Intelligent unlocking data analysis system and method based on television

Abstract

The invention discloses an intelligent unlocking data analysis system and method based on a television, and relates to the technical field of television management.The intelligent unlocking data analysis method comprises the steps that a user entrance timestamp is obtained through user authorization; acquiring spatial movement behavior data after the user enters the home and a television startup and shutdown event; analyzing behavior characteristics of the user, and integrating the behavior characteristics into an original behavior data stream; extracting typical user behaviors through behavior state discretization and sequence clustering; constructing a behavior state transition diagram with timestamp probability distribution; performing fusion calculation on a real-time watching intention probability by calculating the morphological similarity between a real-time behavior state sequence and a historical typical behavior sequence and combining the transition intensity and time goodness of fit embodied on a state transition diagram; analyzing and generating a predictive wake-up decision; and the system generates a feedback signal according to the subsequent actual active interaction behavior of the user, and optimizes a dynamic threshold value and a historical behavior model. And the time sequence sensitivity and accuracy of intention recognition are improved.

Description

Technical Field

[0001] This invention relates to the field of television management technology, specifically a smart unlocking data analysis system and method based on a television set. Background Technology

[0002] With the popularization of smart home technology, televisions, as the core device for home entertainment, are gradually shifting from traditional manual operation to intelligent and automated control. While existing television control solutions have improved in convenience, they are still essentially still operating on a passive response model, unable to truly understand user intent, resulting in significant shortcomings in user experience and room for improvement in energy efficiency.

[0003] Some advanced solutions attempt to use sensors to detect user presence and automatically power on / off. However, most of these solutions rely on single, isolated trigger conditions, such as simply detecting a user's presence in front of the television to power it on. This simplistic logic has significant flaws: a user's activities in the living room are diverse and complex; actions such as passing in front of the television, searching for items, or resting on the sofa do not equate to a clear intention to watch television. Systems based on such simplistic rules are prone to misjudgments, leading to frequent false awakenings of the television. This not only disrupts the user's normal life and creates an unpleasant interactive experience but also results in unnecessary energy consumption. Furthermore, these static rules cannot learn and adapt to the unique behavioral habits of different users, lacking the ability to self-optimize and evolve over time, resulting in rigid and limited intelligent performance.

[0004] Therefore, this invention discloses a smart unlocking data analysis system and method based on a television set to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide a smart unlocking data analysis system and method based on a television set, so as to solve the problems raised in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a smart unlocking data analysis method based on a television set, the method comprising the following steps: S1: With user authorization, obtain the user's entry timestamp; obtain the user's spatial movement behavior data after entering the home and the TV's power on / off events; analyze the user's behavioral characteristics and integrate them into a raw behavioral data stream; S2: Based on the original behavioral data stream, typical user behaviors are extracted through behavioral state discretization and sequence clustering; a behavioral state transition graph with timestamp probability distribution is constructed. S3: Match the real-time collected and processed behavioral state sequences with historical behavioral state sequences; calculate the morphological similarity between the real-time behavioral state sequences and historical typical behavioral sequences, and combine the transition intensity and time consistency of the real-time behavioral state sequences on the state transition diagram to calculate the probability of real-time viewing intent. S4: Compare the real-time viewing intent probability with the dynamic threshold to generate a predictive wake-up decision; the system generates feedback signals based on the user's subsequent actual active interaction behavior to optimize the dynamic threshold and historical behavior model.

[0007] According to the above scheme, S1 includes the following: S101: With user authorization, obtain the user's entry timestamp, denoted as T. in ; Acquire spatial movement behavior data of the user after entering the home; The spatial movement behavior data includes location coordinates and body orientation, forming a time series trajectory Traj={(x a y a θ a , t a )|a∈[1,A]}; where, t a This represents the a-th timestamp after the user enters their home (x a y a ) represents the timestamp t a Corresponding position coordinates; θ a Represents timestamp t a Corresponding body orientation; monitor changes in the television's status, record the power-on event timestamp, denoted as T. on ; S102: Define a sensitive area centered on the television set, and determine the core behavior observation window length W based on historical data. core The core behavior observation window length is equal to the sum of the mode of the time interval from the entry timestamp to the power-on event timestamp and the buffer time; the buffer time is a system preset time; for each user's entry event, a time series trajectory is uniformly extracted within a time period starting from the entry timestamp and lasting for the length of the core behavior observation window; the behavioral characteristics within the extracted time window are analyzed; the behavioral characteristics include the number of times N is traversed sensitive areas. cross The set of durations of all dwelling events within the sensitive area {D} stay}; and when staying in a sensitive area, the angle θ between the body's orientation and the normal to the television screen. face Simultaneously, the actual time interval ΔT from the user entering the door to turning on the device is recorded. on If the core behavior observation window length is not enabled, it will be marked as a preset special value. S103: Encapsulate the data into a raw behavioral data stream.raw ={T in Traj core , {D stay}, {θ face}, ΔT on}, where Traj core This represents trajectory data within the length of the core behavior observation window.

[0008] This invention determines the window length by using the mode of historical data and supplements it with a buffer time, achieving a balance between logical rationality and engineering feasibility; it provides a fixed-length, comparable analysis period for all user samples (regardless of whether the device is ultimately powered on), making subsequent sequence clustering and model training possible; by recording additional time intervals, it retains the important information of the timing of behavior as a feature of the model, thereby improving the model's discriminative ability.

[0009] According to the above scheme, S2 includes the following: S201: Divide the trajectory data within the core behavior observation window into continuous time slices at fixed time intervals; discretize and encode the behavior features within each time slice to generate a behavior state vector; the behavior state vector includes movement state, dwell time level, and orientation state; integrate the historical behavior state vectors within the core behavior observation window in chronological order into a behavior state sequence, denoted as S. hist ={B n |n∈[1,N]}, where N is the sequence length, B n This represents the state vector of the nth historical behavior. A density-based sequence clustering method is used to extract representative typical behavior sequences from all behavioral state sequences. Specifically, the clustering process is based on sequence similarity, and sequences in regions with a density greater than a corresponding threshold are identified as typical behavior sequences. The extracted typical behavior sequences constitute a typical behavior sequence library S={S typ m |m∈[1,M]}, where M is the number of typical sequences; S typ m This represents the m-th typical behavior sequence; S202: Based on all historical behavior state sequences, construct a directed behavior state transition graph, denoted as G = (V, E); where the node set V represents the set of all distinct behavior state vectors, and the edge set E represents the transitions between behavior states; where each edge e （i，j） Represents the historical behavior state vector B i Transition to historical behavior state vector B j The directed edges; Statistical historical behavior state vector B iTransition to historical behavior state vector B j The number of times, after normalization, is denoted as the transition weight W. （i，j） Extracting edge e （i，j） For each time stamp of the transfer, the probability density function value P(t|e) is estimated using a kernel function. （i，j） ), where t is the relative time relative to the user's entry timestamp.

[0010] This invention transforms continuous trajectory data into behavioral state vectors composed of movement state, dwell time level, and orientation state. This discretization simplifies complex data into semantically rich symbols, reducing data noise and computational complexity, while making the model easier to understand and interpret. By extracting typical behavioral sequences through density-based sequence clustering, it can automatically discover and summarize several core behavioral scripts that users tend to watch TV, avoiding the one-sidedness and high cost of manually defined rules, and enabling the model to adapt to the behavioral habits of different users. By modeling user behavior as a directed graph, it intuitively depicts the possible paths of behavioral evolution, improving the temporal sensitivity and accuracy of intent recognition.

[0011] According to the above scheme, S3 includes the following: S301: Convert real-time acquired behavioral data into a real-time behavioral state sequence; calculate the dynamic time warp distance between the real-time behavioral state sequence and each typical behavioral sequence in the typical behavioral sequence library; and convert the real-time behavioral state sequence S... real The dynamic time-warped distance between the m-th typical behavior sequence and the m-th typical behavior sequence is denoted as d. dtw (S real S typ m ); and convert it into similarity sim m =exp(-γd dtw (S real S typ m The maximum similarity score is taken as the sequence morphological similarity score, denoted as Sim. seq Where γ is the scaling factor; S302: For each state transition that occurs in the real-time behavioral state sequence, and its specific relative time t. real ; Find the corresponding edge in the behavior state transition graph; if the state transition exists in the behavior state transition graph, calculate the matching score of the state transition, which is equal to the product of the transition weight and the probability density function value; if the state transition does not exist in the behavior state transition graph, the matching score of the state transition adopts a preset non-zero value; real-time behavior state sequence S real The overall state transition diagram matching degree is equal to the geometric mean of the matching scores of all state transitions; S303: The sequence morphological similarity score and the overall state transition graph matching degree are weighted and fused, and denoted as the real-time viewing intention probability.

[0012] This invention uses dual indicators to calculate the probability of intent, achieving an organic combination of macroscopic behavioral paths and microscopic state transition dynamics; it prevents the system from completely negating the entire sequence due to encountering a previously unseen but potentially harmless minor behavioral transition, thus improving the fault tolerance for marginal or new behaviors.

[0013] According to the above scheme, S4 includes the following: S401: If the probability of real-time viewing intent is greater than or equal to the current dynamic threshold, the TV generates a predictive wake-up decision, determines that the user has a clear viewing intent, and the TV enters STR standby mode; if the probability of real-time viewing intent is less than the current dynamic threshold, the TV remains in standby mode and continues to monitor user behavior. S402: After each predictive wake-up operation, the TV immediately starts a verification time window of a preset duration; if the system detects the user's active interaction behavior through the built-in sensor or interface log within the verification time window, the decision correctness label is marked as successful; the current dynamic threshold is updated, and the updated dynamic threshold is equal to the product of the current dynamic threshold and the attenuation factor, wherein the attenuation factor is preset by the system. If the system does not detect any active interaction behavior within the verification time window, and the TV automatically turns off or returns to standby mode due to timeout, the decision correctness label is marked as false wake-up; the current dynamic threshold is updated, and the updated dynamic threshold is equal to the product of the current dynamic threshold and the enhancement factor, which is preset by the system; All real-time behavioral state sequences and their final decision correctness labels obtained within the verification time window are synchronously stored in the historical database; the system periodically or after accumulating a certain amount of new data automatically triggers and executes incremental updates to the typical behavioral sequence library and behavioral state transition diagram.

[0014] The system of this invention automatically adapts to changes in the behavioral sensitivity of different users or the same user at different times; it evolves together with the slow changes in user behavior habits, avoiding the problem of performance degradation due to "outdated concepts" and improving intelligence.

[0015] Another aspect of this application provides a smart unlocking data analysis system based on a television set. The system is applied to the above-mentioned smart unlocking data analysis method based on a television set. The system includes a behavior data acquisition and integration module, a historical behavior analysis module, a viewing intention probability analysis module, and a feedback update module. The behavior data acquisition and integration module is used to obtain the user's entry timestamp with user authorization; obtain the user's spatial movement behavior data after entering the home and the TV's power on / off events; analyze the user's behavioral characteristics, and integrate them into a raw behavior data stream; The historical behavior analysis module is used to extract typical user behaviors based on the original behavior data stream through behavior state discretization and sequence clustering; and to construct a behavior state transition graph with timestamp probability distribution. The viewing intent probability analysis module is used to match the real-time collected and processed behavioral state sequence with the historical behavioral state sequence; by calculating the morphological similarity between the real-time behavioral state sequence and the historical typical behavioral sequence, and combining the transition intensity and time consistency of the real-time behavioral state sequence on the state transition diagram, the real-time viewing intent probability is calculated. The feedback update module is used to compare the real-time viewing intent probability with the dynamic threshold to generate a predictive wake-up decision; the system generates feedback signals based on the user's subsequent actual active interaction behavior to optimize the dynamic threshold and historical behavior model.

[0016] According to the above scheme, the behavior data acquisition and integration module includes a behavior data acquisition unit and a behavior data analysis unit; The behavior data acquisition unit is used to obtain the user's entry timestamp with user authorization; to obtain the user's spatial movement behavior data after entering the home and the TV's power on / off events; The behavioral data analysis unit is used to uniformly capture the time series trajectory of each user's entry event, starting from the entry timestamp and lasting for a period equal to the length of the core behavior observation window; and analyze the behavioral characteristics within the captured time window.

[0017] According to the above scheme, the historical behavior analysis module includes a behavior state sequence analysis unit and a behavior state transition graph construction unit; The behavior state sequence analysis unit is used to divide the trajectory data within the core behavior observation window length into continuous time slices at fixed time intervals; to discretize and encode the behavior features within each time slice to generate a behavior state vector; and to integrate the historical behavior state vectors within the core behavior observation window length into a behavior state sequence in chronological order. The behavior state transition graph construction unit is used to construct a directed behavior state transition graph based on all historical behavior state sequences.

[0018] According to the above scheme, the viewing intent probability analysis module includes a morphological similarity analysis unit and an overall state transition diagram matching degree analysis unit; The morphological similarity analysis unit is used to convert real-time collected behavioral data into real-time behavioral state sequences; calculate the dynamic time warp distance between the real-time behavioral state sequence and each typical behavioral sequence in the typical behavioral sequence library; convert it into similarity; and take the maximum similarity as the sequence morphological similarity score. The overall state transition graph matching degree analysis unit is used to find the corresponding edge in the behavior state transition graph for each state transition that occurs in the real-time behavior state sequence and its specific relative time; if the state transition exists in the behavior state transition graph, the matching score of the state transition is calculated, and the matching score is equal to the product of the transition weight and the probability density function value; if the state transition does not exist in the behavior state transition graph, the matching score of the state transition is a preset non-zero value; the overall state transition graph matching degree of the real-time behavior state sequence is equal to the geometric mean of the matching scores of all state transitions.

[0019] According to the above scheme, the feedback update module includes an intent analysis unit and a verification update unit; The intent analysis unit is used to generate a predictive wake-up decision if the real-time viewing intent probability is greater than or equal to the current dynamic threshold, determine that the user has a clear viewing intent, and the TV enters STR standby mode; if the real-time viewing intent probability is less than the current dynamic threshold, the TV remains in standby mode and continues to monitor user behavior. The verification update unit updates the current dynamic threshold based on a preset verification time window and the user's active interaction behavior; all real-time behavior state sequences and their final decision correctness labels obtained within the verification time window are synchronously stored in the historical database; the system periodically or after accumulating a certain amount of new data automatically triggers and executes incremental updates to the typical behavior sequence library and behavior state transition diagram.

[0020] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention determines the window length by using the mode of historical data and supplements it with a buffer time, achieving a balance between logical rationality and engineering feasibility; it provides a fixed-length, comparable analysis period for all user samples, making subsequent sequence clustering and model training possible; by recording additional time intervals, it retains the important information of the timing of behavior occurrences as a feature of the model, improving the model's discriminative ability. This invention transforms continuous trajectory data into a behavioral state vector composed of movement state, dwell time level, and orientation state. This discretization process simplifies complex data into semantically rich symbols, reducing data noise and computational complexity, while making the model easier to understand and interpret; by extracting typical behavioral sequences through density-based sequence clustering, it can automatically discover and summarize several core behavioral scripts that users tend to watch TV, avoiding the one-sidedness and high cost of manually defined rules, enabling the model to adapt to the behavioral habits of different users; it models user behavior as a directed graph, intuitively depicting the possible paths of behavioral evolution; and it improves the temporal sensitivity and accuracy of intent recognition. This invention uses dual metrics to calculate intent probability, achieving an organic combination of macroscopic behavioral paths and microscopic state transition dynamics. It prevents the system from completely rejecting an entire sequence due to a previously unseen but potentially harmless minor behavioral transition, thus improving its tolerance for marginal or novel behaviors. The system automatically adapts to changes in behavioral sensitivity among different users or even the same user at different times; it evolves alongside the slow changes in user behavior habits, avoiding performance degradation due to outdated concepts and enhancing intelligence. Attached Figure Description

[0021] 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 illustrating a smart unlocking data analysis method based on a television set according to the present invention. Figure 2 This is a schematic diagram of the structure of a smart unlocking data analysis system based on a television set according to the present invention. Detailed Implementation

[0022] 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.

[0023] Please see Figure 1The present invention provides a technical solution: a smart unlocking data analysis method based on a television set, the method comprising the following steps: S1: With user authorization, obtain the user's entry timestamp; obtain the user's spatial movement behavior data after entering the home and the TV's power on / off events; analyze the user's behavioral characteristics and integrate them into a raw behavioral data stream; S1 includes the following: S101: With user authorization, obtain the user's entry timestamp, denoted as T. in ;Acquire spatial movement behavior data of users after entering their homes; Spatial movement behavior data includes location coordinates and body orientation, forming a time series trajectory Traj={(x a y a θ a , t a )|a∈[1,A]}; where, t a This represents the a-th timestamp after the user enters their home (x a y a ) represents the timestamp t a Corresponding position coordinates; θ a Represents timestamp t a Corresponding body orientation; monitor changes in the television's status, record the power-on event timestamp, denoted as T. on ; S102: Define a sensitive area centered on the television set, and determine the core behavior observation window length W based on historical data. core The core behavior observation window length is equal to the sum of the mode of the time interval from the entry timestamp to the power-on event timestamp and the buffer time; the buffer time is a system preset time; for each user's entry event, a time series trajectory is uniformly extracted within a time period starting from the entry timestamp and lasting for the length of the core behavior observation window; the behavioral characteristics within the extracted time window are analyzed; the behavioral characteristics include the number of times N is traversed sensitive areas. cross The set of durations of all dwelling events within the sensitive area {D} stay}; and when staying in a sensitive area, the angle θ between the body's orientation and the normal to the television screen. face Simultaneously, the actual time interval ΔT from the user entering the door to turning on the device is recorded. on If the core behavior observation window length is not enabled, it will be marked as a preset special value. S103: Encapsulate the data into a raw behavioral data stream. raw ={T in Traj core , {D stay}, {θ face}, ΔT on}, where Trajcore This represents trajectory data within the length of the core behavior observation window.

[0024] S2: Based on the original behavioral data stream, typical user behaviors are extracted through behavioral state discretization and sequence clustering; a behavioral state transition graph with timestamp probability distribution is constructed. S2 includes the following: S201: The trajectory data within the core behavior observation window is divided into continuous time slices at fixed time intervals; the behavioral features within each time slice are discretized and encoded to generate a behavior state vector; the behavior state vector includes movement state, dwell time level, and orientation state; the historical behavior state vectors within the core behavior observation window are integrated into a behavior state sequence in chronological order, denoted as S. hist ={B n |n∈[1,N]}, where N is the sequence length, B n This represents the state vector of the nth historical behavior. Example 1: In this example, the fixed time interval is 5 seconds; The movement status indicates the user's movement trend relative to the sensitive area, including three discrete values: crossing, staying, and moving away; where crossing indicates that the user crosses the target area; staying indicates that the user remains stationary within the target area; and moving away indicates that the user moves away from the target area. The dwell time level indicates the length of time a user stays in the target area, including three discrete values: none, short, and long; where none indicates no dwell time; short indicates a brief dwell time, which is less than 3 seconds in this embodiment; and long indicates a long dwell time, which is greater than or equal to 3 seconds in this embodiment. Orientation status indicates the user's orientation relative to the target object, including three discrete values: facing, back to, and side; facing means the user is facing the target directly; back to means the user is facing the target from behind; and side to means the user is facing the target from the side. A density-based sequence clustering method is used to extract representative typical behavior sequences from all behavioral state sequences. Specifically, the clustering process is based on sequence similarity, and sequences in regions with a density greater than a corresponding threshold are identified as typical behavior sequences. The extracted typical behavior sequences constitute a typical behavior sequence library S={S typ m |m∈[1,M]}, where M is the number of typical sequences; S typ m This represents the m-th typical behavior sequence; S202: Based on all historical behavior state sequences, construct a directed behavior state transition graph, denoted as G = (V, E); where the node set V represents the set of all distinct behavior state vectors, and the edge set E represents the transitions between behavior states; where each edge e （i，j）Represents the historical behavior state vector B i Transition to historical behavior state vector B j The directed edges; Statistical historical behavior state vector B i Transition to historical behavior state vector B j The number of times, after normalization, is denoted as the transition weight W. （i，j） Extracting edge e （i，j） For each time stamp of the transfer, the probability density function value P(t|e) is estimated using a kernel function. （i，j） ), where t is the relative time relative to the user's entry timestamp.

[0025] S3: Match the real-time collected and processed behavioral state sequences with historical behavioral state sequences; calculate the morphological similarity between the real-time behavioral state sequences and historical typical behavioral sequences, and combine the transition intensity and time consistency of the real-time behavioral state sequences on the state transition diagram to calculate the probability of real-time viewing intent. S3 includes the following: S301: Convert real-time acquired behavioral data into a real-time behavioral state sequence; calculate the dynamic time warp distance between the real-time behavioral state sequence and each typical behavioral sequence in the typical behavioral sequence library; and convert the real-time behavioral state sequence S... real The dynamic time-warped distance between the m-th typical behavior sequence and the m-th typical behavior sequence is denoted as d. dtw (S real S typ m ); and convert it into similarity sim m =exp(-γd dtw (S real S typ m The maximum similarity score is taken as the sequence morphological similarity score, denoted as Sim. seq Where γ is the scaling factor; S302: For each state transition that occurs in the real-time behavioral state sequence, and its specific relative time t. real ; Find the corresponding edge in the behavior state transition graph; if the state transition exists in the behavior state transition graph, calculate the matching score of the state transition, which is equal to the product of the transition weight and the probability density function value; if the state transition does not exist in the behavior state transition graph, the matching score of the state transition adopts a preset non-zero value; real-time behavior state sequence S real The overall state transition diagram matching degree is equal to the geometric mean of the matching scores of all state transitions; S303: The sequence morphological similarity score and the overall state transition graph matching degree are weighted and fused, and denoted as the real-time viewing intention probability.

[0026] S4: Compare the real-time viewing intent probability with the dynamic threshold to generate a predictive wake-up decision; the system generates feedback signals based on the user's subsequent actual active interaction behavior to optimize the dynamic threshold and historical behavior model.

[0027] S4 includes the following: S401: If the probability of real-time viewing intent is greater than or equal to the current dynamic threshold, the TV generates a predictive wake-up decision, determines that the user has a clear viewing intent, and the TV enters STR standby mode; if the probability of real-time viewing intent is less than the current dynamic threshold, the TV remains in standby mode and continues to monitor user behavior. S402: After each predictive wake-up operation, the TV immediately starts a verification time window of a preset duration; if the system detects the user's active interaction behavior through the built-in sensor or interface log within the verification time window, the decision correctness label is marked as successful; the current dynamic threshold is updated, and the updated dynamic threshold is equal to the product of the current dynamic threshold and the attenuation factor, which is preset by the system. If the system does not detect any active interaction behavior within the verification time window, and the TV automatically turns off or returns to standby mode due to timeout, the decision correctness label is marked as false wake-up; the current dynamic threshold is updated, and the updated dynamic threshold is equal to the product of the current dynamic threshold and the enhancement factor, which is a system preset; All real-time behavioral state sequences and their final decision correctness labels obtained within the verification time window are synchronously stored in the historical database; the system periodically or after accumulating a certain amount of new data automatically triggers and executes incremental updates to the typical behavioral sequence library and behavioral state transition diagram.

[0028] Please see Figure 2 The present invention provides a technical solution: a smart unlocking data analysis system based on a television set, the system comprising a behavior data collection and integration module, a historical behavior analysis module, a viewing intention probability analysis module, and a feedback update module; The behavior data acquisition and integration module is used to obtain the user's entry timestamp with user authorization; obtain the user's spatial movement behavior data after entering the home and the TV's power on / off events; analyze the user's behavioral characteristics and integrate them into a raw behavior data stream; The historical behavior analysis module is used to extract typical user behaviors based on the original behavior data stream through behavior state discretization and sequence clustering; and to construct a behavior state transition graph with timestamp probability distribution. The viewing intent probability analysis module is used to match the real-time collected and processed behavioral state sequences with historical behavioral state sequences. By calculating the morphological similarity between the real-time behavioral state sequences and historical typical behavioral sequences, and combining the transition intensity and time consistency of the real-time behavioral state sequences on the state transition diagram, the real-time viewing intent probability is calculated. The feedback update module compares the real-time viewing intent probability with the dynamic threshold to generate a predictive wake-up decision. The system generates feedback signals based on the user's subsequent actual active interaction behavior to optimize the dynamic threshold and historical behavior model.

[0029] The behavioral data acquisition and integration module includes a behavioral data acquisition unit and a behavioral data analysis unit; The behavior data acquisition unit is used to obtain the user's entry timestamp with user authorization; to obtain the user's spatial movement behavior data after entering the home and the TV's power on / off events; The behavioral data analysis unit is used to uniformly capture the time series trajectory of each user's entry event, starting from the entry timestamp and lasting for a period equal to the length of the core behavior observation window; and analyze the behavioral characteristics within the captured time window.

[0030] The historical behavior analysis module includes a behavior state sequence analysis unit and a behavior state transition diagram construction unit; The behavior state sequence analysis unit is used to divide the trajectory data within the core behavior observation window into continuous time slices at fixed time intervals; to discretize and encode the behavior features within each time slice to generate a behavior state vector; and to integrate the historical behavior state vectors within the core behavior observation window into a behavior state sequence in chronological order. The behavior state transition graph construction unit is used to construct a directed behavior state transition graph based on all historical behavior state sequences.

[0031] The viewing intent probability analysis module includes a morphological similarity analysis unit and an overall state transition diagram matching degree analysis unit; The morphological similarity analysis unit is used to transform real-time collected behavioral data into real-time behavioral state sequences; calculate the dynamic time warp distance between the real-time behavioral state sequence and each typical behavioral sequence in the typical behavioral sequence library; and convert it into similarity; the maximum similarity is taken as the sequence morphological similarity score. The overall state transition graph matching degree analysis unit is used to find the corresponding edge in the behavior state transition graph for each state transition that occurs in the real-time behavior state sequence and its specific relative time. If the state transition exists in the behavior state transition graph, the matching score of the state transition is calculated. The matching score is equal to the product of the transition weight and the probability density function value. If the state transition does not exist in the behavior state transition graph, the matching score of the state transition is a preset non-zero value. The overall state transition graph matching degree of the real-time behavior state sequence is equal to the geometric mean of the matching scores of all state transitions.

[0032] The feedback update module includes an intent analysis unit and a verification update unit; The intent analysis unit is used to generate a predictive wake-up decision if the real-time viewing intent probability is greater than or equal to the current dynamic threshold, determine that the user has a clear viewing intent, and the TV enters STR standby mode; if the real-time viewing intent probability is less than the current dynamic threshold, the TV remains in standby mode and continues to monitor user behavior. The verification update unit updates the current dynamic threshold based on the user's active interaction behavior within a preset verification time window; all real-time behavior state sequences and their final decision correctness labels obtained within the verification time window are synchronously stored in the historical database; the system periodically or after accumulating a certain amount of new data automatically triggers and executes incremental updates to the typical behavior sequence library and behavior state transition diagram.

[0033] 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.

[0034] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A smart unlocking data analysis method based on a television set, characterized in that, The method includes the following steps: S1: With user authorization, obtain the user's entry timestamp; obtain the user's spatial movement behavior data after entering the home and the TV's power on / off events; analyze the user's behavioral characteristics and integrate them into a raw behavioral data stream; S2: Based on the original behavioral data stream, typical user behaviors are extracted through behavioral state discretization and sequence clustering; Construct a behavioral state transition graph with timestamped probability distributions; S3: Match the real-time collected and processed behavioral state sequences with historical behavioral state sequences; calculate the morphological similarity between the real-time behavioral state sequences and historical typical behavioral sequences, and combine the transition intensity and time consistency of the real-time behavioral state sequences on the state transition diagram to calculate the probability of real-time viewing intent. S4: Compare the real-time viewing intent probability with the dynamic threshold to generate a predictive wake-up decision; the system generates feedback signals based on the user's subsequent actual active interaction behavior to optimize the dynamic threshold and historical behavior model.

2. The smart unlocking data analysis method based on a television set according to claim 1, characterized in that: S1 includes the following: S101: With user authorization, obtain the user's entry timestamp, denoted as T. in ; Acquire spatial movement behavior data of the user after entering the home; The spatial movement behavior data includes location coordinates and body orientation, forming a time series trajectory Traj={(x a y a θ a , t a )|a∈[1,A]}; where, t a This represents the a-th timestamp after the user enters their home (x a y a ) represents the timestamp t a Corresponding position coordinates; θ a Represents timestamp t a Corresponding body orientation; monitor changes in the television's status, record the power-on event timestamp, denoted as T. on ; S102: Define a sensitive area centered on the television set, and determine the core behavior observation window length W based on historical data. core The core behavior observation window length is equal to the sum of the mode of the time interval from the entry timestamp to the power-on event timestamp and the buffer time; the buffer time is a system preset time; for each user's entry event, a time series trajectory is uniformly extracted within a time period starting from the entry timestamp and lasting for the length of the core behavior observation window; the behavioral characteristics within the extracted time window are analyzed; the behavioral characteristics include the number of times N is traversed sensitive areas. cross The set of durations of all dwelling events within the sensitive area {D} stay }; and when staying in a sensitive area, the angle θ between the body's orientation and the normal to the television screen. face Simultaneously, the actual time interval ΔT from the user entering the door to turning on the device is recorded. on If the core behavior observation window length is not enabled, it will be marked as a preset special value. S103: Encapsulate the data into a raw behavioral data stream. raw ={T in Traj core , {D stay }, {θ face }, ΔT on }, where Traj core This represents trajectory data within the length of the core behavior observation window.

3. The smart unlocking data analysis method based on a television set according to claim 2, characterized in that: S2 includes the following: S201: Divide the trajectory data within the core behavior observation window into continuous time slices at fixed time intervals; discretize and encode the behavior features within each time slice to generate a behavior state vector; the behavior state vector includes movement state, dwell time level, and orientation state; integrate the historical behavior state vectors within the core behavior observation window in chronological order into a behavior state sequence, denoted as S. hist ={B n |n∈[1,N]}, where N is the sequence length, B n This represents the state vector of the nth historical behavior. A density-based sequence clustering method is used to extract representative typical behavior sequences from all behavioral state sequences. Specifically, the clustering process is based on sequence similarity, and sequences in regions with a density greater than a corresponding threshold are identified as typical behavior sequences. The extracted typical behavior sequences constitute a typical behavior sequence library S={S typ m |m∈[1,M]}, where M is the number of typical sequences; S typ m This represents the m-th typical behavior sequence; S202: Based on all historical behavior state sequences, construct a directed behavior state transition graph, denoted as G = (V, E); where the node set V represents the set of all distinct behavior state vectors, and the edge set E represents the transitions between behavior states; where each edge e （i，j） Represents the historical behavior state vector B i Transition to historical behavior state vector B j The directed edges; Statistical historical behavior state vector B i Transition to historical behavior state vector B j The number of times, after normalization, is denoted as the transition weight W. （i，j） Extracting edge e （i，j） For each time stamp of the transfer, the probability density function value P(t|e) is estimated using a kernel function. （i，j） ), where t is the relative time relative to the user's entry timestamp.

4. The smart unlocking data analysis method based on a television set according to claim 3, characterized in that: S3 includes the following: S301: Convert real-time collected behavioral data into a real-time behavioral state sequence; Calculate the dynamic time warp distance between the real-time behavior state sequence and each typical behavior sequence in the typical behavior sequence library; The real-time behavioral state sequence S real The dynamic time-warped distance between the m-th typical behavior sequence and the m-th typical behavior sequence is denoted as d. dtw (S real S typ m ); and convert it into similarity sim m =exp(-γd dtw (S real S typ m The maximum similarity score is taken as the sequence morphological similarity score, denoted as Sim. seq Where γ is the scaling factor; S302: For each state transition that occurs in the real-time behavioral state sequence, and its specific relative time t. real ; Find the corresponding edge in the behavior state transition graph; if the state transition exists in the behavior state transition graph, calculate the matching score of the state transition, which is equal to the product of the transition weight and the probability density function value; If a state transition does not exist in the behavioral state transition graph, the matching score of the state transition is set to a preset non-zero value. Real-time behavioral state sequence S real The overall state transition diagram matching degree is equal to the geometric mean of the matching scores of all state transitions; S303: The sequence morphological similarity score and the overall state transition graph matching degree are weighted and fused, and denoted as the real-time viewing intention probability.

5. The intelligent unlocking data analysis method based on a television set according to claim 4, characterized in that: S4 includes the following: S401: If the probability of real-time viewing intent is greater than or equal to the current dynamic threshold, the TV generates a predictive wake-up decision, determines that the user has a clear viewing intent, and the TV enters STR standby mode; if the probability of real-time viewing intent is less than the current dynamic threshold, the TV remains in standby mode and continues to monitor user behavior. S402: After each predictive wake-up operation, the TV immediately starts a verification time window of a preset duration; if the system detects the user's active interaction behavior through the built-in sensor or interface log within the verification time window, the decision correctness label is marked as successful; the current dynamic threshold is updated, and the updated dynamic threshold is equal to the product of the current dynamic threshold and the attenuation factor, wherein the attenuation factor is preset by the system. If the system does not detect any active interaction behavior within the verification time window, and the TV automatically turns off or returns to standby mode due to timeout, the decision correctness label is marked as false wake-up; the current dynamic threshold is updated, and the updated dynamic threshold is equal to the product of the current dynamic threshold and the enhancement factor, which is a system preset; All real-time behavioral state sequences and their final decision correctness labels obtained within the verification time window are synchronously stored in the historical database; the system periodically or after accumulating a certain amount of new data automatically triggers and executes incremental updates to the typical behavioral sequence library and behavioral state transition diagram.

6. A smart unlocking data analysis system based on a television set, wherein the system is applied to the smart unlocking data analysis method based on a television set as described in any one of claims 1-5, characterized in that, The system includes a behavioral data collection and integration module, a historical behavior analysis module, a viewing intent probability analysis module, and a feedback update module; The behavior data collection and integration module is used to obtain the user's entry timestamp with user authorization; Acquire user spatial movement behavior data after entering the home and TV power-on / off events; Analyze user behavior characteristics and integrate them into a raw behavioral data stream; The historical behavior analysis module is used to extract typical user behaviors based on the original behavior data stream through behavior state discretization and sequence clustering. Construct a behavioral state transition graph with timestamped probability distributions; The viewing intent probability analysis module is used to match the real-time collected and processed behavioral state sequences with historical behavioral state sequences; by calculating the morphological similarity between the real-time behavioral state sequences and historical typical behavioral sequences, and combining the transition intensity and time consistency of the real-time behavioral state sequences on the state transition diagram, the real-time viewing intent probability is calculated. The feedback update module is used to compare the real-time viewing intent probability with the dynamic threshold to generate a predictive wake-up decision; the system generates feedback signals based on the user's subsequent actual active interaction behavior to optimize the dynamic threshold and historical behavior model.

7. The intelligent unlocking data analysis system based on a television set according to claim 6, characterized in that: The behavior data acquisition and integration module includes a behavior data acquisition unit and a behavior data analysis unit; The behavior data collection unit is used to obtain the user's entry timestamp with user authorization; Acquire user spatial movement behavior data after entering the home and TV power-on / off events; The behavioral data analysis unit is used to uniformly capture the time series trajectory of each user's entry event, starting from the entry timestamp and lasting for a period equal to the length of the core behavior observation window; and analyze the behavioral characteristics within the captured time window.

8. The intelligent unlocking data analysis system based on a television set according to claim 6, characterized in that: The historical behavior analysis module includes a behavior state sequence analysis unit and a behavior state transition graph construction unit; The behavior state sequence analysis unit is used to divide the trajectory data within the core behavior observation window length into continuous time slices at fixed time intervals; to discretize and encode the behavior features within each time slice to generate a behavior state vector; and to integrate the historical behavior state vectors within the core behavior observation window length into a behavior state sequence in chronological order. The behavior state transition graph construction unit is used to construct a directed behavior state transition graph based on all historical behavior state sequences.

9. A smart unlocking data analysis system based on a television set according to claim 6, characterized in that: The viewing intent probability analysis module includes a morphological similarity analysis unit and an overall state transition diagram matching degree analysis unit; The morphological similarity analysis unit is used to convert real-time collected behavioral data into real-time behavioral state sequences; and to calculate the dynamic time warp distance between the real-time behavioral state sequence and each typical behavioral sequence in the typical behavioral sequence library. This is then converted into a similarity score; the maximum similarity score is taken as the sequence morphological similarity score. The overall state transition graph matching degree analysis unit is used to analyze each state transition that occurs in the real-time behavioral state sequence and its specific relative time. Find the corresponding edge in the behavior state transition graph; if the state transition exists in the behavior state transition graph, calculate the matching score of the state transition, which is equal to the product of the transition weight and the probability density function value; If a state transition does not exist in the behavioral state transition graph, the matching score of the state transition is set to a preset non-zero value. The overall state transition graph matching degree of the real-time behavioral state sequence is equal to the geometric mean of the matching scores of all state transitions.

10. A smart unlocking data analysis system based on a television set according to claim 6, characterized in that: The feedback update module includes an intent analysis unit and a verification update unit; The intent analysis unit is used to generate a predictive wake-up decision if the real-time viewing intent probability is greater than or equal to the current dynamic threshold, determine that the user has a clear viewing intent, and the TV enters STR standby mode; if the real-time viewing intent probability is less than the current dynamic threshold, the TV remains in standby mode and continues to monitor user behavior. The verification update unit updates the current dynamic threshold based on a preset verification time window and the user's active interaction behavior; all real-time behavior state sequences and their final decision correctness labels obtained within the verification time window are synchronously stored in the historical database; the system periodically or after accumulating a certain amount of new data automatically triggers and executes incremental updates to the typical behavior sequence library and behavior state transition diagram.

Images