intimacy value, and get the positive count; and count the number of interaction actions marked as negative for intimacy value, and get the negative count;

[0164] Based on the number of positive and negative interactions, the intimate interaction ratio is calculated using the following formula:

[0165]

[0166] Where R is the ratio of intimate interactions, P is the number of positive interactions, and N is the number of negative interactions;

[0167] Based on the intimacy interaction ratio, the basic adjustment range value is obtained by consulting the personality adjustment level table;

[0168] Based on personality type, the corresponding intimacy value adjustment rule is invoked to adjust the base adjustment value and obtain the intimacy value adjustment amount.

[0169] Furthermore, the dual-axis parameter module 803 is also used for:

[0170] Based on scene parameters and personality type, calculate basic dual-axis parameter values; scene parameters include time scene, interaction scene, environmental scene and device scene;

[0171] If the emotion type determination result matches the preset forced overriding emotion, then based on the emotion type determination result, the corresponding fixed dual-axis parameter value is called to overwrite the basic dual-axis parameter value to obtain the dual-axis parameter value;

[0172] If the emotion type determination result does not match the forced overriding emotion, then the basic dual-axis parameter value will be determined as the dual-axis parameter value.

[0173] Furthermore, the scene emotion module 802 is also used for:

[0174] Extract voice and touch signals from user interaction signals; and preprocess the voice and touch signals to obtain a voice data frame sequence and a touch event sequence.

[0175] The speech data frame sequence is subjected to feature extraction to obtain a speech feature vector; the speech feature vector is then input into a speech emotion classifier to obtain a speech emotion determination result; the speech feature vector includes at least one of fundamental frequency, energy, speech rate and spectral features.

[0176] Feature extraction is performed on the touch event sequence to obtain touch feature vectors; and the touch feature vectors are input into a touch behavior classifier to obtain the touch emotion judgment result.

[0177] Based on the voice emotion determination results and the touch emotion determination results, an emotion type determination result is generated.

[0178] Furthermore, the behavior instruction module 804 is also used for:

[0179] Based on the dual-axis parameter values, the basic emotional state is obtained by querying the unique discrete emotion label in the emotion mapping table.

[0180] Using personality type and basic emotional state as the joint key, the corresponding basic behavioral templates are retrieved from the behavioral rule base;

[0181] Based on the emotion type determination results, the amplitude and frequency of the basic behavior template are fine-tuned to obtain the fine-tuned behavior;

[0182] By mapping fine-tuning behaviors to the corresponding robot execution parameters, a set of behavior instructions is obtained.

[0183] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a power supply safety management method as described above.

[0184] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0185] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0186] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. An interactive control method for a robot, characterized in that, The method includes: Acquire physical environment signals and user interaction signals; and perform standardization processing on the physical environment signals and user interaction signals to obtain multimodal scene data packets; Based on the user interaction signals in the multimodal scene data package, the user's current emotional state is identified to obtain the emotion type determination result; and based on the multimodal scene data package, scene parameters are calculated; the user interaction signals include voice signals and touch signals to the robot; Based on the scene parameters and the emotion type determination results, the core state parameters of the robot are calculated to obtain dual-axis parameter values; the dual-axis parameter values ​​include comfort and activity values, the comfort value reflects the robot's physical comfort in the current environment and interaction, and the activity value reflects the robot's energy level and willingness to be active; The dual-axis parameter values ​​and the robot's personality type are matched with preset emotional behaviors to obtain a set of behavioral instructions; this set of instructions is used to direct the robot to perform actions. Based on the scene parameters and the emotion type determination result, the robot's core state parameters are calculated to obtain dual-axis parameter values, including: Based on the scenario parameters and the personality type, calculate the basic dual-axis parameter values; the scenario parameters include time scenario, interaction scenario, environmental scenario, and device scenario; If the emotion type determination result matches the preset forced overriding emotion, then based on the emotion type determination result, the corresponding fixed dual-axis parameter value is called to overwrite the basic dual-axis parameter value to obtain the dual-axis parameter value; If the emotion type determination result does not match the forced coverage emotion, then the basic dual-axis parameter value is determined as the dual-axis parameter value.

2. The method according to claim 1, characterized in that, The robot's personality type is updated using the following method: The user interaction signal is acquired based on a fixed time window; if the cumulative duration of the user interaction signal exceeds a preset threshold, a personality adjustment identifier is generated; the personality adjustment identifier indicates that a personality type update is required. Based on the personality adjustment identifier and the user interaction signal, the personality parameter adjustment amount is calculated through interaction ratio analysis. The personality parameters are updated based on the aforementioned personality parameter adjustment amount to obtain the updated personality parameters; The updated personality parameters are then matched with the four-temperament personality determination threshold table to obtain the updated personality type.

3. The method according to claim 2, characterized in that, The step of calculating the personality parameter adjustment amount based on the personality adjustment identifier and the user interaction signal through interaction ratio analysis includes: Based on the personality adjustment identifier, the voice signal and the touch signal in the user interaction signal are extracted to obtain a detailed interaction record; Based on a predefined interactive action rule library, each interactive action in the detailed interactive record is judged for both intimacy value and outward value interaction to obtain a dual interactive record list. Based on the dual interaction record list, calculate the intimacy value adjustment amount; and based on the dual interaction record list, calculate the extroversion value adjustment amount; By integrating the intimacy value adjustment amount and the extroversion value adjustment amount, the personality parameter adjustment amount is obtained.

4. The method according to claim 3, characterized in that, The calculation of the intimacy value adjustment based on the dual interaction record list includes: Iterate through the list of dual interaction records, count the number of interaction actions marked as positive intimacy values ​​to obtain the positive count; and count the number of interaction actions marked as negative intimacy values ​​to obtain the negative count. Based on the positive and negative counts, the intimate interaction ratio is calculated using the following formula: Where R is the ratio of intimate interactions, P is the number of positive interactions, and N is the number of negative interactions; Based on the aforementioned intimate interaction ratio, the basic adjustment range value is obtained by querying the personality adjustment level table. Based on the personality type, the intimacy value adjustment rule corresponding to the personality type is invoked to adjust the basic adjustment range value, thereby obtaining the intimacy value adjustment amount.

5. The method according to claim 1, characterized in that, The step of identifying the user's current emotional state based on the user interaction signals in the multimodal scene data packet and obtaining an emotion type determination result includes: Extract voice signals and touch signals from the user interaction signals; and preprocess the voice signals and touch signals to obtain a voice data frame sequence and a touch event sequence; The speech data frame sequence is subjected to feature extraction to obtain a speech feature vector; the speech feature vector is then input into a speech emotion classifier to obtain a speech emotion determination result; the speech feature vector includes at least one of fundamental frequency, energy, speech rate and spectral features. The touch event sequence is subjected to feature extraction to obtain a touch feature vector; the touch feature vector is then input into a touch behavior classifier to obtain a touch emotion determination result. Based on the voice emotion determination result and the touch emotion determination result, the emotion type determination result is generated.

6. The method according to claim 5, characterized in that, The step of matching the dual-axis parameter values ​​and the robot's personality type with preset emotional behaviors to obtain a set of behavioral instructions includes: Based on the dual-axis parameter values, the basic emotional state is obtained by querying the unique discrete emotion label in the emotion mapping table. Using the personality type and the basic emotional state as a joint key, the corresponding basic behavioral template is retrieved from the behavioral rule base; Based on the emotion type determination result, the amplitude and frequency of the basic behavior template are fine-tuned to obtain the fine-tuned behavior; The fine-tuning behavior is mapped to the corresponding execution parameters of the robot to obtain the behavior instruction set.

7. An interactive control system for a robot, characterized in that, The system includes: A standardization module is used to acquire physical environment signals and user interaction signals; and to standardize the physical environment signals and user interaction signals to obtain multimodal scene data packets; The scene emotion module is used to identify the user's current emotional state based on the user interaction signals in the multimodal scene data package, and obtain the emotion type determination result; and to calculate scene parameters based on the multimodal scene data package; the user interaction signals include voice signals and touch signals to the robot; The dual-axis parameter module is used to calculate the core state parameters of the robot based on the scene parameters and the emotion type determination result, and obtain dual-axis parameter values. The dual-axis parameter values ​​include comfort and activity values. The comfort value reflects the robot's physical comfort in the current environment and interaction, and the activity value reflects the robot's energy level and willingness to be active. The behavior instruction module is used to match the dual-axis parameter values ​​and the robot's personality type with preset emotional behaviors to obtain a behavior instruction set; the behavior instruction set is used to instruct the robot to perform actions. Based on the scene parameters and the emotion type determination result, the robot's core state parameters are calculated to obtain dual-axis parameter values, including: Based on the scenario parameters and the personality type, calculate the basic dual-axis parameter values; the scenario parameters include time scenario, interaction scenario, environmental scenario, and device scenario; If the emotion type determination result matches the preset forced overriding emotion, then based on the emotion type determination result, the corresponding fixed dual-axis parameter value is called to overwrite the basic dual-axis parameter value to obtain the dual-axis parameter value; If the emotion type determination result does not match the forced coverage emotion, then the basic dual-axis parameter value is determined as the dual-axis parameter value.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images