A companion device-based emotional interaction standardization adaptation method

Abstract

The application discloses a kind of based on the standardization adaptation method of mood interaction of companion device, comprising the following steps: obtaining the multimodal data of companion device, and standardization processing is generated adaptation state parameter;Identify current mood state, determine corresponding role quota constraint;Identify current device response capability, determine the coupling constraint relationship between each device capability unit;Determine the proposal rule, acceptance rule and update rule of current round stable matching, form the basis of current round stable matching;Using Gale-Shapley stable matching algorithm executes multiple-to-one stable matching, obtains current round stable matching result;Around current round stable matching result, determine local broken stable area;For local broken stable area, again using Gale-Shapley stable matching algorithm executes local reconfiguration, generates target mood interaction adaptation result.The application improves the accuracy, continuity and coordination of mood interaction response.

Description

Technical Field

[0001] This invention relates to the field of emotion interaction adaptation technology, and in particular to a standardized adaptation method for emotion interaction based on companion devices. Background Technology

[0002] With the development of artificial intelligence, emotion computing, and smart companionship terminal technologies, companionship devices have been gradually applied to scenarios such as home care, child companionship, elderly care, emotional comfort, and interactive assistance. Existing companionship devices typically collect user voice, image, text, and behavioral information during interactions, and determine the user's emotions based on a preset recognition model. The device then outputs voice responses, screen displays, light feedback, action execution, or haptic prompts according to the recognition results, thereby achieving basic emotional interaction. Meanwhile, some existing technologies have begun to explore the fusion of multimodal data for emotion recognition and, combined with device functional configurations, generate corresponding interactive actions to improve the responsiveness of companionship devices to users' emotional states.

[0003] However, most existing technologies still rely on a direct mapping approach of "emotion recognition - response triggering," typically focusing on improving emotion recognition accuracy or enriching device feedback formats, lacking a unified adaptation mechanism for the heterogeneous capabilities of companion devices. On the one hand, different companion devices exhibit significant differences in voice, display, lighting, motion, tactile, and network capabilities. The execution of the same emotional state across different devices lacks a unified standard, easily leading to fragmented interaction logic, disjointed response formats, and inconsistent adaptation results. On the other hand, existing solutions typically treat device response capabilities as independent output resources, rarely considering the shared, alternative, prohibited coexistence, and priority / exclusivity relationships between different device capabilities, making it difficult to achieve stable and coordinated emotional interaction output under complex coupling conditions.

[0004] Therefore, how to provide a standardized adaptation method for emotional interaction based on companion devices is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a standardized adaptation method for emotional interaction based on companion devices. This invention constructs a many-to-one stable matching mechanism between standard emotional roles and device capability units, and combines it with continued matching, local instability identification, and local rematch processing to achieve standardized adaptation, stable carrying, and dynamic repair in the emotional interaction process of companion devices, thereby improving the accuracy, continuity, and coordination of emotional interaction responses.

[0006] A method for standardizing and adapting emotional interaction based on companion devices according to an embodiment of the present invention includes the following steps: Acquire multimodal data of companion devices during interaction, perform standardization processing, and generate adaptive state parameters; The system identifies the current emotional state based on the adaptation state parameters, generates standardized emotional semantic tags, divides the current round of emotional interaction needs, and determines the corresponding role quota constraints. Identify the current device response capability based on the adaptation status parameters, divide the device capability units that can participate in matching, and determine the coupling constraint relationship between each device capability unit; Based on role quota constraints and coupling constraints, determine the proposal rules, acceptance rules and update rules for stable matching in the current round, forming the basis for stable matching in the current round; Based on the current round of stable matching, the Gale-Shapley stable matching algorithm is used to perform many-to-one stable matching. After the device capability unit accepts the proposal for the standard emotion role, the occupancy propagation operator is triggered to adjust the matching state of the associated device capability unit and obtain the current round of stable matching result. Based on the current round of stable matching results, the continued occupation matching structure is imported, and the comprehensive instability quantity is calculated in combination with the adaptive state parameters. The local instability region is determined according to the comparison results between the comprehensive instability quantity and the instability threshold. For the locally unstable regions, the Gale-Shapley stable matching algorithm is used again to perform local rematching, update the stable matching results of the current round of role carrying, and generate the target emotion interaction adaptation results.

[0007] Optionally, obtaining the adaptation state parameters specifically includes: Acquire multimodal data from companion devices during interaction, and perform source-specific collection, time stamping, and raw data archiving. Associate the collection description fields to form a raw multimodal data set with a unified time identifier. The original multimodal dataset is processed by time alignment, missing data completion, anomaly removal, unit unification and standardization. The multimodal data is mapped to the corresponding time window, and mean normalization and standard deviation normalization are performed to generate a standardized modal feature set. Adaptive representation parameters are extracted based on a standardized modal feature set, combined, and the combination results are processed by range mapping to generate adaptive state parameters.

[0008] Optionally, determining the corresponding role quota constraint specifically includes: Obtain the adaptation state parameters, load the adaptation state parameters in a fixed order, and perform missing item blocking, out-of-bounds interception and conflict resolution in sequence. Write the retained results as the sentiment discrimination base value. Read the set of emotion categories stored in the companion device, perform matching calculation and strength comparison for each emotion category in the set based on the emotion discrimination base value, determine the emotion category with the largest matching result as the current emotion state, and generate the current emotion state recognition result; Read the current emotion state recognition result and the adaptation state parameters, perform intensity grading and trend orientation on the current emotion state recognition result, and write and encapsulate the obtained emotion category, intensity level and trend identifier according to a unified label format to generate standardized emotion semantic labels; Read standardized emotion semantic tags, call the demand role mapping table stored in companion devices, perform demand mapping and role placement on standardized emotion semantic tags, determine the current round of emotion interaction demand, and generate role quota constraints according to the lower limit, upper limit and prohibition restrictions corresponding to the current round of emotion interaction demand.

[0009] Optionally, determining the coupling constraint relationships between the various equipment capability units specifically includes: The device operation status of the companion device in the current interaction cycle is obtained. The status of the device response capability item is read in combination with the adaptation status parameters. The enable status judgment, occupancy status judgment and availability status judgment are performed on each device response capability item in a fixed reading order to generate a set of device capability status. Read the device capability status set, perform response eligibility judgment on each device response capability item, retain the device response capability items that pass the judgment, and write the retention results into the device response capability set; Read the set of device response capabilities, perform capability splitting, unit division and unit registration on the set of device response capabilities according to the bearing function, divide the set of device response capabilities into a set of device capability units that can participate in matching, and write the bearing category, bearing capacity and occupancy status for each device capability unit respectively. Read the set of equipment capability units, perform pairwise relationship discrimination on any two equipment capability units in the set, generate corresponding relationship records in the order of cooperative discrimination, substitution discrimination, mutual exclusion discrimination and overriding discrimination, and determine the coupling constraint relationship between each equipment capability unit based on the corresponding relationship records.

[0010] Optionally, the criteria for obtaining the current round of stable matching specifically include: Read the role quota constraints, perform proposal qualification sorting on the role set, perform acceptance qualification sorting on the equipment capability unit set, and generate the current round candidate role sequence and the current round candidate equipment capability unit sequence; Read the current round candidate role sequence and the current round candidate device capability unit sequence, and sequentially point each role to the corresponding device capability unit to initiate a bearer request. If the request is not accepted, move to the next device capability unit to continue initiating a bearer request, and generate a proposal record. Read the proposal record and coupling constraint relationship, perform comparison and acceptance processing on the device capability unit that received the role proposal, and generate an acceptance record; Read the acceptance record, role quota constraints, and coupling constraints; increase the current number of accepted roles and the current number of accepted equipment capacity units occupied; rewrite the participation status of equipment capacity units affected by the acceptance result; and generate the basis for stable matching in the current round.

[0011] Optionally, obtaining the stable matching result of the current round specifically includes: Read the current round of stable matching criteria, write the standard emotion roles whose current number of loads is less than the maximum number of loads into the proposed sequence, write the device capability units whose current number of loads is less than the capacity and are in an acceptable state into the accept sequence, and allocate the corresponding number of load positions to each device capability unit according to the capacity, and generate the current round of delayed acceptance initial state. The standard sentiment roles in the proposed sequence are proposed to the target device capability unit in the order of proposal. The target device capability unit performs delayed acceptance processing according to the Gale-Shapley stable matching algorithm. For newly arrived standard sentiment roles, a temporary holding comparison is performed first. If there are free bearer slots in the target device capability unit, the newly arrived standard sentiment roles are temporarily held in the free bearer slots. If there are no free bearer slots in the target device capability unit, the newly arrived standard sentiment roles are compared with the standard sentiment roles that have been temporarily held in each bearer slot in the target device capability unit. The standard sentiment roles with higher role carrying priority continue to occupy the corresponding bearer slots, and the standard sentiment roles with lower role carrying priority are released from the corresponding bearer slots, generating a temporary holding acceptance result for a single device capability unit. After a single device capability unit forms a temporary acceptance result, the occupancy propagation operator is immediately triggered. The temporary acceptance result of the single device capability unit is propagated to the associated device capability units along the coupling constraint relationship. The associated device capability units are then subjected to matchable domain compression processing. Standard emotion roles that are prohibited from being carried together with the current temporary acceptance result are deleted, while standard emotion roles that are allowed to be carried together with the current temporary acceptance result are retained. The number of available bearer bits of associated device capability units that have a bearer bit competition relationship with the current temporary acceptance result is reduced. The associated device capability units whose bearer functions have been covered by the current temporary acceptance result are rewritten to an unacceptable state, and the propagated associated device capability unit matching state is generated. For standard sentiment roles released in the bearer contention adjudication, a backtracking and redirection process is performed. Based on the matching status of the associated device capability units after propagation, device capability units that still retain the corresponding bearer category and still have available bearer bits are re-screened. The released standard sentiment roles are then directed to the next device capability unit after rescreening to continue initiating proposals. The process of proposal, delayed acceptance, bearer contention adjudication, occupancy propagation operator triggering, and backtracking and redirection is repeated until there are no standard sentiment roles in the proposal sequence that can continue to initiate proposals, or no device capability units in the acceptance sequence that can continue to accept proposals. The final temporary retention result of each bearer bit within each device capability unit is written as the stable matching result of the current round.

[0012] Optionally, the determination of the locally unstable region specifically includes: Extract all role-bearing pairs from the current round of stable matching results, and perform renewal registration for each role-bearing pair. Write the standard emotion role, device capability unit, bearer position, and the corresponding adaptation state parameters when forming the current round of stable matching results into the corresponding renewal unit to generate a renewal matching structure. The continued occupation matching structure is sequentially subjected to continued occupation retention determination. The difference between the current adaptation status parameter and the adaptation status parameter registered by the corresponding continued occupation unit is compared. The comprehensive instability amount is calculated in combination with the role quota constraint and the current occupation status of the equipment capacity unit. When the comprehensive instability amount is less than the instability threshold, the continued occupation status of the corresponding role bearing pair is retained. When the comprehensive instability amount is not less than the instability threshold, the corresponding role bearing pair is written as the instability candidate bearing pair. For the candidate bearer pairs that are to break stability, perform boundary stability propagation judgment. Check whether the role bearer pairs directly adjacent to the candidate bearer pairs along the coupling constraint relationship have lost their continued occupation state. Continue to write the adjacent role bearer pairs that meet the propagation conditions as candidate bearer pairs that are to break stability. Repeat the adjacent check and candidate writing to generate a stability propagation chain. The unstable transmission chain is partially closed and defined. The unstable transmission chain that is completely de-occupied inside and surrounded by role carriers that maintain the state of continuous occupation outside is written as a local unstable region. The role carriers that maintain the state of continuous occupation outside the local unstable region and are directly adjacent to the local unstable region are written as boundary carrier pairs, thus generating a local unstable region.

[0013] Optionally, obtaining the target emotion interaction adaptation result specifically includes: Screen out the role bearer pairs to be released from the locally unstable region, perform bearer release and gap registration on the role bearer pairs to be released, write the original equipment capacity unit bearer bits occupied by the role bearer pairs to be released as reconfiguration gaps, and write the equipment capacity units corresponding to the boundary bearer pairs to be boundary lock bits, and generate a local reconfiguration boundary set; Read the local reconfiguration boundary set, filter out the standard sentiment roles that are allowed to participate in local reconfiguration and write them into the local proposal sequence, filter out the equipment capability units that are allowed to participate in gap filling and write them into the local acceptance sequence, and restrict the standard sentiment roles in the local proposal sequence to only make proposals to the equipment capability units corresponding to the reconfiguration gap; The standard emotion roles in the local proposed sequence are proposed to the device capability unit corresponding to the rematch gap in order of proposal order. The device capability unit that receives the proposal re-executes the local delayed acceptance processing according to the Gale-Shapley stable matching algorithm, performs local matchable domain compression and gap redirection according to the coupling constraint relationship, until a local rematch result is generated, and the local rematch result is written back to the current round of role-bearing stable matching result to generate the target emotion interaction adaptation result.

[0014] The beneficial effects of this invention are: This application divides the responsiveness of companion devices into matching device capability units and further determines the coupling constraints between each device capability unit. This makes the device no longer participate in the interaction as a simple overall execution terminal, but is refined into a capability structure that can be carried, replaced, restricted, and linked. This allows for a more accurate reflection of the relationship between voice carrying, display carrying, light carrying, motion carrying, tactile carrying, and network carrying, thereby improving the accuracy of emotional interaction adaptation and output coordination under complex device conditions.

[0015] This application constructs a stable matching basis for the current round based on role quota constraints and coupling constraints, and uses the Gale-Shapley stable matching algorithm to perform many-to-one stable matching on standard emotional roles and device capability units. At the same time, after a device capability unit accepts a proposal, it triggers an occupancy propagation operator to synchronously adjust the matching status of associated device capability units. This makes the matching result no longer a static single-point acceptance, but can be updated in a linkage with the overall matching space as the acceptance result occurs. This effectively reduces the probability of carrying conflicts, capability overlaps and response imbalances, and improves the stability and rationality of the matching results.

[0016] After obtaining the current round of stable matching results, this application imports the continued occupation matching structure, calculates the comprehensive instability quantity in combination with the adaptation state parameters, and determines the local instability region based on the comparison result of the comprehensive instability quantity and the instability threshold. This enables the system to determine which role-bearing pairs should continue to occupy and which role-bearing pairs have lost their stable foundation based on the established stable matching results, thereby avoiding direct global recalculation in the case of local instability and improving the system's ability to maintain and repair the dynamic environment. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a standardized adaptation method for emotional interaction based on companion devices proposed in this invention; Figure 2 This diagram illustrates the delayed acceptance, bearer contention adjudication, and occupancy propagation processing of the Gale-Shapley stable matching algorithm, which is proposed in this invention as a standardized adaptation method for emotional interaction in companion devices. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figures 1-2 A standardized adaptation method for emotional interaction based on companion devices includes the following steps: Acquire multimodal data of companion devices during interaction, perform standardization processing, and generate adaptive state parameters; The system identifies the current emotional state based on the adaptation state parameters, generates standardized emotional semantic tags, divides the current round of emotional interaction needs, and determines the corresponding role quota constraints. Identify the current device response capability based on the adaptation status parameters, divide the device capability units that can participate in matching, and determine the coupling constraint relationship between each device capability unit; Based on role quota constraints and coupling constraints, determine the proposal rules, acceptance rules and update rules for stable matching in the current round, forming the basis for stable matching in the current round; Based on the current round of stable matching, the Gale-Shapley stable matching algorithm is used to perform many-to-one stable matching. After the device capability unit accepts the proposal for the standard emotion role, the occupancy propagation operator is triggered to adjust the matching state of the associated device capability unit and obtain the current round of stable matching result. Based on the current round of stable matching results, the continued occupation matching structure is imported, and the comprehensive instability quantity is calculated in combination with the adaptive state parameters. The local instability region is determined according to the comparison results between the comprehensive instability quantity and the instability threshold. For the locally unstable regions, the Gale-Shapley stable matching algorithm is used again to perform local rematching, update the stable matching results of the current round of role carrying, and generate the target emotion interaction adaptation results.

[0020] In this embodiment, obtaining the adaptation state parameters specifically includes: Acquire multimodal data from companion devices during interaction, and perform source-specific collection, time stamping, and raw data archiving. Associate the collection description field to form a raw multimodal data set with a unified time identifier. The multimodal data includes voice data, image data, text data, and touch behavior data. The collection description field includes collection time, duration, sampling source, and sampling order. The original multimodal dataset is processed by time alignment, missing data completion, anomaly removal, unit unification and standardization. The multimodal data is mapped to the corresponding time window, and mean normalization and standard deviation normalization are performed to generate a standardized modal feature set. Adaptive representation parameters are extracted based on a standardized modal feature set. These parameters include interaction intensity, emotion fluctuation, semantic tendency, and behavioral perturbation. The interaction intensity parameter is obtained by weighted summation of the average volume and the average frequency of touch actions. The emotion fluctuation parameter is obtained by calculating the average absolute deviation of the emotion confidence score at each sampling time from the average emotion confidence score. The semantic tendency parameter is obtained by weighted averaging of the emotion polarity values ​​of each text unit in the text unit sequence according to their corresponding semantic weights. The behavioral perturbation parameter is obtained by the ratio between the sum of the number of abnormal touch events, the number of continuous pressing events, and the number of drastic posture change events and the total number of behavioral events. The adaptive representation parameters are combined, and the combined results are processed by range mapping to generate adaptive state parameters.

[0021] In this embodiment, the determination of the corresponding role quota constraint specifically includes: The system acquires the adaptation state parameters, loads them sequentially according to a fixed arrangement, and performs missing parameter blocking, out-of-bounds interception, and conflict resolution in sequence. The retained results are written as the sentiment discrimination base value. Missing parameter blocking is used to stop the current parameter loading process when there are missing parameters in the adaptation state parameters and to prevent missing parameters from entering the sentiment discrimination base value generation process. Out-of-bounds interception is used to stop parameters that exceed the corresponding preset value range from continuing to participate in the sequential loading when any parameter in the adaptation state parameters exceeds the corresponding preset value range. Conflict resolution is used to retain the effect results of parameters with higher priority according to the preset priority order when there are parameters in the adaptation state parameters with opposite value directions, inconsistent change trends, or combination relationships that do not meet the preset discrimination rules. Read the set of emotion categories stored in the companion device, perform matching calculation and strength comparison for each emotion category in the set based on the emotion discrimination base value, determine the emotion category with the largest matching result as the current emotion state, and generate the current emotion state recognition result; Read the current emotion state recognition result and the adaptation state parameters, perform intensity grading and trend orientation on the current emotion state recognition result, and write and encapsulate the obtained emotion category, intensity level and trend identifier according to a unified label format to generate standardized emotion semantic labels; Read standardized emotion semantic tags, call the demand role mapping table stored in companion devices, perform demand mapping and role placement on standardized emotion semantic tags, determine the current round of emotion interaction demand, and generate role quota constraints according to the lower limit, upper limit and prohibition restrictions corresponding to the current round of emotion interaction demand.

[0022] In this embodiment, the determination of the coupling constraint relationship between each equipment capability unit specifically includes: The system obtains the device operating status of companion devices during the current interaction cycle, reads the status of device response capability items in combination with the adaptation status parameters, and performs enable status judgment, occupancy status judgment and availability status judgment on each device response capability item in a fixed reading order to generate a set of device capability status. The device response capability items are used to characterize the voice carrying capacity, display carrying capacity, light carrying capacity, motion carrying capacity, tactile carrying capacity and network carrying capacity of companion devices that can participate in emotional interaction response. Read the set of device capability statuses, perform response eligibility judgment on each device response capability item, retain the device response capability items that pass the judgment, and write the retention result into the set of device response capabilities. The response eligibility judgment is based on the enable status, occupancy status, and availability status of the device response capability item, and combined with the stimulus constraints corresponding to the adaptation status parameters. The device response capability items are sequentially performed on the enable retention, occupancy screening, availability retention, and adaptation correction to determine whether the device response capability item enters the set of device response capabilities. Read the device response capability set, perform capability splitting, unit division and unit registration according to the bearing function, divide the device response capability set into a set of device capability units that can participate in matching, and write the bearing category, bearing capacity and occupancy status for each device capability unit. The bearing category is used to characterize the output bearing type corresponding to the device capability unit, the bearing capacity is used to characterize the number of roles that the device capability unit can bear in the current interaction cycle, and the occupancy status is used to characterize the occupancy status of the device capability unit in the current interaction cycle. Read the set of equipment capability units, perform pairwise relationship discrimination on any two equipment capability units in the set, generate corresponding relationship records according to the processing order of cooperative discrimination, substitution discrimination, mutual exclusion discrimination and coverage discrimination, and determine the coupling constraint relationship between each equipment capability unit based on the corresponding relationship records. Cooperative discrimination is used to determine the common bearing relationship, substitution discrimination is used to determine the supplementary bearing relationship, mutual exclusion discrimination is used to determine the prohibited coexistence relationship, and coverage discrimination is used to determine the priority exclusive relationship.

[0023] In this embodiment, the criteria for obtaining the current round of stable matching specifically include: Read the role quota constraints, perform proposal qualification sorting on the role set, and perform acceptance qualification sorting on the equipment capacity unit set. The proposal qualification sorting retains roles whose current carrying capacity is less than the maximum carrying capacity, and the acceptance qualification sorting retains equipment capacity units whose carrying capacity is greater than the current occupied capacity and are not in a prohibited state. Generate the current round candidate role sequence and the current round candidate equipment capacity unit sequence. Read the current round candidate role sequence and the current round candidate device capability unit sequence, and sequentially point each role to the corresponding device capability unit to initiate a bearer request. If the request is not accepted, move to the next device capability unit to continue initiating a bearer request, and generate a proposal record. Read the proposal record and coupling constraint relationship, perform comparison and acceptance processing on the equipment capability units that received role proposals. The comparison and acceptance processing includes retaining the role proposals with higher priority, and after completing the acceptance, retaining the equipment capability units that are allowed to participate in the matching in the current round within the scope of proposals, removing the equipment capability units that are prohibited from participating in the matching from the scope of proposals in the current round, marking the equipment capability units whose carrying functions have been replaced by the current acceptance result as no longer acceptable, and generating an acceptance record. Read the acceptance record, role quota constraints, and coupling constraints; increase the current number of accepted roles and the current number of accepted equipment capacity units occupied; rewrite the participation status of equipment capacity units affected by the acceptance result; and generate the basis for stable matching in the current round.

[0024] This invention constructs a rule-based basis for stable matching in the current round by performing proposal qualification sorting and acceptance qualification sorting on the role set and the equipment capability unit set respectively, and combining proposal records, acceptance records and participation status rewriting affected by acceptance results. This enables effective screening of candidate objects, orderly limitation of proposal paths and state linkage control after acceptance before matching begins, thereby improving the executability, stability and coordination of the role-based matching process.

[0025] In this embodiment, obtaining the stable matching result of the current round specifically includes: Read the current round of stable matching criteria, write the standard emotion roles whose current number of loads is less than the maximum number of loads into the proposed sequence, write the device capability units whose current number of loads is less than the capacity and are in an acceptable state into the accept sequence, and allocate the corresponding number of load positions to each device capability unit according to the capacity, and generate the current round of delayed acceptance initial state. The standard sentiment roles in the proposed sequence are proposed to the target device capability unit in the order of proposal. The target device capability unit performs delayed acceptance processing according to the Gale-Shapley stable matching algorithm. For newly arrived standard sentiment roles, a temporary holding comparison is performed first. If there is an idle bearer bit in the target device capability unit, the newly arrived standard sentiment role is temporarily held in the idle bearer bit. If there is no idle bearer bit in the target device capability unit, the newly arrived standard sentiment role is matched with the standard sentiment roles that have been temporarily held in each bearer bit in the target device capability unit. The standard sentiment roles with higher role bearing priority continue to occupy the corresponding bearer bit, and the standard sentiment roles with lower role bearing priority are released from the corresponding bearer bit. The temporary holding acceptance result of a single device capability unit is generated. A single device capability unit refers to any one device capability unit in the set of device capability units, which is used as the single acceptance object corresponding to the current standard sentiment role when it makes a proposal. After a single device capability unit forms a temporary acceptance result, the occupancy propagation operator is immediately triggered to propagate the temporary acceptance result of the single device capability unit to the associated device capability units along the coupling constraint relationship. The associated device capability units refer to the device capability units that have common bearing restrictions, prohibition of common bearing restrictions, bearing substitution restrictions, or bearing coverage restrictions with the single device capability unit in the coupling constraint relationship. It is used to perform matching state adjustment after the single device capability unit forms a temporary acceptance result, and to perform matching domain compression processing on the associated device capability units. It deletes the standard emotion roles that are prohibited from being co-beared with the current temporary acceptance result, retains the standard emotion roles that are allowed to be co-beared with the current temporary acceptance result, reduces the number of available bearing bits of the associated device capability units that have a bearing bit competition relationship with the current temporary acceptance result, and rewrites the associated device capability units whose bearing functions have been covered by the current temporary acceptance result into an unacceptable state, generating the propagated associated device capability unit matching state. For standard sentiment roles released in the bearer contention adjudication, a backtracking and redirection process is performed. Based on the matching status of the associated device capability units after propagation, device capability units that still retain the corresponding bearer category and still have available bearer bits are re-screened. The released standard sentiment roles are then directed to the next device capability unit after rescreening to continue initiating proposals. The process of proposal, delayed acceptance, bearer contention adjudication, occupancy propagation operator triggering, and backtracking and redirection is repeated until there are no standard sentiment roles in the proposal sequence that can continue to initiate proposals, or no device capability units in the acceptance sequence that can continue to accept proposals. The final temporary retention result of each bearer bit within each device capability unit is written as the stable matching result of the current round.

[0026] This invention employs the Gale–Shapley stable matching algorithm, combined with delayed acceptance processing, bearer bit contention adjudication, occupancy propagation operator, matchable field compression, and backoff redirection processing, to achieve dynamic and stable matching between standard emotional roles and device capability units. Under coupling constraints, it can reduce bearer conflicts and erroneous acceptance, and improve matching stability, response accuracy, and overall coordination in the emotional interaction adaptation process.

[0027] In this embodiment, the determination of the locally unstable region specifically includes: Extract all role-bearing pairs from the current round of stable matching results, and perform renewal registration on each role-bearing pair. Write the standard emotion role, device capability unit, bearer position, and the corresponding adaptation state parameters when forming the current round of stable matching results into the corresponding renewal unit to generate a renewal matching structure. A renewal unit is a record unit formed after binding and registering a single role-bearing pair and the corresponding adaptation state parameters when forming the role-bearing pair. A renewal matching structure is a registration structure formed by combining all renewal units to characterize the current round of stable matching results and their corresponding parameter reference relationships.

[0028] The continued occupation matching structure is sequentially subjected to continued occupation retention determination. The difference between the current adaptation status parameter and the adaptation status parameter registered by the corresponding continued occupation unit is compared. The comprehensive instability amount is calculated in combination with the role quota constraint and the current occupation status of the equipment capacity unit. When the comprehensive instability amount is less than the instability threshold, the continued occupation status of the corresponding role bearing pair is retained. When the comprehensive instability amount is not less than the instability threshold, the corresponding role bearing pair is written as the instability candidate bearing pair. For the candidate bearer pairs that are to break stability, perform boundary stability propagation judgment. Check whether the role bearer pairs directly adjacent to the candidate bearer pairs along the coupling constraint relationship have lost their continued occupation state. Continue to write the adjacent role bearer pairs that meet the propagation conditions as candidate bearer pairs that are to break stability. Repeat the adjacent check and candidate writing to generate a stability propagation chain. The unstable transmission chain is partially closed and defined. The unstable transmission chain that is completely de-occupied inside and surrounded by role carriers that maintain the state of continuous occupation outside is written as a local unstable region. The role carriers that maintain the state of continuous occupation outside the local unstable region and are directly adjacent to the local unstable region are written as boundary carrier pairs, thus generating a local unstable region.

[0029] This invention establishes a continued occupation matching structure based on the current round's stable matching results. Combined with adaptation state parameters, role quota constraints, and equipment capacity unit occupancy status, it performs continued occupation retention determination, stability transmission identification, and local closure delineation for role-bearing pairs. This can accurately identify the local instability range and its boundary relationships, avoiding the need for overall recalculation of the stable bearing part. As a result, it improves the stability, continuity, local repair efficiency, and resource utilization rationality in the emotional interaction adaptation process.

[0030] In this embodiment, obtaining the target emotion interaction adaptation result specifically includes: Screen out the role bearer pairs to be released from the locally unstable region, perform bearer release and gap registration on the role bearer pairs to be released, write the original equipment capacity unit bearer bits occupied by the role bearer pairs to be released as reconfiguration gaps, and write the equipment capacity units corresponding to the boundary bearer pairs to be boundary lock bits, and generate a local reconfiguration boundary set; Read the local reconfiguration boundary set, filter out the standard sentiment roles that are allowed to participate in local reconfiguration and write them into the local proposal sequence, filter out the equipment capability units that are allowed to participate in gap filling and write them into the local acceptance sequence, and restrict the standard sentiment roles in the local proposal sequence to only make proposals to the equipment capability units corresponding to the reconfiguration gap; The standard emotion roles in the local proposed sequence are proposed to the device capability unit corresponding to the rematch gap in order of proposal order. The device capability unit that receives the proposal re-executes the local delayed acceptance processing according to the Gale-Shapley stable matching algorithm, performs local matchable domain compression and gap redirection according to the coupling constraint relationship, until a local rematch result is generated, and the local rematch result is written back to the current round of role-bearing stable matching result to generate the target emotion interaction adaptation result.

[0031] This invention achieves targeted repair and boundary isolation control of unstable bearing relationships by performing bearing release, gap registration, boundary locking, and gap backfilling on the role bearing pairs within the locally unstable region, and by combining the Gale-Shapley stable matching algorithm to re-execute local delayed acceptance processing, local matchable domain compression, and gap redirection. This improves the local rematching efficiency, matching stability, boundary preservation ability, and overall interaction continuity in the target emotion interaction adaptation result generation process.

[0032] Example 1: To verify the feasibility of the present invention in practice, it was applied to a companion device with voice acquisition, screen display, light feedback, haptic feedback, and network connectivity. In a family companionship interaction scenario, the device identifies, matches, and dynamically repairs the user's emotional changes during continuous companionship. Most existing companion devices adopt a processing method of "directly triggering a fixed response after identifying an emotion." They can usually only output preset comforting content based on a single voice result, a single facial expression result, or a single text result. It is difficult to achieve continuous adaptation by combining differences in device capabilities, role-bearing relationships, and local instability. Therefore, problems such as repetitive response content, device output conflicts, sudden changes in interaction rhythm, and the interruption of the established comforting relationship by subsequent commands are prone to occur.

[0033] In this embodiment, the companion device continuously collects the user's voice data, image data, text data, and touch behavior data after the companionship session begins. It then performs time alignment, missing data completion, anomaly removal, unit unification, and standardization on the collected multimodal data to generate adaptation state parameters. Based on these parameters, the device identifies the current emotional state, performs intensity grading and trend orientation on the identification results, generates standardized emotional semantic tags, and then maps these tags to a demand role mapping table to determine the current round of emotional interaction demands and corresponding role quota constraints. Simultaneously, the device identifies its responsiveness based on its current operating state and categorizes its voice carrying capacity, display carrying capacity, lighting carrying capacity, motion carrying capacity, tactile carrying capacity, and network carrying capacity into matching device capability units. Through pairwise relationship discrimination, it determines the common carrying relationship, supplementary carrying relationship, prohibited coexistence relationship, and priority exclusive relationship between these device capability units, thereby forming coupling constraint relationships between them.

[0034] The system establishes a stable matching basis for the current round based on role quota constraints and coupling constraints. Then, standard emotion roles initiate bearer requests to equipment capability units according to their proposal order. Equipment capability units perform delayed acceptance processing according to the Gale-Shapley stable matching algorithm. If there are available bearer slots, the newly arrived standard emotion roles are temporarily reserved. If there are no available bearer slots, a bearer slot competition ruling is performed between the newly arrived standard emotion roles and the already reserved standard emotion roles. Standard emotion roles with higher bearer priority continue to occupy the corresponding bearer slots, while standard emotion roles with lower priority are released. After a single equipment capability unit forms a temporary acceptance result, the occupancy propagation operator immediately propagates the temporary acceptance result to associated equipment capability units along the coupling constraints. The associated equipment capability units are then subjected to matchable domain compression processing, which deletes standard emotion roles that are prohibited from being co-beared, retains standard emotion roles that are allowed to be co-beared, reduces the number of available bearer slots in associated equipment capability units with bearer slot competition, and rewrites associated equipment capability units whose bearer functions have been covered as unacceptable.

[0035] For example, in a typical companionship conversation, users initially exhibit characteristics such as decreased voice energy, negative text polarity, increased touch frequency but longer duration of continuous pressing. Based on this, the system identifies the current emotional state as depressed and accompanied by a tendency towards anxiety, and encapsulates it into corresponding standardized emotional semantic tags. Subsequently, the system categorizes the current round of emotional interaction needs into emotional soothing roles, companionship maintenance roles, and rhythmic easing roles, and sets lower and upper limits for each role, as well as prohibition restrictions. The device identifies that the voice, light, display, and tactile capabilities are available, while the motion capability is deemed unavailable due to current constraints. After the standard emotion roles sequentially initiate capability requests to the device's capability units, the voice capability prioritizes accepting the emotion-soothing role, the light capability accepts the rhythm-releasing role, and the display capability accepts the companionship-maintaining role. As the light capability completes its acceptance, the occupancy propagation operator compresses the matchable domain of the display capability according to the coupling constraint relationship to prevent high-frequency dynamic visual stimuli from occurring simultaneously with soothing voices. This results in the device output being low-speed soothing voices, low-brightness, slowly changing lights, and a static companionship interface, rather than simultaneously triggering high-stimulation lights, dynamic interfaces, and strong reminder feedback.

[0036] During the ongoing companionship session, the user's input state may dynamically change, and the established stable matching results may no longer be suitable for maintenance due to parameter offsets, changes in device capacity utilization, or disturbances in coupling relationships. In this case, the present invention does not directly overturn all bearer relationships. Instead, it extracts all role bearer pairs from the current round of stable matching results, performs continuation registration on each role bearer pair, generates a continuation matching structure, and then identifies locally unstable regions through continuation retention determination, boundary instability propagation determination, and local closure delineation. The system only classifies bearer relationships that have truly lost their continuation state and constitute a propagation chain into locally unstable regions, while simultaneously writing role bearer pairs directly adjacent to the locally unstable region and maintaining their continuation state as boundary bearer pairs. Subsequently, the system only performs bearer release, gap registration, boundary locking, gap backfilling, local delayed acceptance processing, local matchable domain compression, and gap redirection on the locally unstable regions, and then writes the obtained local rematch results back to the current round of stable role bearer matching results to generate the target emotion interaction adaptation result.

[0037] To verify the practical application effect of this invention in emotional interaction scenarios of companion devices, a comparative test was conducted with a fixed template response scheme, a conventional priority matching scheme, and the solution of this invention. The fixed template response scheme refers to the comparison scheme that directly calls a preset response template to output the corresponding interactive content after recognizing the user's current emotional state. The conventional priority matching scheme refers to the comparison scheme that selects the device's response capability to execute emotional interaction output according to a preset priority order after recognizing the user's current emotional state. Specific comparison data is shown in Table 1: Table 1. Key performance comparison of different emotion interaction adaptation schemes in continuous companionship scenarios.

[0038] As shown in Table 1, under the same session sample size, the overall performance of the proposed solution in continuous companionship scenarios is significantly better than that of the fixed template response solution and the conventional priority matching solution. The average number of interaction rounds of the proposed solution reaches 18.9 rounds, which is higher than the 13.4 rounds of the fixed template response solution and the 15.8 rounds of the conventional priority matching solution, indicating that the proposed solution can support longer duration and more stable continuous interaction. In terms of matching stability, the proposed solution reduces the proportion of unnecessary role replacement to 6.9%, which is significantly lower than 31.7% and 18.6%, respectively. At the same time, it reduces the number of device output conflicts to 0.5 times / session, indicating that the proposed solution... The invention effectively reduces load disturbances and multi-channel conflicts. Regarding dynamic repair capabilities, the invention achieves a local instability identification accuracy of 91.2% and reduces the average local repair time to 1.7 seconds, both significantly better than the comparative schemes. This demonstrates that the invention is more efficient in local instability identification and local rematching. In terms of interaction effects, the invention achieves 17.3 effective consecutive interaction rounds and a mood improvement rate of 78.9%, both higher than the fixed template response scheme and the conventional priority matching scheme. This indicates that the invention can improve the continuity, coordination, and actual improvement effect of emotional interaction responses while ensuring matching stability.

[0039] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A standardized adaptation method for emotional interaction based on companion devices, characterized in that, Includes the following steps: Acquire multimodal data of companion devices during interaction, perform standardization processing, and generate adaptive state parameters; The system identifies the current emotional state based on the adaptation state parameters, generates standardized emotional semantic tags, divides the current round of emotional interaction needs, and determines the corresponding role quota constraints. Identify the current device response capability based on the adaptation status parameters, divide the device capability units that can participate in matching, and determine the coupling constraint relationship between each device capability unit; Based on role quota constraints and coupling constraints, determine the proposal rules, acceptance rules and update rules for stable matching in the current round, forming the basis for stable matching in the current round; Based on the current round of stable matching, the Gale-Shapley stable matching algorithm is used to perform many-to-one stable matching. After the device capability unit accepts the proposal for the standard emotion role, the occupancy propagation operator is triggered to adjust the matching state of the associated device capability unit and obtain the current round of stable matching result. Based on the current round of stable matching results, the continued occupation matching structure is imported, and the comprehensive instability quantity is calculated in combination with the adaptive state parameters. The local instability region is determined according to the comparison results between the comprehensive instability quantity and the instability threshold. For the locally unstable regions, the Gale-Shapley stable matching algorithm is used again to perform local rematching, update the stable matching results of the current round of role carrying, and generate the target emotion interaction adaptation results.

2. The standardized adaptation method for emotional interaction based on companion devices according to claim 1, characterized in that, The acquisition of the adaptation state parameters specifically includes: Acquire multimodal data from companion devices during interaction, and perform source-specific collection, time stamping, and raw data archiving. Associate the collection description fields to form a raw multimodal data set with a unified time identifier. The original multimodal dataset is processed by time alignment, missing data completion, anomaly removal, unit unification and standardization. The multimodal data is mapped to the corresponding time window, and mean normalization and standard deviation normalization are performed to generate a standardized modal feature set. Adaptive representation parameters are extracted based on a standardized modal feature set, combined, and the combination results are processed by range mapping to generate adaptive state parameters.

3. The standardized adaptation method for emotional interaction based on companion devices according to claim 1, characterized in that, The determination of the corresponding role quota constraints specifically includes: Obtain the adaptation state parameters, load the adaptation state parameters in a fixed order, and perform missing item blocking, out-of-bounds interception and conflict resolution in sequence. Write the retained results as the sentiment discrimination base value. Read the set of emotion categories stored in the companion device, perform matching calculation and strength comparison for each emotion category in the set based on the emotion discrimination base value, determine the emotion category with the largest matching result as the current emotion state, and generate the current emotion state recognition result; Read the current emotion state recognition result and the adaptation state parameters, perform intensity grading and trend orientation on the current emotion state recognition result, and write and encapsulate the obtained emotion category, intensity level and trend identifier according to a unified label format to generate standardized emotion semantic labels; Read standardized emotion semantic tags, call the demand role mapping table stored in companion devices, perform demand mapping and role placement on standardized emotion semantic tags, determine the current round of emotion interaction demand, and generate role quota constraints according to the lower limit, upper limit and prohibition restrictions corresponding to the current round of emotion interaction demand.

4. The standardized adaptation method for emotional interaction based on companion devices according to claim 1, characterized in that, The determination of the coupling constraint relationship between the various equipment capability units specifically includes: The device operation status of the companion device in the current interaction cycle is obtained. The status of the device response capability item is read in combination with the adaptation status parameters. The enable status judgment, occupancy status judgment and availability status judgment are performed on each device response capability item in a fixed reading order to generate a set of device capability status. Read the device capability status set, perform response eligibility judgment on each device response capability item, retain the device response capability items that pass the judgment, and write the retention results into the device response capability set; Read the set of device response capabilities, perform capability splitting, unit division and unit registration on the set of device response capabilities according to the bearing function, divide the set of device response capabilities into a set of device capability units that can participate in matching, and write the bearing category, bearing capacity and occupancy status for each device capability unit respectively. Read the set of equipment capability units, perform pairwise relationship discrimination on any two equipment capability units in the set, generate corresponding relationship records in the order of cooperative discrimination, substitution discrimination, mutual exclusion discrimination and overriding discrimination, and determine the coupling constraint relationship between each equipment capability unit based on the corresponding relationship records.

5. The standardized adaptation method for emotional interaction based on companion devices according to claim 1, characterized in that, The criteria for obtaining the current round of stable matching specifically include: Read the role quota constraints, perform proposal qualification sorting on the role set, perform acceptance qualification sorting on the equipment capability unit set, and generate the current round candidate role sequence and the current round candidate equipment capability unit sequence; Read the current round candidate role sequence and the current round candidate device capability unit sequence, and sequentially point each role to the corresponding device capability unit to initiate a bearer request. If the request is not accepted, move to the next device capability unit to continue initiating a bearer request, and generate a proposal record. Read the proposal record and coupling constraint relationship, perform comparison and acceptance processing on the device capability unit that received the role proposal, and generate an acceptance record; Read the acceptance record, role quota constraints, and coupling constraints; increase the current number of accepted roles and the current number of accepted equipment capacity units occupied; rewrite the participation status of equipment capacity units affected by the acceptance result; and generate the basis for stable matching in the current round.

6. The standardized adaptation method for emotional interaction based on companion devices according to claim 1, characterized in that, The specific steps to obtain the stable matching result of the current round include: Read the current round of stable matching criteria, write the standard emotion roles whose current number of loads is less than the maximum number of loads into the proposed sequence, write the device capability units whose current number of loads is less than the capacity and are in an acceptable state into the accept sequence, and allocate the corresponding number of load positions to each device capability unit according to the capacity, and generate the current round of delayed acceptance initial state. The standard sentiment roles in the proposed sequence are proposed to the target device capability unit in the order of proposal. The target device capability unit performs delayed acceptance processing according to the Gale-Shapley stable matching algorithm. For newly arrived standard sentiment roles, a temporary holding comparison is performed first. If there are free bearer slots in the target device capability unit, the newly arrived standard sentiment roles are temporarily held in the free bearer slots. If there are no free bearer slots in the target device capability unit, the newly arrived standard sentiment roles are compared with the standard sentiment roles that have been temporarily held in each bearer slot in the target device capability unit. The standard sentiment roles with higher role carrying priority continue to occupy the corresponding bearer slots, and the standard sentiment roles with lower role carrying priority are released from the corresponding bearer slots, generating a temporary holding acceptance result for a single device capability unit. After a single device capability unit forms a temporary acceptance result, the occupancy propagation operator is immediately triggered. The temporary acceptance result of the single device capability unit is propagated to the associated device capability units along the coupling constraint relationship. The associated device capability units are then subjected to matchable domain compression processing. Standard emotion roles that are prohibited from being carried together with the current temporary acceptance result are deleted, while standard emotion roles that are allowed to be carried together with the current temporary acceptance result are retained. The number of available bearer bits of associated device capability units that have a bearer bit competition relationship with the current temporary acceptance result is reduced. The associated device capability units whose bearer functions have been covered by the current temporary acceptance result are rewritten to an unacceptable state, and the propagated associated device capability unit matching state is generated. For standard sentiment roles released in the bearer contention adjudication, a backtracking and redirection process is performed. Based on the matching status of the associated device capability units after propagation, device capability units that still retain the corresponding bearer category and still have available bearer bits are re-screened. The released standard sentiment roles are then directed to the next device capability unit after rescreening to continue initiating proposals. The process of proposal, delayed acceptance, bearer contention adjudication, occupancy propagation operator triggering, and backtracking and redirection is repeated until there are no standard sentiment roles in the proposal sequence that can continue to initiate proposals, or no device capability units in the acceptance sequence that can continue to accept proposals. The final temporary retention result of each bearer bit within each device capability unit is written as the stable matching result of the current round.

7. The standardized adaptation method for emotional interaction based on companion devices according to claim 1, characterized in that, The determination of the locally unstable region specifically includes: Extract all role-bearing pairs from the current round of stable matching results, and perform renewal registration for each role-bearing pair. Write the standard emotion role, device capability unit, bearer position, and the corresponding adaptation state parameters when forming the current round of stable matching results into the corresponding renewal unit to generate a renewal matching structure. The continued occupation matching structure is sequentially subjected to continued occupation retention determination. The difference between the current adaptation status parameter and the adaptation status parameter registered by the corresponding continued occupation unit is compared. The comprehensive instability amount is calculated in combination with the role quota constraint and the current occupation status of the equipment capacity unit. When the comprehensive instability amount is less than the instability threshold, the continued occupation status of the corresponding role bearing pair is retained. When the comprehensive instability amount is not less than the instability threshold, the corresponding role bearing pair is written as the instability candidate bearing pair. For the candidate bearer pairs that are to break stability, perform boundary stability propagation judgment. Check whether the role bearer pairs directly adjacent to the candidate bearer pairs along the coupling constraint relationship have lost their continued occupation state. Continue to write the adjacent role bearer pairs that meet the propagation conditions as candidate bearer pairs that are to break stability. Repeat the adjacent check and candidate writing to generate a stability propagation chain. The unstable transmission chain is partially closed and defined. The unstable transmission chain that is completely de-occupied inside and surrounded by role carriers that maintain the state of continuous occupation outside is written as a local unstable region. The role carriers that maintain the state of continuous occupation outside the local unstable region and are directly adjacent to the local unstable region are written as boundary carrier pairs, thus generating a local unstable region.

8. The standardized adaptation method for emotional interaction based on companion devices according to claim 1, characterized in that, The specific steps to obtain the target emotion interaction adaptation result include: Screen out the role bearer pairs to be released from the locally unstable region, perform bearer release and gap registration on the role bearer pairs to be released, write the original equipment capacity unit bearer bits occupied by the role bearer pairs to be released as reconfiguration gaps, and write the equipment capacity units corresponding to the boundary bearer pairs to be boundary lock bits, and generate a local reconfiguration boundary set; Read the local reconfiguration boundary set, filter out the standard sentiment roles that are allowed to participate in local reconfiguration and write them into the local proposal sequence, filter out the equipment capability units that are allowed to participate in gap filling and write them into the local acceptance sequence, and restrict the standard sentiment roles in the local proposal sequence to only make proposals to the equipment capability units corresponding to the reconfiguration gap; The standard emotion roles in the local proposed sequence are proposed to the device capability unit corresponding to the rematch gap in order of proposal order. The device capability unit that receives the proposal re-executes the local delayed acceptance processing according to the Gale-Shapley stable matching algorithm, performs local matchable domain compression and gap redirection according to the coupling constraint relationship, until a local rematch result is generated, and the local rematch result is written back to the current round of role-bearing stable matching result to generate the target emotion interaction adaptation result.

Images