Electronic skin sensing system and method for animal bionic interaction

Abstract

The invention relates to the field of electronic skin perception, and discloses an electronic skin perception system and method for animal bionic interaction, and the method comprises the steps: collecting tactile signals of different contact intensities, directions and vibration spectrums, and forming a basic feedback database; mapping tactile signal vectors collected in real time to a basic feedback database through a matching algorithm, and quickly determining candidate feedback situations according to strategy rules; on the basis of executing the candidate feedback situation, preprocessing the recorded data under a preset triggering condition, and uploading the preprocessed recorded data to a central processing platform to generate a new feedback situation; integrating the new feedback situation into a basic feedback library, and constructing a front feedback situation set and a rear feedback situation set in combination with historical use data of a user and a current environment condition; and performing multi-modal verification according to the recently acquired tactile execution record and the user behavior and physiological response data, and generating an updated pre-feedback situation strategy. The method has the advantage of improving the interaction precision and efficiency.

Description

Technical Field

[0001] This invention relates to the field of electronic skin sensing, specifically to an electronic skin sensing system and method for animal bionic interaction. Background Technology

[0002] With the development of bionic robots and intelligent interaction technologies, electronic skin has significant application value in animal bionic interaction. It collects multi-dimensional tactile information such as pressure, vibration, and tangential force through multi-channel tactile sensors, providing perception capabilities for the interaction system. However, existing electronic skin still has significant shortcomings in practical applications. The tactile signal processing method is simplistic, lacking comprehensive analysis of amplitude, spectrum, vibration mode, and temporal characteristics, resulting in low tactile feedback accuracy, response delay, and a lack of dynamic weighting of environmental conditions and user behavior preferences in matching real-time signals with historical feedback. The feedback selection is not adaptive enough, and there is a lack of complete recording of tactile output, user behavior, and physiological reactions during execution, making it impossible to form a data accumulation and strategy optimization mechanism. The generation of new tactile states and database updates rely heavily on manual or fixed rules, lacking automation, multimodal processing, and dynamic priority management, making it difficult to achieve personalization and long-term optimization. Therefore, existing technologies suffer from low accuracy, poor adaptability, and lagging strategy updates. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an electronic skin sensing system and method for animal bionic interaction, which has the advantages of improving interaction accuracy and efficiency, and solves the problems mentioned in the background technology.

[0004] To achieve the aforementioned goals of improving interaction accuracy and efficiency, this invention provides the following technical solution: an electronic skin sensing method for animal bionic interaction, comprising the following steps: Tactile signals with different contact intensities, directions, and vibration spectra are collected, and the signals are encoded into an indexable set of feedback states through offline batch processing to form a basic feedback database; The real-time acquired tactile signal vectors are mapped to the basic feedback database through a matching algorithm. Multi-dimensional interval judgment is performed by combining signal amplitude, frequency spectrum characteristics and environmental weights, and candidate feedback states are quickly determined according to policy rules. Based on the candidate feedback states, the system records the current tactile signal characteristics, execution patterns, and changes in user behavior and emotions. After preprocessing the recorded data under preset trigger conditions, the data is uploaded to the central processing platform to generate new feedback states. By integrating new feedback scenarios into the basic feedback library and combining historical user data with current environmental conditions, the probability and success rate of using different feedback scenarios are statistically analyzed to construct a set of pre- and post-feedback feedback scenarios. The pre- and post-feedback posture sets are combined and multimodal validation is performed based on recently collected tactile execution records and user behavior and physiological response data. The execution strategy is continuously optimized and fine-tuned to generate an updated pre-feedback posture strategy.

[0005] The preferred process for forming the basic feedback database is as follows: Multi-channel piezoelectric and resistive sensing units are deployed on the surface of electronic skin to simultaneously collect multi-dimensional characteristics of tactile signals, including contact intensity, direction, vibration spectrum, contact area, and coefficient of friction. The acquired signals undergo time synchronization, anomaly removal, frequency spectrum filtering, and amplitude normalization. By employing a high-dimensional vector coding algorithm and locality-sensitive hash index, the processed tactile signals are mapped to a fast-searchable index space, generating a standardized, multimodal basic feedback database for real-time tactile signal matching, candidate feedback state generation, and abnormal state identification.

[0006] Preferably, the process of mapping the real-time acquired tactile signal vectors to the basic feedback database using a matching algorithm is as follows: Multi-dimensional feature extraction, including amplitude distribution, spectral composition, vibration mode, and temporal variation, is performed on the real-time acquired tactile signal vector to generate a complete real-time feature vector; The real-time feature vector is compared with the feature vectors of each feedback situation in the basic feedback database. The similarity is calculated by dimension-wise feature matching to identify the feedback situation that is closest to the current signal and generate a set of candidate feedback situations. By combining haptic feedback environment information, user behavior preferences, and the context of the current interaction task, the states in the candidate set are sorted according to their matching degree and applicability, and a subset of candidate feedback states to be executed with priority is output.

[0007] Preferably, the process of multi-dimensional interval judgment by combining signal amplitude, frequency spectrum characteristics, and environmental weights is as follows: The candidate feedback situation subset is compared with the real-time tactile signal vector in multiple dimensions, including amplitude range matching degree, spectral similarity, vibration mode correlation and time sequence offset alignment; A multi-dimensional environmental weighting factor is introduced based on ambient light, temperature, material reflection characteristics, and tactile surface friction characteristics to normalize and penalize the matching degree. By using threshold determination rules and multi-level confidence fusion mechanism, all feedback states with matching degree exceeding the adaptive threshold are screened out, and a weighted candidate feedback state set is generated.

[0008] Preferably, the process of quickly determining the candidate feedback situation based on the strategy rules is as follows: Based on a weighted set of candidate feedback states, and combining user behavior history, current task objectives, and security constraints, a multi-objective optimization algorithm is used to calculate the priority score of the candidate feedback states. The policy rule base is invoked, and power consumption limits, response latency, and execution security boundaries are combined to filter candidate situations with high scores and generate the final candidate feedback situation.

[0009] Preferably, the process of recording current tactile signal characteristics, execution patterns, and changes in user behavior and emotions is as follows: The final candidate feedback situation execution data, including contact pressure, vibration intensity, tangential force, and response time, are collected in parallel by an electronic skin controller and embedded sensors. Real-time acquisition of user behavior and physiological signals, including hand movement trajectory, electromyography, heart rate, skin conductance response, and facial expression indicators, and synchronous storage with tactile execution data to form a multimodal execution record; The multimodal execution record is bound to the candidate feedback situation list to generate a record data package.

[0010] Preferably, when the recorded data packet reaches a preset time window, the tactile execution data, user behavior data, and physiological response signals contained in the data packet are subjected to anomaly removal, normalization, and feature enhancement to obtain a clean feature vector; Clean feature vectors are transmitted to the central processing platform using encryption. By combining historical feedback trends, historical execution records, and behavioral trends, the uploaded data is compared and analyzed to generate new feedback trends.

[0011] Preferably, the process of constructing the pre- and post-feedback situation sets is as follows: The new feedback situation is combined with all historical feedback situations in the basic feedback database to form a complete set of feedback situations; A multidimensional comparative analysis was conducted on each feedback state in the set, including tactile signal amplitude, spectral composition, vibration mode, and temporal variation characteristics. By combining historical user data, the response performance of each situation under different interaction scenarios is statistically analyzed, including user response time, satisfaction score, frequency of repeated use, and consistency of physiological signals, to generate a situation performance score; Based on the current environmental conditions, user behavior preferences, and task context, assign dynamic weights to each feedback state; Based on the situation performance score and environmental weight, the feedback situations in the entire set are divided to construct pre-feedback and post-feedback situation sets.

[0012] Preferably, the process of generating the updated forward feedback situational strategy is as follows: The pre- and post-feedback situation sets are then aligned with recently collected tactile execution records, user behavior data, and physiological response signals for feature alignment and trend analysis. Based on the alignment and analysis results, a multi-factor weighted evaluation method is used to comprehensively score the execution order, tactile intensity, spectral parameters, and triggering conditions of the feedback situation, and to evaluate the applicability and priority of each situation under different environmental conditions and user behaviors. Fine-tuning was performed on low-scoring situations, including adjusting tactile intensity, spectral parameters, execution latency, and sequence rearrangement. By combining the triggering relationships and dependencies between the preceding and following situations, an updated preceding feedback situation strategy is generated.

[0013] An electronic skin sensing system for animal bionic interaction, comprising: Data acquisition module: Collects tactile signals with different contact intensities, directions, and vibration spectra from the multi-channel tactile sensing units on the surface of the electronic skin, and performs batch processing to form a basic feedback database; Signal matching module: compares the real-time acquired tactile signal vectors with the situation vectors in the basic feedback database using multi-dimensional features and performs environmental weighting to generate priority candidate feedback situations; Execution Record Module: Collects tactile signals of execution status, user behavior and emotional changes in real time, processes and uploads the records under trigger conditions, and generates new feedback status; Situation building module: Combines new feedback situations with historical feedback databases, calculates usage probability and success rate, and builds sets of pre- and post-feedback situations; Strategy optimization module: Based on the set of pre- and post-feedback situations and recent multimodal execution records, the execution order, tactile intensity and spectral parameters are verified and fine-tuned to generate an updated pre-feedback situation strategy.

[0014] Compared with the prior art, the present invention provides an electronic skin sensing system and method for animal bionic interaction, which has the following beneficial effects: This invention establishes an indexable basic feedback database through multi-channel tactile signal acquisition and offline batch encoding, achieving systematic management and rapid retrieval of complex tactile information. This significantly improves the response accuracy and real-time performance in animal bionic interaction. By performing multi-dimensional matching and environment-weighted judgment between real-time tactile signals and database states, and combining this with policy rules to generate candidate feedback states, dynamic and adaptive tactile feedback selection is achieved. This addresses the problems of single tactile response and lack of environmental perception and task context association in existing technologies. In actual execution, tactile output, user behavior, and physiological responses are recorded in real time, and the data is processed and uploaded to generate... The new feedback posture enables continuous learning and posture updates driven by multimodal data. Furthermore, by statistically analyzing historical user data and current environmental conditions, a set of pre- and post-feedback postures is constructed. Based on multimodal verification and strategy optimization, the execution order, tactile intensity, and spectral parameters are fine-tuned to form an updated pre-feedback posture strategy. This achieves adaptive optimization of tactile feedback and continuous improvement of user experience. Technically, this invention overcomes the problems of existing electronic skin systems having single tactile feedback and difficulty in dynamically adjusting to user behavior and environment, and provides a biomimetic interaction solution with high intelligence, scalability, and adaptability. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method of the present invention; Figure 2 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: Please refer to Figure 1 As shown in the embodiment of the present invention, an electronic skin sensing method for animal bionic interaction includes the following steps: S1: Collect tactile signals with different contact intensities, directions, and vibration spectra, and encode the signals into an indexable set of feedback states through offline batch processing to form a basic feedback database.

[0018] The process of forming the basic feedback database in S1 is as follows: Multi-channel piezoelectric and resistive sensing units are deployed on the surface of electronic skin to simultaneously collect multi-dimensional characteristics of tactile signals, including contact intensity, direction, vibration spectrum, contact area, and coefficient of friction. Multi-channel piezoelectric and resistive sensing units are arranged in a designed spatial layout on the surface of the electronic skin. Each unit independently collects local contact information. The piezoelectric sensor is used to detect dynamic tactile changes, including contact intensity and vibration spectrum, while the resistive sensor is used to measure static pressure distribution, contact area, and coefficient of friction. The acquisition system adopts synchronous sampling control and marks the signals of each channel with a unified timestamp. The sampling frequency can be set between 100 Hz and 1 kHz according to material and application requirements to ensure the capture of rapid tactile changes. At the same time, parallel data acquisition is achieved through a multi-channel hardware interface to ensure that the tactile information of the entire electronic skin surface is fully recorded.

[0019] The acquired signals undergo time synchronization, anomaly removal, frequency spectrum filtering, and amplitude normalization. The acquired multi-channel tactile signals undergo time synchronization processing to ensure that data from different channels are aligned at the same time point, avoiding signal errors caused by sampling deviations. Abnormal signals, including zero drift, sudden noise, and missing acquisitions, are removed through sliding window anomaly detection and threshold filtering methods. The signals are then filtered by frequency spectrum to remove high-frequency interference and low-frequency drift while maintaining the integrity of tactile dynamic characteristics. Amplitude normalization processes map the signals from each channel to a unified range, ensuring cross-channel comparability and standardization. When necessary, sensor drift compensation algorithms and temperature correction curves are combined to correct for the effects of long-term sensor drift and environmental changes, ensuring long-term data stability and reliability.

[0020] A high-dimensional vector coding algorithm and locality-sensitive hash index are used to map the processed tactile signals to a fast-searchable index space, generating a standardized, multimodal basic feedback database for real-time tactile signal matching, candidate feedback state generation, and abnormal state recognition. The processed tactile signals are mapped to a numerical feature space using a high-dimensional vector encoding algorithm. Each tactile event generates a multi-dimensional feature vector containing contact intensity, direction, vibration spectrum, contact area, friction coefficient, and derived features such as instantaneous power spectral density and local pressure gradient. To accelerate retrieval and matching, a locality-sensitive hash index is used to construct an index table, mapping the high-dimensional vector to a low-dimensional hash bucket to achieve near-nearest neighbor fast search. The index space design considers the dynamic range and distribution characteristics of the tactile signals, allowing for rapid location of similar feature vectors in complex interaction scenarios. It supports real-time signal matching and candidate feedback situation retrieval, while providing benchmark data for abnormal tactile situation identification. The indexed high-dimensional vectors are organized according to tactile event type and acquisition conditions to form a standardized, multi-modal basic feedback database that supports multi-level access, including indexes classified by time series, tactile intensity range, and environmental conditions, as well as an incremental storage mechanism that supports real-time updates.

[0021] S2: The real-time acquired tactile signal vectors are mapped to the basic feedback database through a matching algorithm. Multi-dimensional interval judgment is performed by combining signal amplitude, frequency spectrum characteristics and environmental weights, and candidate feedback states are quickly determined according to the strategy rules.

[0022] In S2, the process of mapping the real-time acquired tactile signal vectors to the basic feedback database using a matching algorithm is as follows: Multi-dimensional feature extraction, including amplitude distribution, spectral composition, vibration mode, and temporal variation, is performed on the real-time acquired tactile signal vector to generate a complete real-time feature vector; In the real-time acquisition phase of the electronic skin, a comprehensive analysis of the tactile signal is performed to extract key features such as amplitude distribution, spectral composition, vibration modes, and temporal variations. The amplitude distribution is generated by normalizing and statistically analyzing the pressure or contact intensity signal to produce an interval histogram, reflecting the local force intensity and distribution uniformity. The spectral composition uses Fast Fourier Transform to map the time-domain signal to the frequency domain, obtaining the energy distribution of each frequency band and capturing different contact and vibration characteristics. The vibration modes are generated by analyzing the changes in the spectrum over time to produce short-time energy spectra and peak frequency sequences, reflecting the dynamic tactile response. The temporal variations are calculated using sliding window and time series analysis methods to calculate the instantaneous rate of change and short-time trend of the tactile signal, thus characterizing the dynamic characteristics of the continuous interaction process. All features are merged to form a complete real-time feature vector, providing a multi-dimensional reference basis for matching.

[0023] The real-time feature vector is compared with the feature vectors of each feedback situation in the basic feedback database. The similarity is calculated by dimension-wise feature matching to identify the feedback situation that is closest to the current signal and generate a set of candidate feedback situations. The real-time feature vectors are compared with the feature vectors of each feedback situation stored in the basic feedback database in multiple dimensions. The feature similarity is calculated by dimension-wise, including amplitude interval overlap, spectrum peak matching degree, vibration mode consistency and temporal similarity score. The overall matching degree value of each feedback situation is generated by weighted fusion. In order to improve matching efficiency, the feature vectors are indexed and a high-dimensional vector fast retrieval algorithm is used to quickly locate candidate situations in the database. At the same time, the matching position, confidence score and historical usage data are recorded for each candidate situation to support sorting and strategy selection.

[0024] Combining haptic feedback environment information, user behavior preferences, and the contextual conditions of the current interaction task, the candidate situations are sorted according to their matching degree and applicability, and a subset of candidate feedback situations to be executed with priority is output. After obtaining the matching results, the candidate feedback situation set is sorted in a multi-dimensional weighted manner by combining tactile acquisition environmental information, user behavior preferences, and the current interaction task context. Environmental information includes temperature and humidity, contact surface material, interaction force, and system latency parameters. User behavior preferences are weighted by statistical analysis of historical interaction data. The current task context indicates the operation stage, target response requirements, and safety constraints. The sorting algorithm weights and integrates matching degree, environmental adaptability, and user preferences to generate a priority candidate feedback situation subset. The output subset provides the execution module with a directly usable feedback reference, while retaining the complete candidate set for dynamic adjustment and adaptive optimization to achieve high-precision, low-latency real-time interactive response.

[0025] The process of multidimensional interval judgment in S2, which combines signal amplitude, frequency spectrum characteristics, and environmental weights, is as follows: The candidate feedback situation subset is compared with the real-time tactile signal vector in multiple dimensions, including amplitude range matching degree, spectral similarity, vibration mode correlation and time sequence offset alignment; The candidate feedback state subset is compared dimension-by-dimensionally with the real-time acquired tactile signal vector. The amplitude distribution, spectral composition, vibration mode, and temporal variation of the tactile signal are mapped to the characteristics of each candidate feedback state. The amplitude interval matching degree is obtained by calculating the overlap rate of the pressure intervals of each contact point in the real-time signal and the candidate state, which can reflect the consistency of local tactile intensity. The spectral similarity is calculated by using fast Fourier transform to calculate the similarity coefficient of frequency band energy distribution, which characterizes the frequency response matching degree of different tactile events. The vibration mode correlation is evaluated by short-time energy spectrum analysis and peak frequency comparison to assess the vibration consistency between the candidate state and the real-time signal. The temporal offset alignment uses a dynamic time warping algorithm to align the short-time changes of the real-time signal with the feature sequence of the candidate state, identify possible delayed or premature responses, and thus form a multi-dimensional matching index vector, providing basic data for weighted fusion.

[0026] A multi-dimensional environmental weighting factor is introduced based on ambient light, temperature, material reflection characteristics, and tactile surface friction characteristics to normalize and penalize the matching degree. Based on multidimensional matching, multiple environmental parameters such as ambient light, temperature, material reflection characteristics, and tactile surface friction characteristics are introduced as weighting factors to normalize and correct the matching degree. Changes in light may affect the readings of optical tactile sensors, and errors are reduced through linear normalization. Temperature changes affect material softness and sensor response speed, and tactile amplitude offset is corrected by exponential smoothing. Surface material reflection and friction characteristics can change vibration transmission and sliding friction noise, and environmental adaptability coefficients are used to correct the matching degree of spectrum and vibration mode. The matching index vector after multidimensional weighting considers both real-time tactile characteristics and environmental adaptability, making the matching results more robust and reusable.

[0027] By using threshold determination rules and multi-level confidence fusion mechanism, all feedback states with matching degrees exceeding the adaptive threshold are screened out, generating a weighted candidate feedback state set; By utilizing threshold determination rules and a multi-layer confidence fusion mechanism, the weighted matching index vector is processed, including local confidence assessment and global confidence weighting. The results of amplitude matching, spectral similarity, vibration mode consistency, and environmental weighting are comprehensively calculated to generate an overall matching score for each feedback state. Candidate states are selected according to an adaptive threshold rule. The threshold is automatically adjusted through historical interaction data statistics to balance system response accuracy and interaction security. All feedback states with scores exceeding the threshold form a weighted candidate feedback state set. This set not only retains the matching distribution and environmental adaptation information of each candidate state, but also forms a standardized data structure that can be directly input into the selection of pre-feedback strategies.

[0028] The process of quickly determining candidate feedback states based on policy rules in S2 is as follows: Based on a weighted set of candidate feedback states, and combining user behavior history, current task objectives, and security constraints, a multi-objective optimization algorithm is used to calculate the priority score of the candidate feedback states. Based on a weighted set of candidate feedback states, each candidate state includes information such as amplitude matching, spectral similarity, vibration mode consistency, environmental adaptability, and local and global confidence scores. For each candidate state, it is compared with the user's historical interaction data, including the user's past acceptance of tactile feedback, operation preferences, and response delay records. At the same time, the objectives of the current interaction task are input into the system, such as task operation type, precise tactile requirements, and operation rhythm. Combined with safety constraints, including human comfort thresholds, overload limits, and interaction boundary conditions, a multi-objective optimization algorithm is used to comprehensively score the candidate states, calculating the priority of each candidate state in terms of safety, matching degree, response efficiency, and user preference.

[0029] The policy rule base is invoked, and combined with power consumption limits, response latency, and execution security boundaries, high-scoring candidate situations are filtered to generate the final candidate feedback situation; After priority scoring is completed, candidate feedback states are input into the policy rule base. The rule base includes power consumption limits, execution delay thresholds, feedback intensity boundaries, and operational safety constraints. Each candidate state with a high score is validated according to the rules, including checking whether it can complete execution within a specified time, whether the power consumption meets the system's real-time limits, and whether the feedback mode exceeds the safety boundary. Candidate states that meet all constraints are selected through rule matching, and the score and rule matching results of each candidate state are recorded. After priority scoring and rule filtering, final candidate feedback states are generated from the weighted candidate set. These final candidate states have a clear execution priority order and contain complete matching information, environmental adaptation data, and confidence scores. The final candidate states can be directly used as input for the selection of pre-feedback strategies to achieve fast, accurate, and safe tactile feedback responses.

[0030] S3: Based on the execution of candidate feedback states, record the current tactile signal characteristics, execution mode, and changes in user behavior and emotions. After preprocessing the recorded data under preset trigger conditions, upload it to the central processing platform to generate new feedback states.

[0031] The process of recording current tactile signal characteristics, execution patterns, and changes in user behavior and emotions in S3 is as follows: The final candidate feedback situation execution data, including contact pressure, vibration intensity, tangential force, and response time, are collected in parallel by an electronic skin controller and embedded sensors. The electronic skin controller and embedded sensors acquire the final candidate feedback states in parallel. The pressure value of each contact point is recorded in real time by a piezoelectric or resistive sensing unit. Vibration intensity and tangential force are measured by a high-frequency sampling sensor. The response time is accurately recorded to the millisecond level by the controller's built-in timer. The sensor output signal is marked with a synchronous timestamp and undergoes preliminary filtering, noise reduction, and amplitude normalization to ensure the consistency and comparability of cross-channel data. This ensures that the tactile response characteristics of each execution state are captured completely and accurately, providing basic data for multimodal fusion.

[0032] Real-time acquisition of user behavior and physiological signals, including hand movement trajectory, electromyography, heart rate, skin conductance response, and facial expression indicators, and synchronous storage with tactile execution data to form a multimodal execution record; During the synchronous execution of haptic feedback, user behavior and physiological signals are collected in real time. Hand movement trajectories are recorded by a high-precision inertial measurement unit and optical tracking sensors, electromyography signals are acquired through surface electrodes, heart rate and skin conductance responses are collected through a wearable physiological monitoring module, and facial expression indicators are acquired through a camera and facial key point detection algorithm. All signals are processed in time synchronization to form a unified time series with the haptic feedback data, ensuring that user actions, haptic feedback, and physiological state data correspond one-to-one. Multimodal data acquisition ensures the integrity and analyzability of user interaction responses and supports quantitative analysis of behavioral preferences, emotional states, and physiological responses.

[0033] Bind the multimodal execution record to the candidate feedback situation list to generate a record data package; By associating tactile execution data with user behavior and physiological signals, a multimodal execution record data package is generated for each candidate feedback state. The data package contains synchronous time series of physiological signals such as tactile features, execution mode, behavioral trajectory, electromyography, and heart rate. At the same time, environmental conditions such as temperature, light, and operation material information are recorded. Through a structured data format, various signals are bound to candidate states to form a complete execution record, realizing multidimensional information that is traceable, searchable, and reusable.

[0034] The process of generating a new feedback situation in S3 is as follows: When the recorded data packet reaches the preset time window, the tactile execution data, user behavior data, and physiological response signals contained in the data packet are subjected to anomaly removal, normalization, and feature enhancement to obtain a clean feature vector; When the recorded data packet reaches the preset time window, the tactile execution data, user behavior data, and physiological response signals contained in the data packet are preprocessed in batches. The tactile data is noise-removed through time synchronization correction, amplitude normalization, and spectral filtering. The hand movement trajectory and electromyography signal are processed through low-pass filtering, motion baseline correction, and abnormal peak removal. The heart rate and skin conductance response signals are drift-compensated and standardized. Through feature enhancement algorithms, the time series signal is mapped into a high-dimensional feature vector, including amplitude statistical features, frequency domain features, vibration mode features, and behavioral trajectory mode features. At the same time, local mutation information and trend evolution features are extracted to ensure that the data used for feedback situation generation is accurate, stable, and quantifiable.

[0035] Clean feature vectors are transmitted to the central processing platform using encryption. The enhanced data vectors are transmitted via a secure encryption protocol, including end-to-end encryption and data integrity verification, to ensure that tactile execution information, user behavior, and physiological signals are not tampered with or leaked during transmission. After receiving the data, the central processing platform first performs integrity verification and time series alignment on the transmitted data to ensure that the data order is consistent with the acquisition time sequence. At the same time, it establishes a unified data input format for subsequent comparison and analysis, ensuring the reliability and security of multi-source, multi-modal information when transmitted across devices and nodes.

[0036] By combining historical feedback trends, historical execution records, and behavioral trends, the uploaded data is compared and analyzed to generate new feedback trends; On the central processing platform, clean feature vectors are compared and analyzed in multiple dimensions with the historical feedback state database and historical execution records. The comparison includes tactile feature similarity matching, behavioral trajectory trend analysis, and physiological response correlation assessment. At the same time, weights are added based on the evolution trend of long-term user behavior patterns and task context. Through reinforcement learning, incremental learning, and multi-objective optimization algorithms, new feedback states are generated based on changes in feature vectors and historical feedback effects. This ensures that the new feedback is superior to the historical states in terms of naturalness, accuracy, security, and adaptability. The newly generated feedback states are then returned to the basic feedback database for real-time matching and candidate feedback state generation, forming a closed-loop optimization mechanism.

[0037] S4: Integrate the new feedback scenarios into the basic feedback library, combine historical user data and current environmental conditions, statistically analyze the usage probability and success rate of different feedback scenarios, and construct a set of pre- and post-feedback scenarios.

[0038] The process of constructing the pre- and post-feedback situation sets in S4 is as follows: The new feedback situation is combined with all historical feedback situations in the basic feedback database to form a complete set of feedback situations; The newly generated feedback situation is integrated with all historical feedback situations in the basic feedback database to construct a complete set of feedback situations. During the aggregation process, each feedback situation is uniquely identified and its acquisition time, tactile signal characteristics, and execution context information are recorded. Data integration is completed through high-dimensional vector mapping to ensure that the new situation and the historical situation can be directly compared in a unified multi-dimensional feature space, thereby achieving data standardization and retrieval.

[0039] A multidimensional comparative analysis was conducted on each feedback state in the set, including tactile signal amplitude, spectral composition, vibration mode, and temporal variation characteristics. For each feedback state in the complete set, a multidimensional comparative analysis is performed. The amplitude characteristics of the tactile signal are analyzed by statistical indicators to determine its maximum value, mean, and variance. The spectral composition is extracted by Fourier transform to obtain the dominant frequency component and power spectral density characteristics. The vibration modes are classified by short-time energy analysis and pattern recognition algorithms. The temporal variation characteristics are characterized by time series analysis, trend fitting, and local mutation detection. Through this multidimensional analysis, the similarity and feature integrity of each feedback state in different signal dimensions are quantified, providing basic data for performance evaluation.

[0040] By combining historical user data, the response performance of each situation under different interaction scenarios is statistically analyzed, including user response time, satisfaction score, frequency of repeated use, and consistency of physiological signals, to generate a situation performance score; By combining historical user data, statistical analysis is performed on the performance of each feedback state in different interaction scenarios, including user response time, operation efficiency, satisfaction score, frequency of repeated use, and consistency of physiological signals such as electromyography, heart rate, and skin conductance. A state performance score is generated by multi-indicator weighted calculation, with weight allocation referring to the contribution of each indicator to user experience and task completion efficiency in historical data, thereby quantifying the effectiveness and reliability of each feedback state.

[0041] Based on the current environmental conditions, user behavior preferences, and task context, assign dynamic weights to each feedback state; Based on the situation performance score, dynamic weights are assigned to each feedback situation by combining the current environmental conditions, task context, and user behavior preferences. Environmental conditions include temperature, humidity, light intensity, and contact material characteristics. Task context includes operation objectives, operation rhythm, and interaction complexity. User preferences are obtained by modeling historical interaction data. The calculation of dynamic weights adopts a multi-objective optimization method, which combines performance scores with environmental adaptability, so that feedback situations have the ability to be executed preferentially under different conditions.

[0042] Based on the situation performance score and environmental weight, the feedback situations in the entire set are divided to construct pre-feedback and post-feedback situation sets; Based on situation performance scores and dynamic weights, the complete feedback situation set is divided into a pre-feedback situation set and a post-feedback situation set. The pre-feedback situation set includes feedback situations that are triggered first in the current task stage and interaction context to ensure rapid response and security. The post-feedback situation set contains supplementary and backup situations to handle situations where the environment changes or the user deviates from the expected behavior. The division results are stored in the form of structured data for use in strategy selection, dynamic execution and closed-loop optimization, while retaining complete multi-dimensional feature information to support subsequent learning and abnormal situation identification.

[0043] S5: Combine the pre-feedback and post-feedback postures, perform multimodal verification based on recently collected tactile execution records and user behavior and physiological response data, continuously optimize and fine-tune the execution strategy, and generate an updated pre-feedback posture strategy.

[0044] The process of generating the updated forward feedback situational strategy in S5 is as follows: The pre- and post-feedback situation sets are then aligned with recently collected tactile execution records, user behavior data, and physiological response signals for feature alignment and trend analysis. The pre- and post-feedback state sets are aligned with recently collected tactile execution records, user behavior data, and physiological response signals. This includes standardizing the amplitude, spectral composition, vibration mode, and temporal variation characteristics of tactile signals to ensure comparability of data from different sources within the same multidimensional feature space. User behavior data is time-synchronized using motion trajectories, electromyographic signals, and operation timing. Physiological response signals, such as heart rate, skin conductance, and facial expression indicators, are aligned using timestamps. After alignment, sliding window and trend analysis algorithms are used to identify short-term changes, periodic fluctuations, and local mutations, extracting the response trend of each feedback state under the current environment and user state to provide a basis for scoring.

[0045] Based on the alignment and analysis results, a multi-factor weighted evaluation method is used to comprehensively score the execution order, tactile intensity, spectral parameters, and triggering conditions of the feedback situation, and to evaluate the applicability and priority of each situation under different environmental conditions and user behaviors. Based on feature alignment and trend analysis results, a multi-factor weighted evaluation method is used to comprehensively score candidate feedback situations. The scoring factors include execution order priority, tactile intensity amplitude, spectral principal components, and trigger condition sensitivity. At the same time, environmental condition weights, user behavior context, and historical success rate are combined. Each factor is normalized and weighted to ensure that the score can reflect the applicability and execution efficiency of the situation under different users, different operating environments, and multi-task scenarios. The generated comprehensive scoring matrix provides quantifiable priority indicators for each feedback situation, providing a clear basis for strategy updates.

[0046] Fine-tuning was performed on low-scoring situations, including adjusting tactile intensity, spectral parameters, execution latency, and sequence rearrangement. For situations where the overall score is below the adaptive threshold, fine-tuning is performed. Fine-tuning includes adjusting the tactile intensity to match the user's hand sensitivity, optimizing spectral parameters to improve signal recognition reliability, adjusting the execution delay to synchronize the operation rhythm, and rearranging the situation execution order to enhance the overall feedback smoothness. The fine-tuning process is verified through closed-loop simulation or historical data comparison to ensure that the adjusted situation improves the execution success rate and user acceptance while maintaining safety and comfort.

[0047] By combining the triggering relationships and dependencies between the preceding and following situations, an updated preceding feedback situation strategy is generated; Further considering the triggering relationship and dependency conditions between the preceding and following feedback situations, the triggering effect of each preceding situation on the following situation and possible sequence conflicts are analyzed. By establishing a situation dependency graph and execution path matrix, combined with a strategy rule base and multi-objective optimization algorithm, the selection of preceding situations and execution order are optimized, so that the feedback system maintains response consistency and security under different environmental conditions. The finally generated updated preceding feedback situation strategy not only ensures rapid response and personalized adaptation, but also takes into account closed-loop optimization and multi-scenario adaptability.

[0048] Example 2: Please refer to Figure 2 As shown, an electronic skin sensing system for animal bionic interaction includes: Data acquisition module: Collects tactile signals with different contact intensities, directions, and vibration spectra from the multi-channel tactile sensing units on the surface of the electronic skin, and performs batch processing to form a basic feedback database; Signal matching module: compares the real-time acquired tactile signal vectors with the situation vectors in the basic feedback database using multi-dimensional features and performs environmental weighting to generate priority candidate feedback situations; Execution Record Module: Collects tactile signals of execution status, user behavior and emotional changes in real time, processes and uploads the records under trigger conditions, and generates new feedback status; Situation building module: Combines new feedback situations with historical feedback databases, calculates usage probability and success rate, and builds sets of pre- and post-feedback situations; Strategy optimization module: Based on the set of pre- and post-feedback situations and recent multimodal execution records, the execution order, tactile intensity and spectral parameters are verified and fine-tuned to generate an updated pre-feedback situation strategy.

[0049] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0050] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An electronic skin sensing method for animal bionic interaction, characterized in that... This includes the following steps: Tactile signals with different contact intensities, directions, and vibration spectra are collected, and the signals are encoded into an indexable set of feedback states through offline batch processing to form a basic feedback database; The real-time acquired tactile signal vectors are mapped to the basic feedback database through a matching algorithm. Multi-dimensional interval judgment is performed by combining signal amplitude, frequency spectrum characteristics and environmental weights, and candidate feedback states are quickly determined according to policy rules. Based on the candidate feedback states, the system records the current tactile signal characteristics, execution patterns, and changes in user behavior and emotions. After preprocessing the recorded data under preset trigger conditions, the data is uploaded to the central processing platform to generate new feedback states. By integrating new feedback scenarios into the basic feedback library and combining historical user data with current environmental conditions, the probability and success rate of using different feedback scenarios are statistically analyzed to construct a set of pre- and post-feedback feedback scenarios. The pre- and post-feedback posture sets are combined and multimodal validation is performed based on recently collected tactile execution records and user behavior and physiological response data. The execution strategy is continuously optimized and fine-tuned to generate an updated pre-feedback posture strategy.

2. The electronic skin sensing method for animal bionic interaction according to claim 1, characterized in that... The process of forming the basic feedback database is as follows: Multi-channel piezoelectric and resistive sensing units are deployed on the surface of electronic skin to simultaneously collect multi-dimensional characteristics of tactile signals, including contact intensity, direction, vibration spectrum, contact area, and coefficient of friction. The acquired signals undergo time synchronization, anomaly removal, frequency spectrum filtering, and amplitude normalization. By employing a high-dimensional vector coding algorithm and locality-sensitive hash index, the processed tactile signals are mapped to a fast-searchable index space, generating a standardized, multimodal basic feedback database for real-time tactile signal matching, candidate feedback state generation, and abnormal state identification.

3. The electronic skin sensing method for animal bionic interaction according to claim 2, characterized in that... The process of mapping the real-time acquired tactile signal vectors to the basic feedback database using a matching algorithm is as follows: Multi-dimensional feature extraction, including amplitude distribution, spectral composition, vibration mode, and temporal variation, is performed on the real-time acquired tactile signal vector to generate a complete real-time feature vector; The real-time feature vector is compared with the feature vectors of each feedback situation in the basic feedback database. The similarity is calculated by dimension-wise feature matching to identify the feedback situation that is closest to the current signal and generate a set of candidate feedback situations. By combining haptic feedback environment information, user behavior preferences, and the context of the current interaction task, the states in the candidate set are sorted according to their matching degree and applicability, and a subset of candidate feedback states to be executed with priority is output.

4. The electronic skin sensing method for animal bionic interaction according to claim 3, characterized in that... The process of multi-dimensional interval judgment by combining signal amplitude, frequency spectrum characteristics, and environmental weights is as follows: The candidate feedback situation subset is compared with the real-time tactile signal vector in multiple dimensions, including amplitude range matching degree, spectral similarity, vibration mode correlation and time sequence offset alignment; A multi-dimensional environmental weighting factor is introduced based on ambient light, temperature, material reflection characteristics, and tactile surface friction characteristics to normalize and penalize the matching degree. By using threshold determination rules and multi-level confidence fusion mechanism, all feedback states with matching degree exceeding the adaptive threshold are screened out, and a weighted candidate feedback state set is generated.

5. The electronic skin sensing method for animal bionic interaction according to claim 4, characterized in that... The process of quickly determining the candidate feedback situation based on the strategy rules is as follows: Based on a weighted set of candidate feedback states, and combining user behavior history, current task objectives, and security constraints, a multi-objective optimization algorithm is used to calculate the priority score of the candidate feedback states. The policy rule base is invoked, and power consumption limits, response latency, and execution security boundaries are combined to filter candidate situations with high scores and generate the final candidate feedback situation.

6. The electronic skin sensing method for animal bionic interaction according to claim 5, characterized in that... The process of recording current tactile signal characteristics, execution patterns, and changes in user behavior and emotions is as follows: The final candidate feedback situation execution data, including contact pressure, vibration intensity, tangential force, and response time, are collected in parallel by an electronic skin controller and embedded sensors. Real-time acquisition of user behavior and physiological signals, including hand movement trajectory, electromyography, heart rate, skin conductance response, and facial expression indicators, and synchronous storage with tactile execution data to form a multimodal execution record; The multimodal execution record is bound to the candidate feedback situation list to generate a record data package.

7. The electronic skin sensing method for animal bionic interaction according to claim 6, characterized in that... The process of generating a new feedback situation is as follows: When the recorded data packet reaches the preset time window, the tactile execution data, user behavior data, and physiological response signals contained in the data packet are subjected to anomaly removal, normalization, and feature enhancement to obtain a clean feature vector; Clean feature vectors are transmitted to the central processing platform using encryption. By combining historical feedback trends, historical execution records, and behavioral trends, the uploaded data is compared and analyzed to generate new feedback trends.

8. The electronic skin sensing method for animal bionic interaction according to claim 7, characterized in that... The process of constructing the pre- and post-feedback situation sets is as follows: The new feedback situation is combined with all historical feedback situations in the basic feedback database to form a complete set of feedback situations; A multidimensional comparative analysis was conducted on each feedback state in the set, including tactile signal amplitude, spectral composition, vibration mode, and temporal variation characteristics. By combining historical user data, the response performance of each situation under different interaction scenarios is statistically analyzed, including user response time, satisfaction score, frequency of repeated use, and consistency of physiological signals, to generate a situation performance score; Based on the current environmental conditions, user behavior preferences, and task context, assign dynamic weights to each feedback state; Based on the situation performance score and environmental weight, the feedback situations in the entire set are divided to construct pre-feedback and post-feedback situation sets.

9. The electronic skin sensing method for animal bionic interaction according to claim 8, characterized in that... The process of generating the updated forward feedback situational strategy is as follows: The pre- and post-feedback situation sets are then aligned with recently collected tactile execution records, user behavior data, and physiological response signals for feature alignment and trend analysis. Based on the alignment and analysis results, a multi-factor weighted evaluation method is used to comprehensively score the execution order, tactile intensity, spectral parameters, and triggering conditions of the feedback situation, and to evaluate the applicability and priority of each situation under different environmental conditions and user behaviors. Fine-tuning was performed on low-scoring situations, including adjusting tactile intensity, spectral parameters, execution latency, and sequence rearrangement. By combining the triggering relationships and dependencies between the preceding and following situations, an updated preceding feedback situation strategy is generated.

10. An electronic skin sensing system for animal bionic interaction, applied to the method described in any one of claims 1-9, characterized in that... ,include: Data acquisition module: Collects tactile signals with different contact intensities, directions, and vibration spectra from the multi-channel tactile sensing units on the surface of the electronic skin, and performs batch processing to form a basic feedback database; Signal matching module: compares the real-time acquired tactile signal vectors with the situation vectors in the basic feedback database using multi-dimensional features and performs environmental weighting to generate priority candidate feedback situations; Execution Record Module: Collects tactile signals of execution status, user behavior and emotional changes in real time, processes and uploads the records under trigger conditions, and generates new feedback status; Situation building module: Combines new feedback situations with historical feedback databases, calculates usage probability and success rate, and builds sets of pre- and post-feedback situations; Strategy optimization module: Based on the set of pre- and post-feedback situations and recent multimodal execution records, the execution order, tactile intensity and spectral parameters are verified and fine-tuned to generate an updated pre-feedback situation strategy.

Images