Heterogeneous data timing alignment method and apparatus, terminal and medium

Abstract

The application relates to the technical field of data processing, in particular to a heterogeneous data time sequence alignment method and device, a terminal and a medium. The method comprises the following steps: obtaining a physiological feature flow corresponding to a target object, and obtaining window physiological feature data corresponding to each sliding time window; determining environment natural rhythm data according to current time information and position information, and obtaining time sequence alignment heterogeneous data corresponding to each sliding time window; determining a seasonal calibration operator and a day-night time calibration operator according to the environment natural rhythm data, and obtaining a preset physiological rhythm correction operator; performing cascade transformation and normalization processing on the standardized window physiological feature data, and obtaining a standardized time sequence alignment feature vector corresponding to the target sliding time window. The application can solve the technical problem that in the prior art, high-frequency physiological feature flow and low-frequency environment natural rhythm data are difficult to establish an accurate time corresponding relationship, leading to insufficient consistency of heterogeneous time sequence data fusion.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, terminal and medium for heterogeneous data time sequence alignment. Background Technology

[0002] With the development of smart wearable devices and digital health technologies, the continuous acquisition of users' physiological characteristics such as heart rate, heart rate variability, and exercise status through smartwatches and smart bracelets, and the analysis of users' physiological status based on the acquired physiological characteristics, has been gradually applied to scenarios such as health monitoring and human-computer interaction.

[0003] In practical applications, physiological characteristics generated by smart wearable devices are typically output as continuous data streams. For example, physiological characteristics such as heart rate variability, real-time heart rate, and exercise intensity change continuously over time, forming a physiological characteristic stream with a high sampling frequency. Affected by factors such as data transmission delays, data loss, sensor disturbances, or user movements, the collected physiological characteristic streams may contain missing or abnormal data. Directly processing the raw physiological characteristic stream can easily lead to abnormal sampling results interfering with subsequent data analysis. Furthermore, the lack of a unified time window for continuous physiological characteristic streams hinders the establishment of correspondences between data at different time scales. On the other hand, human physiological characteristics are not only influenced by an individual's current activity state but also vary with seasons, day and night, and time cycles. Therefore, when processing physiological characteristic streams, it is also necessary to consider the corresponding environmental and temporal context.

[0004] However, existing physiological data analysis technologies typically focus primarily on the physiological characteristics themselves, or only record environmental information such as date, season, and time of day as supplementary fields. This makes it difficult to establish a clear temporal match between physiological characteristics and the corresponding environmental natural rhythms at the time of their occurrence. Physiological characteristic streams are usually high-frequency continuous data with sampling frequencies at the second level or higher, while environmental natural rhythm data such as solar terms and time periods are low-frequency, slowly varying data with periods at the hour, day, or even longer. There are significant differences between the two in terms of data change frequency and time scale. If environmental natural rhythm data is directly concatenated or cascaded with physiological characteristic streams, it is easy to encounter the problem of multiple physiological sampling data corresponding to the same environmental natural rhythm state, making it difficult to determine the environmental natural rhythm state corresponding to physiological characteristic data within different sliding time windows.

[0005] Based on this, it is necessary to provide a heterogeneous data time-series alignment method, device, terminal, and medium to address the above-mentioned technical problems. This method can solve the technical problem in the prior art that it is difficult to establish an accurate time correspondence between high-frequency physiological feature streams and low-frequency environmental natural rhythm data, resulting in insufficient consistency in the degree of heterogeneous time-series data fusion. Summary of the Invention

[0006] Based on this, it is necessary to provide a heterogeneous data time-series alignment method, device, terminal, and medium to address the above-mentioned technical problems. This method can solve the technical problem in the prior art that it is difficult to establish an accurate time correspondence between high-frequency physiological feature streams and low-frequency environmental natural rhythm data, resulting in insufficient consistency in the degree of heterogeneous time-series data fusion.

[0007] The technical solution of this invention is as follows: A method for time-series alignment of heterogeneous data, the method comprising: Obtain the physiological feature stream corresponding to the target object, preprocess the physiological feature stream, and divide the preprocessed physiological feature stream according to a preset sliding time window to obtain the window physiological feature data corresponding to each sliding time window; Based on the current time and location information, environmental natural rhythm data is determined, and the environmental natural rhythm data is time-matched with the physiological characteristic data of each window to obtain time-aligned heterogeneous data corresponding to each sliding time window. From the time-aligned heterogeneous data corresponding to the target sliding time window, obtain the window physiological characteristic data and environmental natural rhythm data, perform standardization processing on the window physiological characteristic data, determine the seasonal calibration operator and the day-night time calibration operator based on the environmental natural rhythm data, and obtain the preset physiological rhythm correction operator. Based on the seasonal calibration operator, the day-night time calibration operator, and the physiological rhythm correction operator, the standardized window physiological feature data are subjected to cascade transformation and normalization to obtain the standardized temporal alignment feature vector corresponding to the target sliding time window.

[0008] Optionally, a heterogeneous data timing alignment device is also provided, the device comprising: The physiological feature acquisition module is used to acquire the physiological feature stream corresponding to the target object, preprocess the physiological feature stream, and divide the preprocessed physiological feature stream according to a preset sliding time window to obtain the window physiological feature data corresponding to each sliding time window. The heterogeneous data alignment module is used to determine environmental natural rhythm data based on current time and location information, and to perform time matching between the environmental natural rhythm data and the physiological characteristic data of each window to obtain time-aligned heterogeneous data corresponding to each sliding time window. The calibration operator acquisition module is used to acquire window physiological characteristic data and environmental natural rhythm data from the time-aligned heterogeneous data corresponding to the target sliding time window, perform standardization processing on the window physiological characteristic data, determine the seasonal calibration operator and the diurnal time calibration operator based on the environmental natural rhythm data, and acquire the preset physiological rhythm correction operator. The feature vector generation module is used to perform cascade transformation and normalization on the standardized window physiological feature data according to the seasonal calibration operator, the day and night time calibration operator and the physiological rhythm correction operator, so as to obtain the standardized temporal alignment feature vector corresponding to the target sliding time window.

[0009] Optionally, a computer terminal is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps described in the heterogeneous data timing alignment method above.

[0010] Optionally, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps described in the heterogeneous data timing alignment method.

[0011] The technical effects achieved by this invention are as follows: The aforementioned heterogeneous data time-series alignment method, apparatus, terminal, and medium acquire the physiological feature stream corresponding to the target object, preprocess the physiological feature stream, and divide the preprocessed physiological feature stream according to a preset sliding time window. This converts the continuously generated physiological feature stream into window physiological feature data corresponding to different time ranges, reducing the interference of missing and abnormal data on subsequent processing while providing a unified time window basis for establishing a time correspondence between environmental natural rhythm data and physiological feature data. It determines the environmental natural rhythm data based on current time and location information and then correlates the environmental natural rhythm data with… Time matching is performed on the physiological characteristic data of each window to establish a clear temporal correspondence between low-frequency, slowly changing environmental natural rhythm data and high-frequency window physiological characteristic data within the corresponding time range. This yields time-aligned heterogeneous data corresponding to each sliding time window, thereby reducing the problem of direct correspondence between the two types of data due to differences in change frequency and time scale. Window physiological characteristic data and environmental natural rhythm data are obtained from the time-aligned heterogeneous data corresponding to the target sliding time window. The window physiological characteristic data is standardized, and seasonal calibration operators and diurnal time calibration operators are determined based on the environmental natural rhythm data. Simultaneously, a preset physiological rhythm correction algorithm is obtained. This process transforms environmental circadian rhythm data from merely supplementary data into calibration operators corresponding to the current solar term and time of day. This allows for differentiated calibration of physiological characteristics based on the environmental time context of the target sliding time window. A cascade transformation is performed on the standardized window physiological characteristic data using seasonal, diurnal, and physiological rhythm correction operators. This ensures that environmental circadian rhythm information at different time scales is applied to the window physiological characteristic data through the corresponding calibration operators. Normalization then converts the cascaded multidimensional data to a unified data scale, resulting in standardized time-series alignment corresponding to the target sliding time window. Feature vector; furthermore, this application performs windowing processing on the physiological feature stream, performs time matching between the windowed physiological feature data and the environmental natural rhythm data, and further determines the corresponding calibration operator based on the matched environmental natural rhythm data to perform cascade calibration and normalization processing on the windowed physiological feature data, so that the high-frequency physiological feature stream and the low-frequency environmental natural rhythm data are transformed from simple data paralleling or direct splicing into standardized time-series alignment features with time correspondence and operator calibration relationship, thereby improving the time alignment degree and data fusion degree of heterogeneous time-series data, and providing a unified data input for subsequent data processing based on the standardized time-series alignment feature vector. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating a heterogeneous data timing alignment method in one embodiment; Figure 2This is a structural block diagram of a heterogeneous data timing alignment device in one embodiment. Detailed Implementation

[0013] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0014] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0015] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0016] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0017] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0018] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0019] In one embodiment, such as Figure 1 As shown, a method for time-series alignment of heterogeneous data is provided, the method comprising: Step S100: Obtain the physiological feature stream corresponding to the target object, preprocess the physiological feature stream, and divide the preprocessed physiological feature stream according to a preset sliding time window to obtain the window physiological feature data corresponding to each sliding time window. Step S200: Determine the environmental natural rhythm data based on the current time information and location information, and perform time matching between the environmental natural rhythm data and the physiological characteristic data of each window to obtain the time-aligned heterogeneous data corresponding to each sliding time window; Step S300: Obtain window physiological characteristic data and environmental natural rhythm data from the time-aligned heterogeneous data corresponding to the target sliding time window, perform standardization processing on the window physiological characteristic data, determine the seasonal calibration operator and the day-night calibration operator based on the environmental natural rhythm data, and obtain the preset physiological rhythm correction operator. Step S400: Based on the seasonal calibration operator, day and night time calibration operator, and physiological rhythm correction operator, perform cascade transformation and normalization on the standardized window physiological feature data to obtain the standardized temporal alignment feature vector corresponding to the target sliding time window.

[0020] This application obtains the physiological feature stream corresponding to the target object, preprocesses the physiological feature stream, and divides the preprocessed physiological feature stream according to a preset sliding time window. This converts the continuously generated physiological feature stream into window physiological feature data corresponding to different time ranges. While reducing the interference of missing and abnormal data on subsequent processing, it provides a unified time window basis for establishing a time correspondence between environmental natural rhythm data and physiological feature data. Furthermore, it determines the environmental natural rhythm data based on current time and location information and performs time matching between the environmental natural rhythm data and the physiological feature data of each window. This approach allows for the establishment of a clear temporal correspondence between low-frequency, slowly varying environmental circadian rhythm data and high-frequency window physiological characteristic data within the corresponding time range. This yields time-aligned heterogeneous data corresponding to each sliding time window, reducing the difficulty of directly corresponding the two types of data due to differences in frequency and time scale. By acquiring window physiological characteristic data and environmental circadian rhythm data from the time-aligned heterogeneous data corresponding to the target sliding time window, standardizing the window physiological characteristic data, and determining seasonal and diurnal time-based calibration operators based on the environmental circadian rhythm data, while simultaneously acquiring a preset physiological rhythm correction operator, the environmental circadian rhythm... The data is no longer just supplementary data for processing, but is transformed into calibration operators corresponding to the current solar term and time of day. This enables differentiated calibration of physiological characteristics based on the environmental time background corresponding to the target sliding time window. The standardized window physiological characteristic data is cascaded through seasonal calibration operators, diurnal time calibration operators, and physiological rhythm correction operators. This allows environmental natural rhythm information at different time scales to act on the window physiological characteristic data through the corresponding calibration operators. The cascaded multidimensional data is then converted to a unified data scale through normalization processing to obtain the standardized time-series aligned feature vector corresponding to the target sliding time window. Furthermore, this application performs windowing processing on the physiological feature stream, matches the windowed physiological feature data with the environmental natural rhythm data in time, and further determines the corresponding calibration operator based on the matched environmental natural rhythm data to perform cascade calibration and normalization processing on the windowed physiological feature data. This transforms the high-frequency physiological feature stream and low-frequency environmental natural rhythm data from simple data paralleling or direct splicing into standardized time-series alignment features with time correspondence and operator calibration relationships, thereby improving the time alignment degree and data fusion degree of heterogeneous time-series data, and providing a unified data input for subsequent data processing based on the standardized time-series alignment feature vector.

[0021] In one embodiment, step S100 involves acquiring the physiological feature stream corresponding to the target object, preprocessing the physiological feature stream, and dividing the preprocessed physiological feature stream according to a preset sliding time window, including: Step S110: Obtain the heart rate variability time series index, real-time heart rate and triaxial acceleration data corresponding to the target object, determine the number of steps and exercise intensity based on the triaxial acceleration data, and arrange them according to the acquisition time corresponding to each data to form a physiological feature stream; Step S120: Determine the missing data corresponding to each physiological feature in the physiological feature stream, and perform local forward first-order difference interpolation on the missing data to obtain the missing physiological feature stream; Step S130: Based on the preset effective physiological range corresponding to each physiological feature, determine the abnormal physiological feature data from the missing physiological feature stream, and remove the data frames corresponding to the abnormal physiological feature data to obtain the preprocessed physiological feature stream. Step S140: According to the preset window length and preset sliding step size corresponding to the preset sliding time window, the preprocessed physiological feature stream is divided in chronological order, and the heart rate variability time series index, real-time heart rate, step count and exercise intensity falling into the same sliding time window are determined as the window physiological feature data of the corresponding sliding time window.

[0022] In this embodiment, in step S110, the target object wears a smart wearable device with physiological parameter acquisition function. The smart wearable device continuously collects the physiological data and motion data corresponding to the target object through its built-in physiological sensors and motion sensors, and sends the collected data to the smart terminal through Bluetooth communication.

[0023] The heart rate variability time-series index includes the standard deviation of all RR intervals (SDNN) and the root mean square standard deviation (RMSSD) of the differences between adjacent RR intervals. SDNN characterizes the overall dispersion of RR intervals over a certain time range, while RMSSD characterizes the short-term variation between consecutive adjacent RR intervals. The real-time heart rate characterizes the heart rate state of the target object at the corresponding acquisition time. The triaxial acceleration data characterizes the changes in motion acceleration of the target object in three mutually perpendicular directions. In this embodiment, SDNN and RMSSD are used as two independent feature components in the heart rate variability time-series index for subsequent processing.

[0024] For the triaxial acceleration data, the motion state of the target object is determined based on the changes in triaxial acceleration at multiple consecutive sampling times, and the step count and motion intensity are statistically obtained from the motion state. The step count characterizes the amount of walking activity generated by the target object within a corresponding time range, and the motion intensity characterizes the overall motion amplitude of the target object within the corresponding time range.

[0025] After obtaining the time-series indicators of heart rate variability, real-time heart rate, steps, and exercise intensity, the corresponding acquisition time for each physiological characteristic is retained and arranged in order of acquisition time from earliest to latest, thereby forming a stream of physiological characteristics that changes continuously over time.

[0026] In step S120, due to the impact of Bluetooth communication delay, wireless interference or data packet loss on data transmission between the smart wearable device and the smart terminal, there may be a situation in the continuous physiological feature stream where no valid physiological feature value is obtained at a certain acquisition moment.

[0027] The temporal continuity between adjacent valid data is checked according to the collection time corresponding to each physiological characteristic. When there is no corresponding valid physiological characteristic value at the preset sampling position, the position is determined as the missing data position.

[0028] For the missing data location, adjacent valid physiological feature data preceding the missing data location are obtained, and local forward first-order difference interpolation is performed based on the change trend between the aforementioned adjacent valid physiological feature data to obtain the filling physiological feature value corresponding to the missing data location. In this embodiment, local forward first-order difference interpolation refers to obtaining at least two valid physiological feature values ​​closest to the missing data location preceding the missing data location, determining the local change amount of the corresponding physiological feature based on the at least two valid physiological feature values, and determining the filling physiological feature value corresponding to the missing data location based on the valid physiological feature value closest to the missing data location and the local change amount.

[0029] Among them, various physiological characteristics are identified and interpolated within their respective time series to ensure the continuity of the time series corresponding to heart rate variability indicators, real-time heart rate, steps, and exercise intensity.

[0030] After filling in each missing position, the physiological feature values ​​obtained from the filling and the original valid physiological feature values ​​are rearranged according to the corresponding acquisition time to obtain the missing physiological feature stream.

[0031] In step S130, corresponding effective physiological ranges are preset for different types of physiological characteristics. The preset effective physiological ranges are used to determine whether the corresponding physiological characteristic values ​​are within a reasonable range that can be generated during the normal physiological activities of the target object.

[0032] Each physiological feature data in the missing physiological feature stream is acquired frame by frame, and the temporal index of heart rate variability, real-time heart rate, step count and exercise intensity are compared with their respective preset physiological effective ranges.

[0033] When any physiological characteristic value exceeds the corresponding preset physiological effective range, the physiological characteristic value is determined as abnormal physiological characteristic data, and the data frame corresponding to the abnormal physiological characteristic data is determined as an abnormal data frame.

[0034] For outliers caused by sensor detachment from target object, violent body movement, communication glitches, or transient sensor disturbances, a whole-frame rejection method is used to prevent outlier data from participating in subsequent sliding time window division and timing alignment calculations.

[0035] After removing abnormal data frames, the remaining valid data and their corresponding acquisition times are retained to obtain the preprocessed physiological feature stream.

[0036] In step S140, the preset sliding time window has a pre-set window length and a sliding step size. The window length is used to determine the time range covered by each sliding time window, and the sliding step size is used to determine the starting time interval between two adjacent sliding time windows.

[0037] According to the preset window length, a first sliding time window is determined on the time axis corresponding to the physiological feature stream, and the heart rate variability time series indicators, real-time heart rate, steps and exercise intensity that fall within the first sliding time window are assigned to the first sliding time window.

[0038] Subsequently, the sliding time window is moved along the time axis according to the preset sliding step size to form multiple subsequent sliding time windows, and the physiological characteristic data corresponding to each sliding time window is determined in the same way.

[0039] Therefore, the physiological feature data corresponding to each sliding time window clearly includes the temporal index of heart rate variability, real-time heart rate, steps and exercise intensity within that time range, so that all the physiological features obtained and derived in step S110 can enter the subsequent temporal alignment process.

[0040] In one embodiment, step S200, determining environmental natural rhythm data based on current time information and location information, includes: Step S210: Obtain the current Greenwich Mean Time stamp and the current location information corresponding to the target object, and determine the current local time information corresponding to the location of the target object based on the current Greenwich Mean Time stamp and the current location information; Step S220: Determine the corresponding Julian day based on the current Greenwich Mean Time stamp, and determine the solar longitude degree corresponding to the current time based on the Julian day; Step S230: Determine the current solar term index based on the solar ecliptic longitude, and encode the current solar term index using periodic features to obtain the solar term features; Step S240: Determine the current time index based on the current local time information, and encode the current time index using periodic features to obtain the time feature; Step S250: Establish a correlation between the solar term characteristics and the time characteristics and the corresponding current time information to form the environmental natural rhythm data.

[0041] In this embodiment, in step S210, the smart terminal obtains the current Greenwich Mean Time stamp through the system clock and obtains the location information corresponding to the current location of the target object through the terminal positioning module. The location information is at least used to determine the time zone relationship corresponding to the current location of the target object.

[0042] Based on the current location information, the time zone corresponding to the location of the target object is determined, and the current Greenwich Mean Time is converted according to the time offset relationship between the time zone and Greenwich Mean Time to obtain the current local time information corresponding to the location of the target object.

[0043] The current local time information includes at least the current date, current hour, and current minute, providing a time basis for determining the current hour index.

[0044] In step S220, the Julian date is converted based on the Gregorian calendar date and time corresponding to the current Greenwich timestamp to obtain the Julian date corresponding to the current moment.

[0045] After obtaining the Julian Day, the solar ecliptic longitude corresponding to the Earth's current orbital position is determined according to a pre-defined astronomical calendar relationship. This solar ecliptic longitude is used to characterize the current position within a complete annual natural cycle.

[0046] In this embodiment, the solar longitude is used instead of just the Gregorian calendar month to determine the solar term status, so that the characteristics of subsequent solar terms can be determined according to the natural annual cycle.

[0047] In step S230, the complete solar ecliptic longitude cycle is divided according to the ecliptic longitude intervals corresponding to the twenty-four solar terms, and the current solar term index is determined based on the ecliptic longitude interval to which the current solar ecliptic longitude degree belongs.

[0048] After obtaining the current solar term index, the current solar term index is encoded with periodic features to convert the discrete solar term states into solar term features that can participate in subsequent data processing.

[0049] In this embodiment, periodic cosine encoding is used to convert the current solar term index, so that the solar terms that are adjacent at the beginning and end of the annual cycle maintain the periodic continuity in the encoding space, and reduce the numerical abrupt changes at the boundary position of the annual cycle when sequential integer encoding is used.

[0050] In other embodiments, a one-hot encoding method is used to encode the current solar term index, mapping each of the twenty-four solar terms to a different activation position in a twenty-four-dimensional feature vector.

[0051] In step S240, the time interval to which the current moment belongs is determined based on the hour and minute in the current local time information. In this embodiment, a natural day is divided into twelve consecutive time intervals, and each time interval corresponds to a time index.

[0052] After determining the current time index, the current time index is encoded with periodic features to obtain the time feature.

[0053] In this embodiment, periodic feature encoding is used to maintain the data continuity between adjacent time periods.

[0054] In other embodiments, a twelve-dimensional one-hot encoding is used to represent the current time state, so that the feature dimension corresponding to the current time is in an active state, while the feature dimensions corresponding to other times are in an inactive state.

[0055] In step S250, the solar term features obtained in step S230 and the time features obtained in step S240 are jointly set as the environmental natural rhythm features corresponding to the current time information, and the time information corresponding to both is retained.

[0056] As the current time continuously changes, the solar term characteristics and time-of-day characteristics are redefined according to a preset update time, thereby forming environmental natural rhythm data that changes over time. In this embodiment, the current time information is not limited to a single fixed moment, but is time information that is continuously updated during the data processing. Each time the update time arrives, the corresponding current time information is associated with and saved with the solar term characteristics and time-of-day characteristics determined based on the current time information, so that the environmental natural rhythm data includes solar term characteristics and time-of-day characteristics corresponding to different times.

[0057] Therefore, the environmental natural rhythm data includes at least solar term features and time features with clear time correspondence, enabling subsequent steps to determine the environmental natural rhythm state that should be matched for each sliding time window based on the time range corresponding to that window.

[0058] In one embodiment, step S200, which involves time-matching the environmental natural rhythm data with the physiological characteristic data of each window, includes: Step S260: Determine the window time range corresponding to the physiological characteristic data of each window based on the window start time and window end time corresponding to each sliding time window; Step S270: Based on the current time information associated with the environmental natural rhythm data, determine the solar term characteristics and time characteristics corresponding to each of the window time ranges; Step S280: Establish time association between the physiological feature data of each window and the corresponding solar term features and time features to obtain the time-aligned heterogeneous data corresponding to each sliding time window.

[0059] In this embodiment, in step S260, after the sliding time window of the physiological feature flow is divided in step S140, each sliding time window has a corresponding window start time and window end time.

[0060] The continuous time interval between the start time and end time of any sliding time window is defined as the window time range corresponding to the sliding time window, and the window time range is associated with the window physiological characteristic data corresponding to the sliding time window.

[0061] Therefore, each window's physiological characteristic data has a clear time interval identifier, enabling environmental natural rhythm data to be matched with the corresponding window's physiological characteristic data according to the same time benchmark.

[0062] In step S270, the window time range corresponding to each sliding time window is obtained sequentially, and the time correspondence between different environmental natural rhythm features and each window time range is determined based on the current time information corresponding to the solar term features and time features in the environmental natural rhythm data.

[0063] When the current time information corresponding to a certain solar term feature and time feature falls within the window time range corresponding to the target sliding time window, or when its effective time interval covers the target sliding time window, the solar term feature and time feature are determined as the environmental natural rhythm feature corresponding to the target sliding time window.

[0064] When the same solar term or time period continuously covers multiple consecutive sliding time windows, the same solar term or time period characteristics are associated with the corresponding physiological characteristic data of multiple windows, thereby adapting to the heterogeneous time scale characteristics where the rate of change of the natural rhythm of the environment is lower than the rate of change of physiological characteristic flow.

[0065] In step S280, for any sliding time window, the window physiological feature data obtained in step S140, the solar term features and time features determined in step S270 are jointly associated with the window identifier and window time range corresponding to the sliding time window.

[0066] In this embodiment, the time-aligned heterogeneous data is not a direct concatenation of physiological features and environmental natural rhythm features into a single numerical vector. Instead, it preserves the time correspondence between the physiological feature data of the window and the corresponding solar term features and time features under the same sliding time window.

[0067] Therefore, the time-aligned heterogeneous data corresponding to each sliding time window includes at least the corresponding window physiological characteristic data and environmental natural rhythm data. The subsequent step S300 can directly extract the two types of data from the time-aligned heterogeneous data corresponding to the target sliding time window and perform different data processing.

[0068] Through the above processing, a time correspondence is established between the high-frequency physiological feature streams that change at the millisecond or second level and the environmental natural rhythm data that change at the hourly, daily, or longer periodic levels under a unified sliding time window, thereby providing a data foundation for the subsequent calibration of physiological features by the environmental natural rhythm operator.

[0069] In one embodiment, step S300 involves standardizing the window physiological feature data, including: Step S310: Obtain the heart rate variability time series index, real-time heart rate, step count, and exercise intensity from the window physiological feature data corresponding to the target sliding time window; Step S320: Obtain the baseline mean and baseline standard deviation of the heart rate variability time series index, real-time heart rate, step count and exercise intensity respectively; Step S330: Based on the baseline mean and baseline standard deviation of each physiological characteristic, perform Z-score standardization on the time-series indicators of heart rate variability, real-time heart rate, steps and exercise intensity to obtain the corresponding standardized heart rate variability features, standardized heart rate features, standardized steps features and standardized exercise intensity features. Step S340: Arrange the standardized heart rate variability feature, standardized heart rate feature, standardized step count feature and standardized exercise intensity feature according to a preset feature order to obtain the standardized window physiological feature vector corresponding to the target sliding time window.

[0070] In this embodiment, in step S310, the target sliding time window to be processed for timing alignment calibration is determined from multiple sliding time windows, and the physiological feature data of the window is obtained from the timing alignment heterogeneous data corresponding to the target sliding time window.

[0071] Based on the data type identifiers in the window's physiological feature data, heart rate variability time series indicators, real-time heart rate, step count, and exercise intensity are extracted to obtain four types of physiological features to be standardized within the target sliding time window.

[0072] When the heart rate variability time series index includes SDNN and RMSSD, the corresponding heart rate variability feature data are formed according to a preset data organization method, and the fixed feature arrangement relationship between the heart rate variability and other physiological features is maintained in the subsequent standardization process.

[0073] In step S320, standardized benchmark parameters are pre-established for different types of physiological characteristics, with each type of physiological characteristic having a corresponding benchmark mean and benchmark standard deviation.

[0074] After determining the target sliding time window, the corresponding baseline mean and baseline standard deviation are read according to the data type of each physiological characteristic.

[0075] The baseline mean is used to characterize the central level of the corresponding physiological characteristic within the baseline data range, and the baseline standard deviation is used to characterize the scale of change of the corresponding physiological characteristic relative to the central level.

[0076] In this embodiment, the benchmark mean and benchmark standard deviation are obtained in advance from the data used to establish the system benchmark, and are used as standardized parameters of the corresponding physiological characteristics during the standardization process.

[0077] In step S330, for any physiological feature to be standardized, the relative deviation is determined based on the difference between the current feature value of the physiological feature and the corresponding benchmark mean, and the relative deviation is scaled according to the corresponding benchmark standard deviation to obtain the standardized feature value corresponding to the physiological feature.

[0078] By processing heart rate variability time-series indicators, real-time heart rate, steps, and exercise intensity in the manner described above, various physiological characteristics that originally had different physical units and different numerical ranges are converted into a unified dimensionless data expression form.

[0079] Thus, we obtained the standardized heart rate variability characteristics, standardized heart rate characteristics, standardized step count characteristics, and standardized exercise intensity characteristics, respectively.

[0080] In step S340, the feature arrangement order of each physiological feature in the standardized window physiological feature vector is preset, and the same arrangement order is maintained during model construction, operator parameter configuration and running processing.

[0081] According to the preset feature order, the four types of standardized physiological features obtained in step S330 are sequentially set at the corresponding feature positions to obtain the standardized window physiological feature vector corresponding to the target sliding time window. In this embodiment, the standardized heart rate variability feature includes mutually independent standardized SDNN feature components and standardized RMSSD feature components. Therefore, the standardized window physiological feature vector is respectively set with standardized SDNN feature components, standardized RMSSD feature components, standardized heart rate features, standardized step count features, and standardized exercise intensity features.

[0082] Therefore, the heart rate variability time series indicators, real-time heart rate, steps and exercise intensity extracted in step S310 have all undergone explicit standardization and are finally included in the standardized window physiological feature vector. There is no physiological feature data that is acquired separately but not involved in subsequent calculations.

[0083] In one embodiment, step S300, determining the seasonal calibration operator based on the environmental natural rhythm data, includes: Step S351: Obtain the solar term characteristics from the environmental natural rhythm data corresponding to the target sliding time window, and determine the target solar term state based on the solar term characteristics; Step S352: Based on the target solar term state, obtain the target seasonal sensitivity weights corresponding to each physiological characteristic from the preset seasonal sensitivity parameter library; Step S353: According to the arrangement order of each physiological feature in the standardized window physiological feature vector, set each target seasonal sensitivity weight as a diagonal element to construct the seasonal calibration operator corresponding to the target sliding time window.

[0084] In this embodiment, in step S351, environmental natural rhythm data that has a time correspondence with the window physiological characteristic data is obtained from the time-aligned heterogeneous data corresponding to the target sliding time window, and solar term features are extracted from the environmental natural rhythm data.

[0085] When the solar term features are encoded using one-hot encoding, the corresponding solar term index is determined based on the feature position in the active state, and the solar term corresponding to that index is determined as the target solar term state.

[0086] When the solar term features are encoded using periodic continuous coding, the target solar term status is determined based on the annual periodic position corresponding to the periodic continuous coding.

[0087] In step S352, the preset seasonal sensitivity parameter library stores the correspondence between different solar terms and the seasonal sensitivity weights corresponding to various physiological characteristics.

[0088] Based on the target solar term state obtained in step S351, query the preset seasonal sensitivity parameter library to obtain the target seasonal sensitivity weights corresponding to the heart rate variability feature, heart rate feature, step count feature and exercise intensity feature under the target solar term state.

[0089] In this embodiment, the seasonal sensitivity weights are derived from the offline parameter training process. During training, parameters corresponding to different seasons or solar terms are used as scaling parameters to be learned. The scaling parameters are adjusted according to the residuals of the model processing results under different environmental conditions, and after training convergence, they are solidified as seasonal sensitivity parameters used in the system operation.

[0090] In step S353, based on the arrangement of each physiological feature in the standardized window physiological feature vector determined in step S340, the target seasonal sensitivity weights obtained in step S352 are arranged in the same order.

[0091] Each target seasonal sensitivity weight is set at a diagonal position corresponding to the diagonal matrix, and the remaining off-diagonal positions are set to preset values ​​that do not participate in cross-feature coupling, thereby obtaining a seasonal calibration operator that matches the dimension of the standardized window physiological feature vector. In this embodiment, the preset value of not participating in cross-feature coupling is zero.

[0092] Therefore, each diagonal element in the seasonal calibration operator corresponds one-to-one with a physiological feature component in the standardized window physiological feature vector, enabling differentiated scaling calibration corresponding to the current target solar term state to be implemented for different physiological features in the subsequent cascade transformation process.

[0093] In one embodiment, step S300, determining the diurnal time calibration operator based on the environmental natural rhythm data and obtaining a preset physiological rhythm correction operator, includes: Step S361: Obtain time characteristics from the environmental natural rhythm data corresponding to the target sliding time window, and determine the target time state based on the time characteristics; Step S362: Based on the target time state, obtain the target time calibration weights corresponding to each physiological characteristic from the preset time calibration parameter library; Step S363: According to the arrangement order of each physiological feature in the standardized window physiological feature vector, set each target time calibration weight as a diagonal element to construct the day and night time calibration operator corresponding to the target sliding time window; Step S364: Based on the target solar term state and the target time state, obtain the corresponding physiological rhythm correction weight matrix from the preset physiological rhythm correction parameter library, and determine the physiological rhythm correction weight matrix as the physiological rhythm correction operator.

[0094] In this embodiment, time features are extracted from the environmental natural rhythm data corresponding to the target sliding time window.

[0095] When the time feature uses one-hot encoding, the current time index is determined based on the position of the feature in the active state, and the time corresponding to that time index is determined as the target time state.

[0096] When the time feature uses periodic continuous encoding, the target time state is determined based on the position of the periodic continuous encoding in the daily cycle.

[0097] In step S362, based on the target time state, the target time calibration weights corresponding to each physiological characteristic are obtained from the preset time calibration parameter library.

[0098] Specifically, the preset time calibration parameter library pre-stores the correspondence between different target time states and the time calibration weights corresponding to each physiological characteristic.

[0099] Based on the target time state obtained in step S361, query the preset time calibration parameter library to obtain the target time calibration weights corresponding to the standardized heart rate variability feature, standardized heart rate feature, standardized step count feature, and standardized exercise intensity feature respectively under the current target time state.

[0100] The target time calibration weight is used to represent the relative degree of compensation required for each physiological characteristic of the target object under different day and night periods, so that the same physiological characteristic value corresponds to different time background calibration results under different target time states.

[0101] In step S363, according to the arrangement order of each physiological feature in the standardized window physiological feature vector, the target time calibration weights are set as diagonal elements to construct the day and night time calibration operator corresponding to the target sliding time window.

[0102] Specifically, the target time calibration weights obtained in step S362 are arranged according to the preset feature arrangement order of each physiological feature in step S340.

[0103] The calibration weights for each target time period are set as diagonal elements in a diagonal matrix that match the corresponding physiological feature components, resulting in a diurnal time period calibration operator that matches the dimension of the standardized window's physiological feature vector. In this embodiment, all off-diagonal elements in the diurnal time period calibration operator, except for the diagonal elements corresponding to the calibration weights for each target time period, are zero.

[0104] Therefore, the day-night time calibration operator can apply calibration effects corresponding to the target time state to different physiological characteristics in the subsequent cascade transformation process.

[0105] In this embodiment, a physiological rhythm correction parameter library is pre-established before executing step S364. This physiological rhythm correction parameter library is used to store physiological rhythm correction weight matrices corresponding to different environmental natural rhythm states.

[0106] Specifically, during the system configuration phase, based on the correlation between various physiological characteristics under different environmental natural rhythm states, a corresponding physiological rhythm correction weight matrix is ​​pre-determined, and the physiological rhythm correction weight matrix is ​​then associated with the corresponding environmental natural rhythm state and stored in the physiological rhythm correction parameter library. The environmental natural rhythm state is characterized at least by the solar term state and the time of day state.

[0107] The dimension of the physiological rhythm correction weight matrix matches the feature dimension of the standardized window physiological feature vector, wherein each matrix element is used to characterize the correction relationship between different physiological features under the corresponding environmental natural rhythm state. The physiological rhythm correction weight matrix is ​​predetermined and fixed before system operation, and is not trained in real time during the heterogeneous data temporal alignment process.

[0108] Each physiological rhythm correction weight matrix is ​​indexed with its corresponding solar term and time state, thus forming the preset physiological rhythm correction parameter library. During system operation, based on the target solar term and time state corresponding to the current target sliding time window, a physiological rhythm correction weight matrix matching the current environmental natural rhythm state is determined from the preset physiological rhythm correction parameter library.

[0109] In step S364, based on the target solar term state and the target time state, the corresponding physiological rhythm correction weight matrix is ​​obtained from the preset physiological rhythm correction parameter library, and the physiological rhythm correction weight matrix is ​​determined as the physiological rhythm correction operator.

[0110] Specifically, the target solar term state determined in step S351 and the target time state determined in step S361 are used together as parameter query conditions. Based on the parameter query conditions, a physiological rhythm correction weight matrix that matches the current environmental natural rhythm state is queried from the preset physiological rhythm correction parameter library.

[0111] Each element in the physiological rhythm correction weight matrix has a preset correspondence with each physiological feature in the standardized window physiological feature vector. The diagonal elements are used to characterize the correction relationship of the corresponding physiological feature itself, and the off-diagonal elements are used to characterize the cross correction relationship between different physiological features.

[0112] The physiological rhythm correction weight matrix obtained from the query is determined as the physiological rhythm correction operator corresponding to the target sliding time window, and in the subsequent step S430, the second feature transformation is performed on the environmental calibration physiological feature vector after environmental calibration according to the physiological rhythm correction operator.

[0113] Therefore, the same combination of physiological characteristics corresponds to different physiological rhythm correction relationships under different target solar terms and target time states, so that the physiological rhythm correction operator can keep matching with the environmental natural rhythm state corresponding to the current target sliding time window.

[0114] In one embodiment, step S400 involves performing a cascaded transformation on the standardized windowed physiological characteristic data based on the seasonal calibration operator, the diurnal time calibration operator, and the physiological rhythm correction operator, including: Step S410: Perform a composite operation based on the seasonal calibration operator and the day-night time calibration operator to obtain the environmental calibration composite operator; Step S420: Perform a first feature transformation on the standardized window physiological feature vector according to the environmental calibration composite operator to obtain the environmental calibration physiological feature vector; Step S430: Perform a second feature transformation on the environmental calibration physiological feature vector according to the physiological rhythm correction operator to obtain the cascaded calibration feature vector.

[0115] In this embodiment, in step S410, a composite operation is performed based on the seasonal calibration operator and the day-night time calibration operator to obtain the environmental calibration composite operator.

[0116] Specifically, the seasonal calibration operator obtained in step S353 and the day-night time calibration operator obtained in step S363 are obtained.

[0117] Since both operators are constructed according to the order of physiological features in the standardized window physiological feature vector, they have a matching matrix dimension and feature position correspondence.

[0118] The seasonal calibration operator and the day-night time calibration operator are combined at corresponding positions to form a joint environmental calibration effect at the same feature position, so that the seasonal sensitivity weight and time calibration weight corresponding to the same physiological feature form a joint environmental calibration effect, thus obtaining the environmental calibration composite operator.

[0119] In this embodiment, an environmental calibration composite operator is formed by cascading matrix operations of seasonal calibration operator and day / night time calibration operator, thereby simultaneously introducing environmental background corrections to physiological characteristics based on the target solar term state and target time state.

[0120] In step S420, the standardized window physiological feature vector is subjected to a first feature transformation according to the environmental calibration composite operator to obtain the environmental calibration physiological feature vector.

[0121] Specifically, the standardized window physiological feature vector obtained in step S340 is used as the input of the first feature transformation, and the environmental calibration composite operator obtained in step S410 is used to perform corresponding calibration transformation on each standardized physiological feature component.

[0122] After the first feature transformation, the standardized heart rate variability feature, standardized heart rate feature, standardized step count feature, and standardized exercise intensity feature are adjusted by the environmental weights corresponding to the current target solar term state and the target time state, respectively.

[0123] After completing the above environmental calibration, the multiple physiological feature components are organized according to their original feature arrangement order to obtain the environmental calibration physiological feature vector.

[0124] In step S430, the environmental calibration physiological feature vector is subjected to a second feature transformation according to the physiological rhythm correction operator to obtain the cascaded calibration feature vector.

[0125] Specifically, the environmental calibration physiological feature vector obtained in step S420 is used as the input of the second feature transformation, and the feature components in the environmental calibration physiological feature vector are subjected to matrix transformation according to the physiological rhythm correction operator obtained in step S364.

[0126] Through the matrix transformation, based on the already completed seasonal and time-based background calibration, a cross-correction relationship between different physiological characteristics under the current spatiotemporal background is further introduced.

[0127] The physiological feature components obtained from the second feature transformation are arranged according to their corresponding positions to form the cascaded calibration feature vector.

[0128] Thus, the standardized window physiological feature vector is successively processed by the environmental calibration composite operator and the physiological rhythm correction operator to form a cascaded processing result of seasonal background calibration, time background calibration and physiological feature cross-correction.

[0129] In one embodiment, step S400, normalizing the cascaded calibration feature vector, includes: Step S440: Obtain each level of cascaded calibration feature component in the cascaded calibration feature vector, and perform exponential transformation on each of the cascaded calibration feature components to obtain the exponential transformation result corresponding to each of the cascaded calibration feature components; Step S450: Accumulate the results of each exponential transformation to obtain the normalized benchmark value; Step S460: Determine the normalized characteristic values ​​corresponding to each stage of the joint calibration characteristic component based on the exponential transformation results and the normalized reference value, respectively; Step S470: Arrange the normalized feature values ​​according to the order of the cascaded calibration feature components in the cascaded calibration feature vector to obtain the normalized time alignment feature vector corresponding to the target sliding time window.

[0130] In this embodiment, in step S440, each level of cascade calibration feature component in the cascade calibration feature vector is obtained, and each of the cascade calibration feature components is subjected to exponential transformation to obtain the exponential transformation result corresponding to each of the cascade calibration feature components.

[0131] Specifically, the cascaded calibration feature components are extracted sequentially from the cascaded calibration feature vector obtained in step S430 according to the feature arrangement order.

[0132] An exponential transformation is performed on each cascaded calibration feature component to convert the original cascaded calibration feature value into an exponential transformation result greater than zero, so that subsequent feature components can be proportionally normalized based on a unified positive value scale.

[0133] Based on the arrangement of each cascade calibration feature component in the original cascade calibration feature vector, the correspondence between the results of each exponential transformation is preserved.

[0134] In step S450, the results of each exponential transformation are accumulated to obtain a normalized benchmark value.

[0135] Specifically, all the exponential transformation results obtained in step S440 are accumulated, and the accumulated result is determined as the normalized reference value corresponding to the current target sliding time window.

[0136] The normalized benchmark value is used to characterize the overall numerical scale of all feature components in the current cascaded calibration feature vector after exponential transformation, and serves as a unified benchmark for determining the normalization result of a single feature component in step S460.

[0137] In step S460, the normalized characteristic values ​​corresponding to each stage of the joint calibration characteristic component are determined based on the results of each exponential transformation and the normalized reference value.

[0138] Specifically, for any cascaded calibration feature component, the exponential transformation result corresponding to the cascaded calibration feature component is proportionally processed with the normalized reference value obtained in step S450 to obtain the normalized feature value corresponding to the cascaded calibration feature component.

[0139] All cascaded calibration feature components are processed sequentially in the same manner, so that the resulting multiple normalized feature values ​​together form a unified dimensionless feature distribution.

[0140] In this embodiment, steps S440 to S460 correspond to the Softmax normalization process. This process transforms the multidimensional data after multi-operator cascade calibration into a data representation with a uniform scale.

[0141] In step S470, the normalized feature values ​​are arranged according to the order of the cascaded calibration feature components in the cascaded calibration feature vector to obtain the normalized time alignment feature vector corresponding to the target sliding time window.

[0142] Specifically, the original feature arrangement order of the cascaded calibration feature vector in step S430 is maintained, and the normalized feature values ​​obtained in step S460 are sequentially set to the corresponding feature positions to obtain the standardized time-aligned feature vector.

[0143] Each feature component in the standardized time-aligned feature vector has been successively processed by physiological feature standardization, seasonal calibration, time calibration, physiological rhythm correction and normalization. Therefore, it simultaneously contains physiological state change information within the target sliding time window and environmental natural rhythm calibration information corresponding to that time window.

[0144] Therefore, steps S440 to S470 further convert the cascaded calibration feature vector into a standardized time-aligned feature vector with a uniform scale, thus completing the time-alignment and standardized fusion process of high-frequency physiological feature stream and low-frequency environmental natural rhythm data in claim 1.

[0145] In one embodiment, the method further includes: Step S500: Input the standardized temporally aligned feature vector into a preset manifold transformation layer; Step S510: Spatial projection of the normalized temporally aligned feature vector is performed on the manifold transformation matrix pre-fixed in the manifold transformation layer to obtain a low-dimensional mapped feature vector; Step S520: Determine the emotional valence and emotional arousal of the target object from the low-dimensional mapping feature vector; Step S530: Construct the target state coordinates of the target object in a preset two-dimensional state space based on the emotional valence and the emotional arousal.

[0146] In this embodiment, in step S500, after obtaining the standardized temporal alignment feature vector corresponding to the target sliding time window, the standardized temporal alignment feature vector is used as the input data for subsequent low-dimensional state space mapping and input into the preset manifold transformation layer deployed on the smart terminal.

[0147] The manifold transformation layer is used to transform high-dimensional normalized temporal alignment features that have incorporated information on environmental natural rhythms and physiological state changes into a low-dimensional continuous state space.

[0148] In this embodiment, the operating parameters of the manifold transformation layer are predetermined during the offline training phase and remain fixed during the terminal operation phase, so that the terminal can directly complete the forward spatial projection when performing real-time data processing.

[0149] In step S510, the standardized temporally aligned feature vector is spatially projected using the pre-fixed manifold transformation matrix in the manifold transformation layer to obtain a low-dimensional mapped feature vector.

[0150] Specifically, during the offline training phase, a mapping relationship between high-dimensional temporal alignment features and low-dimensional states is established based on training samples containing different environmental natural rhythm states and different physiological states, and the corresponding manifold transformation matrix is ​​obtained according to the training results.

[0151] After the model training is completed, the manifold transformation matrix is ​​frozen as a parameter and fixed to the manifold transformation layer in the smart terminal.

[0152] During the operation phase, the standardized temporal alignment feature vector input in step S500 is subjected to matrix operation with the manifold transformation matrix, so that the temporal alignment features in the high-dimensional feature space are projected to the preset low-dimensional continuous state space to obtain the low-dimensional mapping feature vector.

[0153] In step S520, the emotional valence and emotional arousal of the target object are determined from the low-dimensional mapping feature vector.

[0154] Specifically, in this embodiment, the low-dimensional mapping feature vector includes two state dimensions, where the first state dimension corresponds to emotional valence and the second state dimension corresponds to emotional arousal.

[0155] Obtain the feature value corresponding to the first state dimension in the low-dimensional mapping feature vector and determine it as the emotional valence corresponding to the current target sliding time window of the target object; obtain the feature value corresponding to the second state dimension and determine it as the corresponding emotional arousal.

[0156] Among them, emotional valence is used to characterize the positive or negative degree of the target object's current state, and emotional arousal is used to characterize the activation degree of the target object's current state.

[0157] Thus, the physiological feature information and environmental natural rhythm calibration information fused in the standardized temporal alignment feature vector are further converted into two continuous values ​​with clear state meanings.

[0158] In step S530, the target state coordinates of the target object in a preset two-dimensional state space are constructed based on the emotional valence and the emotional arousal.

[0159] Specifically, a two-dimensional state space is pre-established, including a first state dimension and a second state dimension, where the first state dimension corresponds to emotional valence and the second state dimension corresponds to emotional arousal.

[0160] The emotional valence obtained in step S520 is used as the coordinate value of the target state coordinate in the first state dimension, and the emotional arousal is used as the coordinate value of the target state coordinate in the second state dimension, thereby forming the target state coordinate corresponding to the current target sliding time window of the target object.

[0161] For multiple consecutive target sliding time windows, the corresponding target state coordinates are determined according to steps S500 to S530 respectively, so that the multiple target state coordinates form a continuous state change trajectory according to the time sequence of the sliding time window.

[0162] Therefore, steps S500 to S530 further map the obtained standardized temporal alignment feature vector into two-dimensional state coordinates corresponding to the target object, so that the physiological characteristics after environmental natural rhythm calibration can be provided to the subsequent data processing process in the form of continuous state.

[0163] In one embodiment, such as Figure 2 As shown, a heterogeneous data timing alignment device is also provided, the device comprising: The physiological feature acquisition module is used to acquire the physiological feature stream corresponding to the target object, preprocess the physiological feature stream, and divide the preprocessed physiological feature stream according to a preset sliding time window to obtain the window physiological feature data corresponding to each sliding time window. The heterogeneous data alignment module is used to determine environmental natural rhythm data based on current time and location information, and to perform time matching between the environmental natural rhythm data and the physiological characteristic data of each window to obtain time-aligned heterogeneous data corresponding to each sliding time window. The calibration operator acquisition module is used to acquire window physiological characteristic data and environmental natural rhythm data from the time-aligned heterogeneous data corresponding to the target sliding time window, perform standardization processing on the window physiological characteristic data, determine the seasonal calibration operator and the diurnal time calibration operator based on the environmental natural rhythm data, and acquire the preset physiological rhythm correction operator. The feature vector generation module is used to perform cascade transformation and normalization on the standardized window physiological feature data according to the seasonal calibration operator, the day and night time calibration operator and the physiological rhythm correction operator, so as to obtain the standardized temporal alignment feature vector corresponding to the target sliding time window.

[0164] In one embodiment, the physiological feature acquisition module is further configured to: acquire heart rate variability time-series indicators, real-time heart rate, and triaxial acceleration data corresponding to the target object; determine the number of steps and exercise intensity based on the triaxial acceleration data; and arrange the data according to the acquisition time corresponding to each data point to form a physiological feature stream; determine the missing data corresponding to each physiological feature in the physiological feature stream; perform local forward first-order difference interpolation on the missing data to obtain a missing-filled physiological feature stream; determine abnormal physiological feature data from the missing-filled physiological feature stream according to the preset effective physiological range corresponding to each physiological feature; remove the data frames corresponding to the abnormal physiological feature data to obtain a preprocessed physiological feature stream; and divide the preprocessed physiological feature stream according to the preset window length and preset sliding step size corresponding to the preset sliding time window, and determine the heart rate variability time-series indicators, real-time heart rate, number of steps, and exercise intensity falling into the same sliding time window as the window physiological feature data of the corresponding sliding time window.

[0165] In one embodiment, the heterogeneous data alignment module is further configured to: obtain the current Greenwich Mean Time (GMT) timestamp and the current location information corresponding to the target object, and determine the current local time information corresponding to the location of the target object based on the current GMT timestamp and the current location information; determine the corresponding Julian day based on the current GMT timestamp, and determine the solar longitude degree corresponding to the current time based on the Julian day; determine the current solar term index based on the solar longitude degree, and perform periodic feature encoding on the current solar term index to obtain solar term features; determine the current hour index based on the current local time information, and perform periodic feature encoding on the current hour index to obtain hour features; and establish a correlation between the solar term features and the hour features and the corresponding current time information to form the environmental natural rhythm data.

[0166] In one embodiment, the heterogeneous data alignment module is further configured to: determine the window time range corresponding to the physiological characteristic data of each window based on the window start time and window end time corresponding to each sliding time window; determine the solar term features and time features corresponding to each window time range based on the current time information associated with the environmental natural rhythm data; and establish time associations between the physiological characteristic data of each window and the corresponding solar term features and time features to obtain the time-series aligned heterogeneous data corresponding to each sliding time window.

[0167] In one embodiment, the calibration operator acquisition module is further configured to: acquire heart rate variability time-series indicators, real-time heart rate, steps, and exercise intensity from the window physiological feature data corresponding to the target sliding time window; acquire the baseline mean and baseline standard deviation corresponding to the heart rate variability time-series indicators, real-time heart rate, steps, and exercise intensity; perform Z-score standardization on the heart rate variability time-series indicators, real-time heart rate, steps, and exercise intensity according to the baseline mean and baseline standard deviation corresponding to each physiological feature, to obtain the corresponding standardized heart rate variability features, standardized heart rate features, standardized step count features, and standardized exercise intensity features; and arrange the standardized heart rate variability features, standardized heart rate features, standardized step count features, and standardized exercise intensity features according to a preset feature order to obtain the standardized window physiological feature vector corresponding to the target sliding time window.

[0168] In one embodiment, the calibration operator acquisition module is further configured to: acquire solar term features from the environmental natural rhythm data corresponding to the target sliding time window, and determine the target solar term state based on the solar term features; acquire target seasonal sensitivity weights corresponding to each physiological feature from a preset seasonal sensitivity parameter library based on the target solar term state; and set each target seasonal sensitivity weight as a diagonal element according to the arrangement order of each physiological feature in the standardized window physiological feature vector to construct the seasonal calibration operator corresponding to the target sliding time window.

[0169] In one embodiment, the calibration operator acquisition module is further configured to: acquire time features from the environmental natural rhythm data corresponding to the target sliding time window, and determine the target time state based on the time features; acquire target time calibration weights corresponding to each physiological feature from a preset time calibration parameter library based on the target time state; set each target time calibration weight as a diagonal element according to the arrangement order of each physiological feature in the standardized window physiological feature vector, and construct the day-night time calibration operator corresponding to the target sliding time window; acquire the corresponding physiological rhythm correction weight matrix from a preset physiological rhythm correction parameter library based on the target solar term state and the target time state, and determine the physiological rhythm correction weight matrix as the physiological rhythm correction operator.

[0170] In one embodiment, the feature vector generation module is further configured to: perform a composite operation on the seasonal calibration operator and the day-night time calibration operator to obtain an environmental calibration composite operator; perform a first feature transformation on the standardized window physiological feature vector according to the environmental calibration composite operator to obtain an environmental calibration physiological feature vector; and perform a second feature transformation on the environmental calibration physiological feature vector according to the physiological rhythm correction operator to obtain a cascaded calibration feature vector.

[0171] In one embodiment, the feature vector generation module is further configured to: obtain each level of cascaded calibration feature component in the cascaded calibration feature vector, and perform exponential transformation on each of the cascaded calibration feature components to obtain the exponential transformation result corresponding to each of the cascaded calibration feature components; accumulate the exponential transformation results to obtain a normalized reference value; determine the normalized feature value corresponding to each level of cascaded calibration feature component based on each of the exponential transformation results and the normalized reference value; and arrange the normalized feature values ​​according to the arrangement order of each of the cascaded calibration feature components in the cascaded calibration feature vector to obtain the standardized time alignment feature vector corresponding to the target sliding time window.

[0172] In one embodiment, the feature vector generation module is further configured to: input the standardized temporally aligned feature vector into a preset manifold transformation layer; spatially project the standardized temporally aligned feature vector through a pre-fixed manifold transformation matrix in the manifold transformation layer to obtain a low-dimensional mapped feature vector; determine the emotional valence and emotional arousal corresponding to the target object from the low-dimensional mapped feature vector; and construct the target state coordinates of the target object in a preset two-dimensional state space based on the emotional valence and the emotional arousal.

[0173] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0174] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0175] This application also provides a network device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above method embodiments.

[0176] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0177] This application provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the various method embodiments above.

[0178] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0179] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0180] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0181] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0182] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0183] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

[0184] One embodiment of this application also provides a computer terminal, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above-described methods.

[0185] The computer terminal may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above description is an example of a computer terminal and does not constitute a limitation on the computer terminal. It may include more or fewer components than described above, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0186] The processor referred to can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0187] In some embodiments, the memory may be an internal storage unit of the computer terminal, such as a hard drive or RAM. In other embodiments, the memory may be an external storage device of the computer terminal, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory may include both internal and external storage units of the computer terminal. The memory is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory can also be used to temporarily store data that has been output or will be output.

[0188] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0189] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for time-series alignment of heterogeneous data, characterized in that, The method includes: Obtain the physiological feature stream corresponding to the target object, preprocess the physiological feature stream, and divide the preprocessed physiological feature stream according to a preset sliding time window to obtain the window physiological feature data corresponding to each sliding time window; Based on the current time and location information, environmental natural rhythm data is determined, and the environmental natural rhythm data is time-matched with the physiological characteristic data of each window to obtain time-aligned heterogeneous data corresponding to each sliding time window. From the time-aligned heterogeneous data corresponding to the target sliding time window, obtain the window physiological characteristic data and environmental natural rhythm data, perform standardization processing on the window physiological characteristic data, determine the seasonal calibration operator and the day-night time calibration operator based on the environmental natural rhythm data, and obtain the preset physiological rhythm correction operator. Based on the seasonal calibration operator, the day-night time calibration operator, and the physiological rhythm correction operator, the standardized window physiological feature data are subjected to cascade transformation and normalization to obtain the standardized temporal alignment feature vector corresponding to the target sliding time window.

2. The heterogeneous data time-series alignment method according to claim 1, characterized in that, Obtain the physiological feature stream corresponding to the target object, preprocess the physiological feature stream, and divide the preprocessed physiological feature stream according to a preset sliding time window, including: Acquire the time-series index of heart rate variability, real-time heart rate and triaxial acceleration data corresponding to the target object, determine the number of steps and exercise intensity based on the triaxial acceleration data, and arrange them according to the acquisition time corresponding to each data to form a physiological feature stream; The missing data corresponding to each physiological feature in the physiological feature stream are determined, and the missing data are subjected to local forward first-order difference interpolation to obtain the missing physiological feature stream. Based on the preset effective physiological range corresponding to each physiological feature, abnormal physiological feature data are determined from the missing physiological feature stream, and the data frames corresponding to the abnormal physiological feature data are removed to obtain the preprocessed physiological feature stream. Based on the preset window length and preset sliding step size corresponding to the preset sliding time window, the preprocessed physiological feature stream is divided according to time sequence, and the heart rate variability time series index, real-time heart rate, step count and exercise intensity falling into the same sliding time window are determined as the window physiological feature data of the corresponding sliding time window.

3. The heterogeneous data time-series alignment method according to claim 2, characterized in that, Based on current time and location information, environmental natural rhythm data are determined, including: Obtain the current Greenwich Mean Time (GMT) timestamp and the current location information corresponding to the target object, and determine the current local time information corresponding to the location of the target object based on the current GMT timestamp and the current location information; The corresponding Julian day is determined based on the current Greenwich Mean Time stamp, and the solar longitude degree corresponding to the current time is determined based on the Julian day. The current solar term index is determined based on the solar ecliptic longitude, and the current solar term index is coded with periodic features to obtain the solar term features; The current time index is determined based on the current local time information, and the current time index is encoded with periodic features to obtain the time feature; The solar term characteristics and time characteristics are associated with the corresponding current time information to form the environmental natural rhythm data.

4. The heterogeneous data time-series alignment method according to claim 3, characterized in that, The environmental natural rhythm data is time-matched with the physiological characteristic data of each window, including: Based on the start and end times of each sliding time window, the window time range corresponding to the physiological characteristic data of each window is determined; Based on the current time information associated with the environmental natural rhythm data, determine the solar term characteristics and time characteristics corresponding to each of the window time ranges; The physiological feature data of each window are associated with the corresponding solar term features and time features to obtain the time-aligned heterogeneous data corresponding to each sliding time window.

5. The heterogeneous data time-series alignment method according to claim 4, characterized in that, The physiological feature data of the window are standardized, including: From the physiological feature data of the window corresponding to the target sliding time window, obtain the time series index of heart rate variability, real-time heart rate, step count and exercise intensity respectively; The baseline mean and baseline standard deviation of the heart rate variability time series index, real-time heart rate, steps, and exercise intensity are obtained respectively; Z-score standardization was performed on the time-series indicators of heart rate variability, real-time heart rate, steps, and exercise intensity based on the baseline mean and baseline standard deviation of each physiological characteristic to obtain the corresponding standardized heart rate variability features, standardized heart rate features, standardized steps features, and standardized exercise intensity features. The standardized heart rate variability feature, standardized heart rate feature, standardized step count feature, and standardized exercise intensity feature are arranged in a preset feature order to obtain the standardized window physiological feature vector corresponding to the target sliding time window.

6. The heterogeneous data time-series alignment method according to claim 5, characterized in that, The seasonal calibration operator is determined based on the environmental natural rhythm data, including: The solar term characteristics are obtained from the environmental natural rhythm data corresponding to the target sliding time window, and the target solar term state is determined based on the solar term characteristics; Based on the target solar term state, the target seasonal sensitivity weights corresponding to each physiological characteristic are obtained from the preset seasonal sensitivity parameter library; Based on the arrangement order of the physiological features in the standardized window physiological feature vector, the seasonal sensitivity weights of each target are set as diagonal elements to construct the seasonal calibration operator corresponding to the target sliding time window.

7. The heterogeneous data time-series alignment method according to claim 6, characterized in that, The diurnal time calibration operator is determined based on the environmental natural rhythm data, and a preset physiological rhythm correction operator is obtained, including: The time characteristics are obtained from the environmental natural rhythm data corresponding to the target sliding time window, and the target time state is determined based on the time characteristics; Based on the target time state, obtain the target time calibration weights corresponding to each physiological characteristic from the preset time calibration parameter library; According to the arrangement order of each physiological feature in the standardized window physiological feature vector, the target time calibration weights are set as diagonal elements to construct the day and night time calibration operator corresponding to the target sliding time window; Based on the target solar term state and the target time state, the corresponding physiological rhythm correction weight matrix is ​​obtained from the preset physiological rhythm correction parameter library, and the physiological rhythm correction weight matrix is ​​determined as the physiological rhythm correction operator.

8. A heterogeneous data timing alignment device, characterized in that, The device includes: The physiological feature acquisition module is used to acquire the physiological feature stream corresponding to the target object, preprocess the physiological feature stream, and divide the preprocessed physiological feature stream according to a preset sliding time window to obtain the window physiological feature data corresponding to each sliding time window. The heterogeneous data alignment module is used to determine environmental natural rhythm data based on current time and location information, and to perform time matching between the environmental natural rhythm data and the physiological characteristic data of each window to obtain time-aligned heterogeneous data corresponding to each sliding time window. The calibration operator acquisition module is used to acquire window physiological characteristic data and environmental natural rhythm data from the time-aligned heterogeneous data corresponding to the target sliding time window, perform standardization processing on the window physiological characteristic data, determine the seasonal calibration operator and the diurnal time calibration operator based on the environmental natural rhythm data, and acquire the preset physiological rhythm correction operator. The feature vector generation module is used to perform cascade transformation and normalization on the standardized window physiological feature data according to the seasonal calibration operator, the day and night time calibration operator and the physiological rhythm correction operator, so as to obtain the standardized temporal alignment feature vector corresponding to the target sliding time window.

9. A computer terminal, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images