Electric toothbrush user behavior analysis system based on big data

Abstract

The invention relates to the technical field of health monitoring, in particular to an electric toothbrush user behavior analysis system based on big data. The big data-based electric toothbrush user behavior analysis system comprises a cleaning behavior change extraction module, a behavior continuous segment alignment module, a rhythm trajectory offset detection module, a behavior rhythm linkage recognition module and an abnormal trend health prompt module. According to the method, the behavior change time period is screened through the vibration starting and ending time difference value, the abnormal period recognition precision is improved, the behavior segments are uniformly adjusted to the standard length and the proportional position is normalized, the alignment consistency of cross-day behavior data is enhanced, time migration and high-frequency time point clustering are used for constructing a time centralized distribution model, the rhythm migration trend is defined, and the accuracy of the abnormal period recognition is improved. The vibration alternating interval sequence is combined with a stability threshold value to recognize jump frequency abnormity, jump and drift information are linked to judge a health risk period, and deep fusion of multi-dimensional behavior characteristics and health trends is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring, and particularly to a system for analyzing the behavior of electric toothbrush users based on big data. Background Art

[0002] The technical field of health monitoring includes related technologies for collecting, analyzing, and evaluating information such as the physiological state, behavior characteristics, and living habits of individuals or groups, aiming to achieve long-term observation and dynamic tracking of health conditions. The core content of this technical field is to collect and process multi-source data collected by wearable devices, smart home products, or daily use terminals, extract the change trends of individual behavior patterns and physiological parameters, and then form basic support information such as health risk warnings and lifestyle intervention suggestions. Its overall technical framework usually covers data collection terminals, information transmission systems, big data processing platforms, and modeling tools for user feature analysis, and is widely used in extended medical services, home care support, and personalized health management.

[0003] Among them, the system for analyzing the behavior of electric toothbrush users based on big data refers to using built-in sensors and communication components to collect behavior information such as the duration, frequency, angle, and strength during the user's brushing process, uploading it to the cloud platform through a communication network, and aggregating the behavior data with the user identifier. In the data processing process, the system uses the brushing time series analysis method to statistically analyze the usage frequency at different time periods, and combines the usage habit classification standard to assign label attributes to the behavior characteristics of the user group. Subsequently, the individual brushing behavior pattern is extracted and classified by means of behavior trajectory modeling, and finally a user behavior database is established on the platform side to support subsequent statistical analysis and health trend identification. This system mainly relies on three methods: data sequence sorting, label classification, and behavior trajectory modeling to complete the systematic analysis of user behavior.

[0004] The prior art relies on the collection of basic information on brushing behavior and time series statistical analysis. Although it can classify and attribute the usage frequency and behavior characteristics, it lacks a trend-sensitive mechanism for identifying periodic behavior changes and is difficult to achieve timely early warning in the stage of continuous decrease or small fluctuations in behavior data. In the process of behavior data processing, there is a lack of a unified duration adjustment and proportional position normalization method for daily behavior segments, and the horizontal comparison consistency of behavior data is insufficient, resulting in low accuracy in the extraction of user behavior patterns. In the aspect of time concentration change detection, no offset trajectory or high-frequency time point attribution logic is constructed, so the identification of rhythmic disturbances depends on static time node judgment, missing the time series offset trend. In the identification of vibration rhythm changes, it is mostly limited to the immediate judgment of local vibration intensity fluctuation points and cannot reflect the continuous jump characteristics formed by vibration frequency in the time series. The existing health reminder mechanism is mainly based on abnormal single-point behavior or static physiological parameters, lacking the modeling of the linkage relationship between jump rhythm and time offset. The identification accuracy and timeliness of health abnormal cycles are both limited and difficult to support the early response to chronic behavior abnormalities and potential risks. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art and propose an electric toothbrush user behavior analysis system based on big data.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: The electric toothbrush user behavior analysis system based on big data includes:

[0007] The cleaning behavior change extraction module extracts the daily brushing duration, coverage area, and vibration intensity records, screens the decreasing cycles of the three types of data, judges whether the start-stop difference of vibration exceeds the limit, marks it as the cleaning habit change cycle, and generates a list of change time periods;

[0008] The behavior continuous segment alignment module reads the vibration start and stop points in the list of change time periods, extracts the daily brushing data, judges the head and tail time difference and uniformly adjusts it to the standard length, updates the proportional interval and then arranges it to generate a time series segment list;

[0009] The rhythm trajectory offset detection module calls the daily first brushing record in the time series segment list, extracts the daily high-frequency operations and classifies them into fixed time period groups, analyzes whether the time points cross time periods during the day, and generates a time concentration change performance;

[0010] The behavior rhythm linkage identification module based on the time concentration change performance extracts the cycles that coincide with the change time periods, analyzes the strong and weak vibration alternation sequence, judges whether it exceeds the stable limit, and generates a jump frequency anomaly flag sequence;

[0011] Based on the jump frequency anomaly flag sequence, the abnormal trend health prompt module counts the anomaly flag cycles, determines whether there is an extended interval and regional drift at the same time, classifies them as healthy abnormal cycles, and outputs the analysis results of the user behavior of the electric toothbrush.

[0012] As a further solution of the present invention, the cleaning habit change time period list includes a decreasing duration cycle, a decreasing coverage area cycle, a decreasing vibration intensity cycle, and an abnormal mark of vibration fluctuation difference. The time sequence segment list includes a standardized brushing behavior segment, a proportional interval position parameter, and a unified arrangement order index. The concentrated change performance of the brushing behavior time includes a first brushing time offset sequence, a high-frequency operation time point group, and a time period attribution distribution feature. The jump frequency anomaly flag sequence includes a strong and weak vibration alternating anomaly segment, a vibration interval overrun record, and an offset amplitude compliance section. The analysis results of the user behavior of the electric toothbrush include a jump interval extension flag, a drift area expansion flag, and a healthy abnormal cycle segment.

[0013] As a further solution of the present invention, the cleaning behavior change extraction module includes:

[0014] The data trend extraction sub-module obtains the user's daily brushing duration record, the number of covered area records, and the vibration intensity change record. Based on the continuous trend changes of each type of data in the three types of record sequences, by judging whether the data of consecutive multiple time nodes show a monotonically decreasing state, it screens the cycle segments in which the three types of records are continuously decreasing at the same time, and generates a combined decreasing cycle interval value;

[0015] The time difference judgment sub-module calls the vibration fluctuation time values at the start and end of brushing in each cycle in the combined decreasing cycle interval value, calculates the difference between the start and end times of the vibration intensity, and compares the difference with the time change limit range to identify the cycles that simultaneously meet the conditions of the decreasing trend and the vibration time difference exceeding the limit, and obtains a list of conforming change cycles;

[0016] The change time period generation sub-module constructs time period information indicating significant changes in the user's brushing behavior by extracting the start and end time nodes in the cycle as boundaries based on the list of conforming change cycles, and obtains a list of change time periods.

[0017] As a further solution of the present invention, the behavior continuous segment alignment module includes:

[0018] The start and end data extraction sub-module obtains the brushing records corresponding to the cycles in the change time period list, extracts the start and end times of the vibration of the daily brushing behavior in the cycle, calculates the start and end interval values of the daily behavior and binds them to the daily data tags, and generates a daily behavior interval data group;

[0019] The time period judgment sub-module identifies whether the absolute difference in the start and end spacing of the behavior is greater than the time threshold reference according to the daily behavior interval data group, determines the time period of the brushing behavior and extracts the corresponding recorded data, using the formula:

[0020]

[0021] Performs operations to obtain the behavior segment adjustment coefficient value, and based on this coefficient, performs a rationality screening on the behavior segment time interval to obtain an adjusted effective behavior interval group;

[0022] Among them, Q i is the behavior segment adjustment coefficient value on the i-th day, Z i is the number of daily brushing vibration wave bands, M i is the average daily vibration intensity, A i is the number of brushing coverage areas, D i is the total duration of the daily behavior segment, R is the fixed reference value of the behavior duration, F i,k is the time fluctuation value of the k-th section on the i-th day, C k is the coverage time constant corresponding to the k-th section;

[0023] The unified time sequence generation sub-module performs equidistant repositioning on the length of the daily behavior segment according to the adjusted effective behavior interval group, converts it into a proportional position under the standard time span, arranges the behavior segments in chronological order, establishes a time period sequence in a unified format, and obtains the behavior time sequence segment column.

[0024] As a further solution of the present invention, the rhythm trajectory offset detection module includes:

[0025] The first brushing time offset sub-module calls the daily first brushing record time in the behavior time sequence segment column, collects the time distribution data, judges whether there is a one-way offset of continuous time values according to the time sequence, and performs record marking according to the consistency of the offset direction to generate a continuous offset time group;

[0026] The high-frequency point grouping sub-module extracts the top three time points with the highest daily vibration frequency in the daily brushing records according to the continuous offset time group, and assigns them to fixed time period labels according to the hour segments where the time points are located, using the formula:

[0027]

[0028] Performs operations to obtain the time attribution difference value, which is used to identify the time period crossing difference between different days, judges whether a change feature of time period classification is formed, and obtains the grouped difference change value;

[0029] Among them, S t is the attribution difference value, N f is the number of high-frequency time points extracted daily, H uis the serial number of the hour period where the u-th high-frequency time point is located, D u is the corresponding brushing duration at this time point, V u is the corresponding vibration intensity value at this time point, G v is the total count of the time period to which the v-th day belongs, N d is the number of participating days in the cycle, U c is the number of time period types, F v is the total number of brushing behaviors on the v-th day;

[0030] The time concentration comparison sub-module combines the grouped difference change values to normalize and compare the repetition of the time period attribution frequency within consecutive cycles, determines whether the time distribution is concentrated in fixed segments, obtains the brushing time classification concentration index, and establishes the time concentration change performance of the brushing behavior.

[0031] As a further solution of the present invention, the behavior rhythm linkage recognition module includes:

[0032] The offset cycle screening sub-module extracts the cycle data with a coincidence degree of more than two days with the cleaning habit change section according to the marked cycle time range in the brushing behavior time concentration change performance, cross-compares the cycle boundaries, screens out the eligible time periods, and generates a combined offset cycle set;

[0033] The vibration sequence extraction sub-module calls the cycle data in the combined offset cycle set, extracts the vibration strength alternating numerical sequence of the daily brushing behavior, constructs a vibration change array in sequence and marks the corresponding time interval positions to obtain a continuous vibration interval sequence;

[0034] The jump frequency judgment sub-module is based on the continuous vibration interval sequence, judges the difference between adjacent vibration interval values, and compares with the stable vibration interval limit value to identify the vibration section with an excessive fluctuation range, using the formula:

[0035]

[0036] where:

[0037]

[0038] ||·||1 represents the 1-norm, that is, the sum of the absolute values of the elements of the vector difference;

[0039] operates to obtain the vibration interval offset value, and judges whether an abnormal jump behavior is formed according to the comparison result between the vibration interval offset value and the offset amplitude threshold, and obtains the jump frequency abnormal flag sequence;

[0040] where, F dev is the vibration interval offset value, N seg is the total number of jump sections, P ris the average vibration intensity of the r-th segment, T r is the time difference between the start and end times of this segment, C r is the number of cleaning areas covered in this segment, Θ is the upper threshold of the vibration stabilization time interval, D r is the behavior duration of the r-th segment.

[0041] As a further solution of the present invention, the abnormal trend health prompt module includes:

[0042] The abnormal cycle statistics sub-module counts all the marked abnormal records in the jump frequency abnormal flag sequence, extracts the corresponding cycle number information, removes duplicates and summarizes the abnormal cycle numbers, and generates a list of abnormal cycle numbers;

[0043] The drift feature comparison sub-module calls the list of abnormal cycle numbers, matches the drift direction label data in the concentrated change performance of the brushing behavior time, identifies whether there are marked types of both extended jump intervals and extended drift regions within the cycle, screens and summarizes the cycles with dual markings, and obtains a classifiable health abnormal cycle sequence;

[0044] The healthy cycle determination sub-module extracts the start and end time fields of the cycle according to the classifiable health abnormal cycle sequence, performs annotation integration in chronological order after unifying the format, establishes a classification information record based on the cycle behavior characteristics, and obtains the analysis result of the electric toothbrush user behavior.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0046] In the present invention, the behavior change period is screened through the time difference between the start and end times of vibration, improving the recognition accuracy of abnormal cycles. The behavior segments are uniformly adjusted to the standard length and normalized to the proportional position, enhancing the alignment consistency of cross-day behavior data. Time offset and high-frequency time point clustering are used to construct a time concentration distribution model to clarify the rhythm offset trend. The vibration alternating interval sequence combines with the stability threshold to identify abnormal jump frequencies, and the linkage between jumps and drifts is used to judge the health risk cycle, realizing the deep integration of multi-dimensional behavior characteristics and health trends. Description of the Drawings

[0047] Figure 1 is the system flow chart of the present invention;

[0048] Figure 2 is the flow chart of the cleaning behavior change extraction module of the present invention;

[0049] Figure 3 is the flow chart of the behavior continuous segment alignment module of the present invention;

[0050] Figure 4 is the flow chart of the rhythm trajectory offset detection module of the present invention;

[0051] Figure 5 Flow chart of the behavior rhythm linkage recognition module of the present invention;

[0052] Figure 6 Flow chart of the abnormal trend health prompt module of the present invention. Detailed implementation manners

[0053] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0054] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more unless otherwise specifically defined.

[0055] Please refer to Figure 1 , the electric toothbrush user behavior analysis system based on big data includes:

[0056] The cleaning behavior change extraction module obtains the daily brushing duration record, the record of the number of covered areas and the record of the vibration intensity change of the user, screens the cycles in which the three types of continuous records all show a decreasing change, calls the starting and ending vibration fluctuation time values during brushing in the cycle, and judges whether the difference exceeds the time change limit range. If both conditions are met, it is marked as a cleaning habit change cycle and a change time period list is generated;

[0057] The behavior continuous segment alignment module obtains the starting and ending time points of the vibration of the brushing behavior within the cycle for the marked cycles in the change time period list, selects the daily behavior data, compares the head and tail time distances of the behavior segments and judges the difference behavior segments, adjusts the time span to a fixed standard length, updates the corresponding proportional interval positions of the behavior segments and then arranges them uniformly to generate a time sequence segment list;

[0058] The rhythm trajectory deviation detection module calls the first brushing record of each day in the time sequence segment list, records the time distribution and judges whether it deviates in the same direction for three consecutive days, then obtains the first three time points of the daily high-frequency operations and assigns them to a fixed time period group, judges whether the time points cross the time period range during different days, and makes a comparison in combination with the repeated attribution situation of the brushing time to generate the concentrated change performance of the brushing behavior time;

[0059] The behavior rhythm linkage recognition module extracts the cycle records with a coincidence degree of time and cleaning habit changes exceeding two days according to the marked offset cycles in the concentrated change performance of brushing behavior time, obtains the sequence of alternating values of the brushing vibration strength and weakness under the cycle, and judges whether the adjacent vibration interval distance exceeds the stable vibration interval change limit value. If the offset amplitude requirement is met, the abnormal segment is recorded, and a jump frequency abnormal flag sequence is generated;

[0060] The abnormal trend health prompt module counts the occurrence cycles corresponding to the abnormal records in the jump frequency abnormal flag sequence, calls the drift direction content in the concentrated change performance of brushing behavior time, and judges whether there are two types of marks, namely, the extension of the jump interval and the expansion of the drift area, at the same time. If the conditions are met, they are integrated into the health abnormal cycle segment, and the analysis result of the electric toothbrush user behavior is generated.

[0061] The cleaning habit change time period list includes the decreasing duration cycle, the decreasing coverage area cycle, the weakening vibration intensity cycle, and the abnormal mark of the vibration fluctuation difference value. The time sequence segment list includes the standardized brushing behavior segment, the proportional interval position parameter, and the unified arrangement order index. The concentrated change performance of brushing behavior time includes the first brushing time offset sequence, the high-frequency operation time point group, and the time period attribution distribution characteristics. The jump frequency abnormal flag sequence includes the strong and weak vibration alternating abnormal segment, the vibration interval exceeding limit record, and the offset amplitude reaching the standard section. The analysis result of the electric toothbrush user behavior includes the jump interval extension flag, the drift area expansion flag, and the health abnormal cycle segment.

[0062] Please refer to Figure 2 , and the cleaning behavior change extraction module includes:

[0063] The data trend extraction sub-module obtains the user's daily brushing duration record, the coverage area quantity record, and the vibration intensity change record. Based on the continuous trend changes of each type of data in the three record sequences, by judging whether the data of consecutive multiple time nodes show a monotonically decreasing state, the cycle segment in which the three records decrease continuously at the same time is screened, and the combined decreasing cycle interval value is generated;

[0064] Obtain the user's daily brushing duration record, the record of the number of covered areas, and the record of vibration intensity changes. Specifically, during implementation, the system first retrieves the original sensor data within a specified time range from the stored user database. For example, retrieve the daily records of user A from March 1, 2025, to March 10, 2025, and identify the daily brushing duration data, such as [125 seconds, 122 seconds, 118 seconds, 115 seconds, 110 seconds, 105 seconds, 100 seconds, 98 seconds, 95 seconds, 92 seconds]. At the same time, extract the corresponding daily record of the number of brushing-covered areas, such as [12, 12, 11, 10, 10, 9, 8, 8, 7, 7], and the record of the number of times of vibration intensity mode switching or manual adjustment during daily brushing, such as [3 times, 3 times, 2 times, 2 times, 1 time, 1 time, 1 time, 0 times, 0 times, 0 times]. Then, the system compares the time series of these three types of data day by day to check whether there is a monotonically decreasing trend for at least three consecutive days (preset). Taking the above data as an example, from the 2nd day to the 7th day (March 2nd to March 7th), the duration continuously decreases (from 122 seconds to 100 seconds), the covered area decreases (from 12 to 8), and the number of changes decreases or remains the same (from 3 times to 1 time). The system records this time range (March 2, 2025, to March 7, 2025), continues to process the subsequent data until all the data within the specified time range is traversed, and pairs up the start and end dates of all the identified cycles that meet the condition of three consecutive indicators decreasing simultaneously for at least 3 days to form multiple cycle intervals. For example, in addition to March 2nd to March 7th, if similar trends appear in the subsequent data from March 20th to March 25th, a new interval will be added. Finally, output a set containing all the identified combined decreasing cycle intervals, such as [2025-03-02, 2025-03-07], [2025-03-20, 2025-03-25], as the combined decreasing cycle interval values.

[0065] The time difference judgment sub-module calls the vibration fluctuation time values at the start and end of brushing within each cycle in the combined decreasing cycle interval value, calculates the difference between the start and end times of the vibration intensity, and compares the difference with the time change limit range to identify the cycles that simultaneously meet the conditions of a decreasing trend and an excessive vibration time difference, obtaining a list of cycles with compliant changes;

[0066] Call the vibration fluctuation time value of the start and end of brushing in each cycle in the combined decreasing cycle interval value. Specifically, for the first interval [2025-03-02, 2025-03-07] in the combined decreasing cycle interval value [2025-03-02, 2025-03-07], [2025-03-20, 2025-03-25] output in the previous step, the system further queries the detailed vibration sensor log corresponding to the daily brushing events in the cycle, and extracts the record at the beginning of each brushing. For example, for March 2, 2025, the start vibration time is 07:05:02 and the end vibration time is 07:07:07. The time difference between the two is 125 seconds. This operation is performed every day in the interval (March 2 to March 7) to obtain the actual vibration duration sequence of each day. Then, a time variation limit range is set. The setting of this range is based on the statistical analysis of a large number of users' normal brushing behavior. It is generally believed that a single brushing time of less than 60 seconds or more than 180 seconds indicates abnormal behavior or recording interruption. Therefore, the limit range is set to [60 seconds, 180 seconds]. The calculated daily actual vibration duration difference is compared with this range. For example, if the calculated vibration duration on March 5 is 55 seconds, which is lower than the lower limit of 60 seconds, then the day is marked as not meeting the condition. The system checks whether the vibration duration difference of all days in the entire period [2025-03-02, 2025-03-07] is within the range. If the vibration duration of each day in the period [2025-03-02, 2025-03-07] is within the limit, then the period is retained. If the vibration duration of any day in the period [2025-03-20, 2025-03-25] is 190 seconds, then the period is eliminated. Finally, a list of all periods that pass the screening is output to obtain a list of periods that meet the change requirements.

[0067] The change period generation submodule extracts the start and end time nodes in the period as boundaries based on the change conformity period list, constructs time period information indicating significant changes in the user's brushing behavior, and obtains a change period list;

[0068] Based on the variable compliance cycle list. For example, assume that the list obtained after the previous screening is [2025-03-02, 2025-03-07]. The system extracts the start time node (March 2, 2025) and the end time node (March 7, 2025) of each cycle in this cycle list as the boundaries of the time period, and directly constructs these start and end dates into a clear time period information record. This record indicates that the user's brushing behavior pattern (according to the previous screening conditions, that is, the duration, coverage, and intensity change jointly decrease and the single duration is within a reasonable range) has changed continuously during this period. All cycles that pass the screening are processed in this way to construct a list, and each element in the list is a time period representing a significant change in behavior. For example, [start date: 2025-03-02, end date: 2025-03-07], to obtain the list of cleaning habit change time periods.

[0069] Please refer to Figure 3 , the behavior continuous segment alignment module includes:

[0070] The start and end data extraction sub-module obtains the brushing records corresponding to the cycles in the list of cleaning habit change time periods, extracts the start time and end time of the vibration of the daily brushing behavior within the cycle, calculates the daily behavior start and end interval value and binds it to the daily data label, and generates a daily behavior interval data group;

[0071] Obtain a list of cleaning habit change time periods. For example, the input is [start date: 2025-03-02, end date: 2025-03-07]. For each time period in the list, such as [2025-03-02, 2025-03-07], the system accesses the corresponding original brushing record database, extracts the vibration start timestamps and end timestamps of all brushing behaviors for each day within this period (from March 2nd to March 7th). For example, on March 2nd, there are multiple records, such as id: rec1, start: 07:05:02, end: 07:07:07, id: rec2, start: 21:30:10, end: 21:32:05. Calculate the start and end time interval values for each brushing behavior on each day, that is, subtract the start timestamp from the end timestamp to get the duration (in seconds). For rec1, the calculated value is 125 seconds, and for rec2, it is 115 seconds. Bind each calculated interval value to its corresponding data record label (such as the date 2025-03-02 and record ID rec1) to form structured data containing the date, record identifier, and duration. For example, [date: 2025-03-02, record ID: rec1, interval value: 125 seconds, date: 2025-03-02, record ID: rec2, interval value: 115 seconds, date: 2025-03-03,..., date: 2025-03-07,...]. Repeat this process for all days and all records within the period to generate a daily behavior interval data group.

[0072] Based on the daily behavior interval data group, the time period judgment sub-module identifies whether the absolute difference in the start and end distances of the behavior is greater than the time threshold benchmark, determines the brushing behavior time period, and extracts the corresponding record data. The formula used is:

[0073]

[0074] Calculate to obtain the behavior segment adjustment coefficient value. Based on this coefficient, perform a rationality screening on the behavior segment time interval to obtain an adjusted effective behavior interval group;

[0075] Among them, Q i is the behavior segment adjustment coefficient value for the i-th day, Z i is the number of daily brushing vibration wave bands, M i is the average daily vibration intensity, A i is the number of brushing coverage areas, D i is the total duration of the daily behavior segment, R is the fixed benchmark value of the behavior duration, F i,k is the time fluctuation value of the k-th segment on the i-th day, C k is the coverage time constant corresponding to the k-th segment;

[0076] Based on the daily behavior interval data set, for example, it contains data items such as record rec1 with an interval value of 125 seconds recorded from user A on March 2, 2025, record rec2 with an interval value of 115 seconds, and record rec3 with an interval value of 55 seconds on March 3, 2025, etc. First, the system sets a time threshold benchmark for an effective brushing behavior. The setting of this benchmark value refers to the recommended single brushing duration in generally accepted oral health guidelines (usually recommended 2 minutes, i.e., 120 seconds) and the statistical analysis of the actual brushing duration data of a large number of users. Considering that too short records are not complete brushing processes but device misoperations or short-term uses, a lower limit is selected as the basis for validity judgment. For example, the time threshold benchmark is set to 60 seconds. Then, the system processes each record in the daily behavior interval data set one by one, comparing its "interval value" with this 60-second benchmark, and specifically performing a judgment action: If the interval value of the record is strictly greater than 60 seconds, it is determined that the time period corresponding to this record is a period of effective brushing behavior; conversely, if it is less than or equal to 60 seconds, it is determined to be invalid;

[0077] For example, records rec1 (125 seconds > 60 seconds) and rec2 (115 seconds > 60 seconds) are both determined to be effective, and record rec3 (55 seconds ≤ 60 seconds) is determined to be invalid. Subsequently, the system extracts all relevant record data corresponding to the determined effective brushing behavior time periods, which includes but is not limited to duration, detailed vibration intensity sequence, coverage area information, etc., to prepare data input for subsequent rationality screening. Then, the system calls the formula

[0078]

[0079] to calculate the behavior segment adjustment coefficient value Q for each effective behavior record i ;

[0080] This formula aims to comprehensively evaluate the quality and standardization of a single brushing behavior. Among them, i is used as a subscript, indicating that this is the calculation for the i-th effective brushing behavior record, and Q i is the calculated adjustment coefficient value for this behavior, and the magnitude of its value reflects the deviation degree of this brushing behavior from the preset ideal model. The absolute value symbol |…| ensures that the final result is non-negative. Z i represents the number of vibration wave bands detected by analyzing the vibration sensor data during this effective brushing behavior, that is, the number of times the intensity mode changes or adjusts significantly during the brushing process. For example, switching from the "cleaning" mode to the "massage" mode is counted as 1 time. This data is obtained by processing the original sensor log. Assuming that for record rec1, Z i = 2 times, M iIt represents the average vibration intensity value recorded by the medical-grade piezoelectric ceramic sensor during this brushing, which can be a quantified gear value (such as from 1 to 5 gears) or a specific average frequency (unit: Hertz, Hz). For example, by calculating the average of all intensity readings during the record rec1, M is obtained. i = 4.0 (assuming it is the average value of the gear, and its range is usually between 2.0 and 5.0), A i is the number of oral regions covered during this brushing, which is judged by the position sensor built into the toothbrush combined with an algorithm. For example, the oral cavity is pre-divided into 12 standard regions, and by tracking the position of the toothbrush, it is estimated how many of them are covered. Assuming that the record rec1 covers A i = 11 regions, D i is the total duration of this effective behavior segment, that is, the aforementioned "interval value" determined to be effective, with the unit of seconds. For the record rec1, D i = 125 seconds. R is a preset fixed reference value for the behavior duration, and its setting is usually based on the standard recommended brushing duration. For example, let R = 120 seconds, which is used to measure the deviation between the actual duration and the recommended duration. m represents the total number of regions into which the oral cavity is divided, which is a fixed value, such as m = 12. The summation symbol indicates that it is necessary to perform a traversal calculation for all m oral regions. The subscript k represents the specific region number, ranging from 1 to m, F i,k represents the vibration time fluctuation value recorded when the toothbrush is located in the k-th oral region during the i-th effective brushing behavior. This value can be quantified by calculating the standard deviation of the amplitude of continuous vibration signals within this region or the amplitude of the change in vibration frequency per unit time, reflecting the stability during brushing in this region. The unit can be seconds or the standard deviation of the intensity. The smaller the value, the more stable it is. For example, assuming for the record rec1, the F i,k values (unit: seconds) for each covered region are obtained as follows:

[0081] F i,1 = 0.5, F i,2 = 0.6, F i,3 = 0.8, F i,4 = 0.9, F i,5 = 1.2, F i,6 = 1.1, F i,7 = 0.4, F i,8 = 0.5, F i,9 = 0.7, F i,10 = 0.8, F i,11 = 1.0;

[0082] These values are usually between 0.1 and 2.0 seconds, C kis the recommended coverage time constant corresponding to the k-th area, with the unit of seconds. This constant is preset according to the cleaning difficulty and importance of different tooth areas. For example, the structure of the posterior molar area is complex and requires a longer time, while the incisors are relatively simple and the time is shorter. Set the C values for all 12 areas k as follows: C1 = 10, C2 = 10, C3 = 12, C4 = 12, C5 = 15, C6 = 15, C7 = 10, C8 = 10, C9 = 12, C 10 = 12, C 11 = 15, C 12 = 15 seconds. These constant values are optimized based on dental expertise and a large amount of user data. The first part of the formula combines the vibration complexity (Z i ), intensity, and coverage breadth ( square root is used to smooth the dimension and numerical range), and the duration deviation (D -R) to evaluate the degree of investment and normativity of the behavior itself; i The second part of the formula

[0083] quantifies the degree of satisfaction of the stability and uniformity of brushing in each area relative to the recommended duration. F / C i,k can be understood as the instability or time allocation deviation within the unit recommended time. After summation, the overall regional brushing quality index is obtained. Subtracting the two parts and taking the absolute value, the obtained Q k comprehensively reflects the deviation degree of this brushing behavior from an ideal model including multiple dimensions such as complexity, intensity, coverage, duration, and regional uniformity. It does not simply rely on the total duration to evaluate the brushing behavior, but by calculating and comparing the internal process indicators (vibration change, intensity, coverage, regional stability) with the external benchmarks (recommended duration, regional recommended duration), it can more meticulously and comprehensively evaluate the internal quality and rationality of a single brushing behavior. Next, calculate the Q i of the record rec1; i Calculation example;

[0084] First, calculate the summation term Since area 12 is not covered (A i = 11), assume its F i,12 = 0 or does not participate in the summation (here, it is processed as only summing the covered areas);

[0085]

[0086] Then, substitute all parameter values for

[0087] Q i calculation:

[0088]

[0089] The calculation process is as follows:

[0090] Q i = |1.7266 - 0.687|,

[0091] Q i = |1.0396| = 1.0396;

[0092] The obtained behavior segment adjustment coefficient value of record rec1 is 1.0396. Subsequently, the system uses a preset rationality screening threshold to evaluate the calculated Q i value. The setting of this threshold is also based on the statistical analysis of the Q i value distribution calculated by a large number of users, aiming to distinguish abnormal behaviors with excessive deviation. For example, this threshold is set to 1.5, which is an empirical value indicating that a certain degree of deviation of the behavior from the ideal model is allowed, but if it exceeds this limit, its rationality is questionable. The system performs a comparison operation: comparing the calculated Q i = 1.0396 with the threshold 1.5. Since 1.0396 < 1.5, it is determined that the behavior segment of record rec1 is within the reasonable range under the multi-dimensional evaluation, and the screening operation is performed: retaining this record. If the Q j value calculated for another valid record is 1.8, then because 1.8 > 1.5, this record will be excluded. The system repeats the Q i calculation and the threshold-based screening and judgment process for all initially determined valid brushing behavior records, and finally pools all the valid behavior records that pass the Q i value screening (i.e., Q i is less than or equal to the threshold) to obtain an adjusted group of valid behavior intervals.

[0093] The unified time series generation sub-module relocates the daily behavior segment lengths at equal intervals according to the adjusted group of valid behavior intervals, converts them into proportional positions under the standard time span, arranges the behavior segments in chronological order, establishes a time period sequence in a unified format, and obtains a behavior time series column;

[0094] According to the adjusted valid behavior interval group, for example, after screening, records rec1 (March 2, duration 125 seconds) and rec2 (March 2, duration 115 seconds) and other valid records on subsequent dates are retained. The system sums up the total lengths of all valid behavior segments retained daily to obtain the total daily valid brushing duration. For example, the total valid duration on March 2 is 125 + 115 = 240 seconds. Then, the daily total valid duration is subjected to equidistant repositioning, that is, it is converted into a proportional position relative to the standard time span (such as 24 hours a day, i.e., 86400 seconds). Or, more commonly, the occurrence time (such as the start time) of each valid brushing behavior is recorded, and these valid behavior segments are arranged in chronological order. For example, if record rec1 on March 2 starts at 07:05:02 and rec2 starts at 21:30:10, then these events are arranged in chronological order while retaining information such as their valid durations. This operation is performed on all valid records within the entire change cycle (such as from March 2 to March 7) to establish a time period sequence in a unified format. This sequence contains the screened valid brushing events sorted by time and their related attributes (such as start time, valid duration), and the behavioral time series segment column is obtained.

[0095] Please refer to Figure 4 , the rhythm trajectory offset detection module includes:

[0096] The first brushing time offset sub-module calls the first brushing record time of each day in the behavioral time series segment column, collects time distribution data, determines whether there is a one-way offset of continuous time values in the time series, makes record marks according to the consistency of the offset direction, and generates a continuous offset time group;

[0097] Call the behavioral time series segment column, which contains the daily valid brushing records arranged in chronological order, such as [Date: 2025-03-02, Start time: 07:05:02, Duration: 125, Date: 2025-03-02, Start time: 21:30:10, Duration: 115, Date: 2025-03-03, Start time: 07:15:40, Duration: 110,...]. The system extracts the first brushing record time of each day, that is, the start time of the record with the earliest timestamp every day. For example, the first brushing time on March 2 is 07:05:02, on March 3 is 07:15:40, on March 4 is 07:25:10, and on March 5 is 07:35:00, and collects these first brushing time points;

[0098] A time series is formed: [07:05:02, 07:15:40, 07:25:10, 07:35:00, ...], and then it is determined in chronological order whether there is a unidirectional shift in the time values for three or more consecutive days. A unidirectional shift is defined as three or more consecutive days in which the first tooth brushing time of each day is delayed or advanced relative to the previous day, and the shift exceeds a set minimum shift threshold. The threshold is set based on the observation of the stability of user habits, for example, it is set to 10 minutes. Checking the above sequence, 07:15:40 is 10 minutes and 38 seconds later than 07:05:02 (>10 minutes), 07:25:10 is 10 minutes and 38 seconds later than 07:15:40, and 07:25:10 is 10 minutes and 38 seconds later than 07:15:40. 5:40 is 9 minutes and 30 seconds late (<10 minutes), and 07:35:00 is 9 minutes and 50 seconds later than 07:25:10 (<10 minutes). Assuming that subsequent dates continue to be postponed and the magnitude meets the requirements, for example, March 4th is changed to 07:30:00, and March 5th is changed to 07:45:00, then March 2nd to March 5th constitute three consecutive days of postponement (all greater than 10 minutes). The system marks the records of this period (March 2nd to 5th) according to the identified shift direction (delay or advance), for example, marking it as "delayed shift". All identified consecutive one-way shift time periods and their marks are aggregated to generate a continuous shift time group.

[0099] The high-frequency point grouping submodule extracts the top three time points with the highest daily vibration frequency based on the daily toothbrushing records in the continuous offset time group. The time points are assigned to fixed time period labels according to the hourly period they fall into, using the formula:

[0100]

[0101] The time attribution difference value is obtained by operation, which is used to identify the time span difference between the differentiated days, determine whether the change characteristics of the time period classification are formed, and obtain the group difference change value;

[0102] Among them, S t is the attributed difference value, N f is the number of high-frequency time points extracted daily, H u is the hour segment number of the u-th high-frequency time point, D u is the brushing time corresponding to the time point, V u is the vibration intensity value corresponding to the time point, G v is the total count of the time period to which the vth day belongs, N d is the number of days of participation in the cycle, U c is the number of time period types, F v is the total number of tooth brushing behaviors on day v;

[0103] Based on a continuous offset time group, for example, it is identified that the period from March 2nd to March 5th is a "postponed offset" cycle. The system further processes all valid brushing records for each day within this cycle (not just the first brushing), extracts the brushing record segments corresponding to the top three time points with the highest internal vibration frequency (or equivalent vibration intensity level value) each day. For example, for March 2nd, assuming the vibration intensity peaks appear at 07:06:00 (intensity 5), 21:31:00 (intensity 5), and 07:06:30 (intensity 4.8), then these three time points are extracted. Then, according to the hour segments in which these time points are located, they are assigned to predefined fixed time period labels. For example, the 24 hours of a day are divided into 4 time periods: early morning (0 - 6 o'clock), morning (6 - 12 o'clock), afternoon (12 - 18 o'clock), evening (18 - 24 o'clock). Then 07:06:00 and 07:06:30 belong to the "morning" label, and 21:31:00 belongs to the "evening" label. This high-frequency point extraction and time period assignment operation are performed for each day within this cycle. Next, calculate the time attribution difference value S t ;

[0104]

[0105] Explanation of the parameters of the formula: S t is the value for evaluating the degree of difference in the distribution of high-frequency brushing time points over dates within a certain continuous offset time group; N f is the number of high-frequency time points extracted daily, set to 3 as required; represents the traversal and summation of the N f high-frequency points extracted daily; H u is the hour segment number of the u-th high-frequency time point. The time period label (such as "morning") needs to be quantified into a numerical number. For example, early morning = 1, morning = 2, afternoon = 3, evening = 4; D u is the total duration (in seconds) of the complete brushing behavior corresponding to this high-frequency time point, obtained from the behavior time sequence column; V u is the instantaneous vibration intensity value corresponding to this high-frequency time point or the average intensity value of its corresponding time period, which needs to be quantified, for example, intensity levels 1 - 5; Combines the duration and intensity information, and taking the square root is used to adjust its magnitude; Assigns a comprehensive weight to each high-frequency point, reflecting its occurrence time period, duration, and intensity; Calculates the average weighted time period index of the high-frequency points daily; The second part is used to calculate the average time period distribution within the entire cycle; N d is the number of participating calculation days included in this continuous offset time group. For example, from March 2nd to 5th, N d = 4; represents the traversal of each day within the cycle; Gv is the total value of the time period tags to which all high-frequency points belong on the v-th day (which can be understood as the sum of the time period numbers to which the high-frequency points belong on that day, or a more complex distribution measure); U c is the total number of predefined time period categories, which are divided into 4 here; max(F v ) is the maximum value among the total number of all brushing behaviors that occurred on the v-th day (if only the maximum value within the period is of concern, then for or here F v represents the total number of brushings on the v-th day, then max(F v ) is the maximum number of daily brushings within the period, which is used as the normalization denominator; Calculates an average measure of the time period distribution within the period; the entire formula quantifies the deviation degree of the daily high-frequency point time distribution relative to the overall average distribution of the period or the distribution difference between days by calculating the absolute value of the difference between the daily weighted average time period index and the average time period measure of the period; by combining the time period, duration, and intensity information of high-frequency brushing behaviors and comparing them with the average distribution pattern of the entire period, the stability or variability characteristics of the user's brushing time rhythm within a specific period can be quantified; now, for the above period (from March 2nd to March 5th, N d = 4), a simplified example of the S t calculation is given (assuming to calculate the difference of March 3rd relative to the overall average of the period): Obtain the parameter values (calculate the first part with March 3rd as an example, assuming the overall period data is used to calculate the second part): N f = 3;

[0106] Assume the high-frequency points and information on March 3rd are: Time point 1: 07:16:00, H1 = 2 (morning), D1 = 110 seconds, V1 = 4.5 gears, Time point 2: 21:40:00, H2 = 4 (evening), D2 = 105 seconds, V2 = 4.8 gears, Time point 3: 07:17:30, H3 = 2 (morning), D3 = 110 seconds, V3 = 4.2 gears;

[0107] Calculate the weighted average time period index for March 3rd:

[0108]

[0109]

[0110] Obtain the overall period parameter values: Assume obtained by calculating the data of all days within the period (G v is calculated in a certain way, for example, G v = ∑H u ), the number of time period categories U c = 4, and the maximum number of daily brushings within the period is assumed to be max(Fv ) = 3 times;

[0111] Calculate the average period metric:

[0112]

[0113] Calculate S t (The concept of the difference between a single day and the average of the period is demonstrated here. The original formula is more like calculating a certain overall difference index for the entire period, and its exact meaning depends on the v specific definition of G. If it is understood as the difference between days, then pairwise calculation or the standard deviation of the relative mean needs to be calculated. Here, an example value is calculated according to the formula structure, assuming it represents a certain average daily deviation for the entire period): S t = |28.23 - 50| = |-21.77| = 21.77 (Note: In actual applications, the calculation of S t involves more complex day-to-day comparisons or statistics) Set a threshold for the change in the group difference. For example, based on the analysis of a large amount of user data, the threshold is set to 15. If the calculated value of S t (or the fluctuation index within the period calculated based on S t ) is greater than 15, it is determined that there is a significant day-to-day variation or deviation from the average pattern in the distribution of high-frequency brushing time points within this period. Record this calculated value of the change in the group difference (such as 21.77) to obtain the value of the change in the group difference. This result indicates that the distribution characteristics of high-frequency brushing time within this period (considering time period, duration, intensity) are significantly different (21.77 > 15) from the overall average of the period or compared with other days, showing instability.

[0114] The time concentration comparison sub-module combines the value of the change in the group difference to perform a normalized comparison of the repetition of the frequency of time period attribution within consecutive periods, determines whether the time distribution is concentrated in a fixed segment, obtains the index of the concentration degree of the brushing time classification, and establishes the performance of the change in the time concentration of the brushing behavior;

[0115] Combining the value of the change in the group difference, such as the value 21.77 obtained in the previous step, and the daily time period attribution information (such as March 2: morning x2, evening x1; March 3: morning x2, evening x1; March 4: morning x1, afternoon x1, evening x1; March 5: morning x1, afternoon x2) obtained when grouping the high-frequency points in the previous step, the system performs a normalized comparative analysis of the repetition of these time period attribution frequencies within consecutive periods (from March 2 to March 5). Specifically, calculate the total number of times each time period label (early morning, morning, afternoon, evening) appears within the entire period. For example, the morning appears 6 times, the afternoon appears 3 times, the evening appears 3 times, and the early morning appears 0 times. Then calculate the frequency or proportion of each time period.

[0116] Please refer toFigure 5 , the behavior rhythm linkage recognition module includes:

[0117] The offset period screening sub-module extracts the period data with a coincidence degree exceeding two days with the cleaning habit change section according to the marked period time range in the concentrated change performance of the brushing behavior time, cross-compares the period boundaries, screens out the eligible time periods, and generates a combined offset period set;

[0118] According to the marked period time range and its concentration judgment result in the concentrated change performance of the brushing behavior time (for example, the period [2025-03-02, 2025-03-05], marked as "postponed offset", and the concentration judgment is "not concentrated"), and the list of cleaning habit change sections obtained in the first module "Cleaning habit change extraction" (for example, [start date: 2025-03-02, end date: 2025-0, 3-07]), the system extracts the period data with overlapping time ranges in these two types of periods, and the coincidence degree requirement exceeds the set number of days threshold, for example, set to at least coincide for 2 days. Compare [2025-03-02, 2025-03-05] with [2025-03-02, 2025-03-07], their overlapping interval is [2025-03-02, 2025-03-05], a total of 4 days, which is greater than 2 days and meets the conditions. The system cross-compares and confirms the boundaries of all period pairs that meet the coincidence conditions, and retains these period time periods that simultaneously have a "rhythm offset feature" (from module three) and a "cleaning habit change feature" (from module one). For example, retain [2025-03-02, 2025-03-05] (or take the overlapping part of the two as the new combined period), and summarize all the screened time periods to generate a combined offset period set.

[0119] The vibration sequence extraction sub-module calls the period data in the combined offset period set, extracts the vibration strength alternating numerical sequence of the daily brushing behavior, constructs a vibration change array in sequence and marks the corresponding time interval position to obtain a continuous vibration interval sequence;

[0120] Call the combined offset period set, such as [2025-03-02, 2025-03-05]. For each period in the set, the system extracts the detailed vibration intensity sensor reading sequences for each valid brushing behavior on each day within that period. These sequences are intensity values recorded in chronological order. For example, for the record rec1 (duration 125 seconds) on March 2nd, an array containing hundreds of intensity readings is obtained [3.0, 3.1,..., 4.5, 4.6,..., 3.5,..., 2.0]. These numerical sequences are constructed into a vibration change array or list according to their occurrence order during the brushing process, and at the same time, the time interval position information corresponding to these vibration values needs to be marked. For example, record the duration of each stable vibration (intensity change less than the threshold), or record the time points where significant intensity jumps occur, forming a sequence that can reflect the alternating pattern of vibration strength and the duration of each pattern. For example, [(intensity: 3.0, duration: 5s), (intensity: 4.5, duration: 30s), (intensity: 3.5, duration: 60s), (intensity: 2.0, duration: 30s)]. Such sequences are generated for all valid brushing records within the period to obtain a continuous vibration interval sequence.

[0121] Based on the continuous vibration interval sequence, the jump frequency judgment sub-module judges the difference between adjacent vibration interval values, compares them with the stable vibration interval boundary values, and identifies the vibration sections with excessive fluctuation amplitude. The formula is used:

[0122]

[0123] Where:

[0124] Vector

[0125] Vector

[0126] ||·||1 represents the 1-norm, that is, the sum of the absolute values of the elements of the vector difference;

[0127] Calculate to obtain the vibration interval offset value, judge whether it constitutes an abnormal jump behavior according to the comparison result of the vibration interval offset value and the offset amplitude threshold, and obtain the jump frequency anomaly flag sequence;

[0128] Among them, F dev is the vibration interval offset value, N seg is the total number of jump sections, P r is the average vibration intensity of the r-th section, T r is the time difference between the start and end of this section, C r is the number of cleaning areas covered in this section, Θ is the upper threshold of the vibration stable time interval, D r is the duration of the r-th behavior;

[0129] Based on a continuous vibration interval sequence, for example, containing multiple records in the form of [(intensity: P1, duration: T1), (intensity: P2, duration: T2),...], the system determines the difference between adjacent vibration interval values (here referring to the vibration intensity value P), for example, |P2 - P1|, |P3 - P2|, etc., and compares it with a preset stable vibration limit value, which is set based on the intensity change range of normal mode switching or fine-tuning. For example, it is set to an intensity gear difference of 1.0. If the absolute value of the adjacent intensity difference, for example, |4.5 - 3.0| = 1.5, is greater than the limit value of 1.0, it is considered that a large-amplitude vibration jump has occurred, and the system identifies these vibration sections with excessive fluctuation amplitudes (i.e., the two interval sections before and after the jump occurs);

[0130] Calculate to obtain the vibration interval offset value F dev ;

[0131]

[0132] Explanation of the parameters of the formula: F dev Is a comprehensive measure of the deviation of the vibration behavior from the stable state within the identified potential abnormal vibration jump section; N seg Is the total number of sections with large jumps identified in a brushing record (for example, if there are two large jumps, involving a total of 4 stable sections before and after, then N seg = 4 or select specific sections according to the definition); P r Is the average vibration intensity value (Power) within the r-th section, such as the intensity gear value or frequency value; T r Is the duration difference (Timedelta) of the r-th section, in seconds; Combines the intensity and duration of the section to form a metric value. The longer the intensity and time, the larger this value; C r Is the number of oral regions covered during brushing within the duration of the r-th section (Coverage), reflecting the spatial range of the behavior in this section; Θ is the upper threshold of the set vibration stable time interval (Threshold), for example, based on user data statistics, the upper limit of the normal stable mode duration is set, such as 45 seconds. Exceeding this value indicates that the user has stayed in a certain mode for too long; D r Is the behavior duration (Duration) of the r-th section, the same as T r Same, indicating the section duration; This term combines the number of covered regions and time factors. Adding Θ to the denominator is to avoid the value being too large when D r Is too small and introduces the consideration of the stable duration. This term can be understood as the coverage efficiency within the unit adjustment time; the subtraction within the parentheses Indicates the comparison between the intensity-time metric and the coverage-time efficiency metric, and the difference reflects a certain "cost performance" or "normative" deviation of this section of behavior;

[0133] The final absolute value ensures that the result is non-negative;

[0134] By focusing on the sections where significant changes in vibration intensity occur and comprehensively considering the intensity (P r ), duration (T r ), coverage (C r ) of these sections and their relationship with the stable duration threshold (Θ), an index is provided to quantitatively evaluate whether there are abnormal jumps in the vibration rhythm during brushing, rather than just looking at the intensity change itself;

[0135] Now, for two jump-related sections (assuming N seg = 2) identified in a certain brushing record, an F dev calculation example: Obtain parameter values: Assume section 1: P1 = 3.0 gears, T1 = 60 seconds, C1 = 4 regions, D1 = 60 seconds; section 2: P2 = 1.5 gears, T2 = 70 seconds, C2 = 3 regions, D2 = 70 seconds; set the threshold Θ = 45 seconds;

[0136] Calculate

[0137] Calculate the terms for section 1:

[0138] Calculate the terms for section 2:

[0139] Calculate F dev :

[0140] Based on the calculated vibration interval offset value F dev (such as 65.0135) and compare it with the preset offset amplitude threshold, which is set based on the F dev distribution calculated from a large amount of normal and abnormal brushing behavior data. For example, set the threshold to 50. If F dev > 50, then it is determined that this brushing behavior constitutes an abnormal jump behavior, and mark this brushing record as "abnormal". If F dev ≤ 50, then mark it as "normal". Repeat this calculation and judgment for all valid brushing records within the combined offset period to obtain a sequence marked with whether each record is an abnormal jump, that is, the jump frequency anomaly flag sequence. The result shows that the vibration interval offset value of this calculated brushing record is 65.0135, which is greater than the threshold of 50, so it is determined to have an abnormal jump behavior.

[0141] Please refer to Figure 6 , the abnormal trend health prompt module includes:

[0142] The abnormal cycle statistics sub-module counts all the marked abnormal records in the jump frequency abnormal flag sequence, extracts the corresponding cycle number information, removes duplicates and summarizes the abnormal cycle numbers to generate a list of abnormal cycle numbers;

[0143] Count all the records marked as "abnormal" in the jump frequency abnormal flag sequence. The form of this sequence is [record ID: rec1, abnormal flag: normal, record ID: rec_x, abnormal flag: abnormal, record ID: rec_y, abnormal flag: abnormal,...]. The system traverses this sequence and extracts the cycle number information corresponding to all the records with the abnormal flag "abnormal" (that is, in which combined offset cycle these records occur). For example, if records rec_x and rec_y both occur in the cycle [2025-03-02, 2025-03-05], then extract this cycle number. Remove duplicates from all the extracted cycle numbers because there are multiple abnormal records in one cycle, but only care about which cycles have abnormalities. Then summarize the deduplicated cycle numbers into a list. For example, if abnormal records are found in both the cycle [2025-03-02, 2025-03-05] and another cycle [2025-04-10, 2025-04-15], the list is [2025-03-02, 2025-03-05], [2025-04-10, 2025-04-15], generating a list of abnormal cycle numbers.

[0144] The drift feature comparison sub-module calls the list of abnormal cycle numbers, matches the drift direction label data in the concentrated change performance of the brushing behavior time, identifies whether there are both the marks of extended jump intervals and extended drift regions in the cycle, screens and summarizes the cycles with dual marks, and obtains a sequence of health abnormal cycles that can be classified;

[0145] Call the list of abnormal cycle numbers, such as [2025-03-02, 2025-03-05], [2025-04-10, 2025-04-15]. The system goes back to match the continuous offset time group generated in the "first brush time offset sub-module" (paragraph 7) of the "rhythm trajectory offset detection module", and the specific label data on the drift direction further analyzed therein (for example, the cycle [2025-03-02, 2025-03-05] is marked as "postponed offset"). At the same time, it is necessary to combine the further analysis of the abnormal jump behavior (paragraph 12) to judge whether the abnormal jump is mainly manifested as "extended vibration interval" (for example, by analyzing whether the T r value is generally too large or F devIs a high value related to a large T r ), and whether there is an "expansion of the drift area" (referring to the area A covered by brushing i or C r shows a trend of expansion or irregular change during the abnormal period, which requires additional analysis steps to obtain such markers). The system identifies the periods in each list of abnormal period numbers, checks whether they have both types of markers, namely "lengthening of the jump interval" and "expansion of the drift area" (or related "rhythm offset" markers, such as "postponed offset" can be regarded as a kind of time drift). For example, assume that the period [2025-03-02, 2025-03-05] is determined to have both the lengthening of the jump interval feature and the postponed offset feature, while the period [2025-04-10, 2025-04-15] only has the lengthening of the jump interval feature. Then only the former is selected. All the periods with such dual-marker features (indicating that the behavior change is abnormal not only in the intensity rhythm but also in the time regularity or spatial coverage with synchronous drift) are screened and summarized to obtain a sequence of healthy abnormal periods that can be classified.

[0146] The healthy period determination sub-module extracts the start and end time fields of the periods from the sequence of healthy abnormal periods that can be classified. After unifying the format, they are marked and integrated in chronological order to establish a classification information record based on the periodic behavior characteristics, and obtain the analysis result of the electric toothbrush user behavior;

[0147] According to the sequence of healthy abnormal periods that can be classified, for example, after screening, [2025-03-02, 2025-03-05] is obtained. The system extracts the start time field (2025-03-02) and the end time field (2025-03-05) of each period in this sequence, unifies the formatting of these date information, for example, unifies the format, and then sorts and marks and integrates them in the chronological order of occurrence. A classification information record based on the periodic behavior characteristics is established. This record clearly indicates that within which specific time period, the user's brushing behavior simultaneously exhibits multiple characteristics such as changes in cleaning habits, offsets in time rhythm, and abnormal jumps in vibration rhythm, and classifies this composite behavior pattern as a health risk signal or a period of drastic change in behavior pattern that needs attention. For example, the output result is: "During the period from March 2, 2025 to March 5, 2025, the user's brushing behavior showed significant changes, specifically manifested as a decrease in the cleaning efficiency index, a continuous postponement of the first brushing time, and frequent and abnormally extended intervals of vibration intensity jumps during the brushing process. It is recommended to pay attention to oral health status or adjust brushing habits", and finally obtain the analysis result of the electric toothbrush user behavior.

[0148] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art can use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A system for analyzing the behavior of electric toothbrush users based on big data, characterized in that, The system includes: The cleaning behavior change extraction module extracts the daily brushing duration, coverage area, and vibration intensity records, screens the decreasing periods of the three types of data, determines whether the difference between the start and end of the vibration exceeds the limit, marks it as the cleaning habit change period, and generates a list of change time periods; The behavior continuous segment alignment module reads the start and end points of the vibration in the list of change time periods, extracts the daily brushing data, determines the time difference between the start and end, and uniformly adjusts it to the standard length. After updating the ratio interval, it arranges and generates a time series segment list; The rhythm trajectory deviation detection module calls the first brushing record of each day in the time series segment list, extracts the daily high-frequency operations and groups them into fixed time period groups, analyzes whether the time points cross time periods during the day, and generates a concentrated time change performance; The behavior rhythm linkage recognition module, based on the concentrated time change performance, extracts the periods that coincide with the change time periods, analyzes the strong and weak vibration alternating sequence, determines whether it exceeds the stable limit, and generates a jump frequency anomaly flag sequence; The abnormal trend health reminder module, based on the jump frequency anomaly flag sequence, counts the abnormal flag periods, determines whether there are both extended intervals and regional drifts at the same time, classifies them as health abnormal periods, and outputs the analysis results of the electric toothbrush user behavior.

2. The system for analyzing the user behavior of an electric toothbrush based on big data according to claim 1, wherein The cleaning habit change time period list includes the decreasing duration period, the decreasing coverage area period, the decreasing vibration intensity period, and the abnormal mark of the vibration fluctuation difference. The time series segment list includes the standardized brushing behavior segment, the ratio interval position parameter, and the unified arrangement order index. The concentrated time change performance of the brushing behavior includes the first brushing time offset sequence, the high-frequency operation time point group, and the time period attribution distribution characteristics. The jump frequency anomaly flag sequence includes the strong and weak vibration alternating abnormal segment, the vibration interval overrun record, and the offset amplitude compliance section. The analysis results of the electric toothbrush user behavior include the jump interval extension flag, the drift area expansion flag, and the health abnormal period segment.

3. The system for analyzing the user behavior of an electric toothbrush based on big data according to claim 2, wherein, The cleaning behavior change extraction module includes: The data trend extraction sub-module obtains the user's daily brushing duration record, the coverage area quantity record, and the vibration intensity change record. Based on the continuous trend changes of each type of data in the three record sequences, by determining whether the data of consecutive multiple time nodes show a monotonically decreasing state, it screens the period segments in which the three records decrease continuously at the same time, and generates a combined decreasing period interval value; The time difference judgment sub-module calls the vibration fluctuation time values at the start and end of brushing in each period of the combined decreasing period interval value, calculates the difference between the start and end times of the vibration intensity, and compares the difference with the time change limit range to identify the periods that meet both the decreasing trend and the vibration time difference overrun conditions, and obtains a list of periods that meet the changes; The change time period generation sub-module, based on the list of periods that meet the changes, extracts the start and end time nodes in the period as boundaries, constructs the time period information indicating significant changes in the user's brushing behavior, and obtains the list of change time periods.

4. The system for analyzing the user behavior of an electric toothbrush based on big data according to claim 3, wherein The behavior continuous segment alignment module includes: The start and end data extraction sub-module obtains the brushing records corresponding to the cycles in the list of change periods, extracts the start time and end time of the brushing behavior every day during the cycle, calculates the start and end interval values of the daily behavior and binds them to the daily data tags, and generates a daily behavior interval data group; The time period judgment sub-module, according to the daily behavior interval data group, identifies whether the absolute difference between the start and end distances of the behavior is greater than the time threshold benchmark, determines the brushing behavior time period and extracts the corresponding record data, using the formula: Calculate to obtain the behavior segment adjustment coefficient value, and based on this coefficient, perform a reasonable screening of the behavior segment time interval to obtain an adjusted effective behavior interval group; Among them, Q i is the adjustment coefficient value of the i-th daily behavior segment, Z i is the number of daily brushing vibration wave bands, M i is the average daily vibration intensity, A i is the number of brushing coverage areas, D i is the total duration of the daily behavior segment, R is the fixed reference value of the behavior duration, F i,k is the time fluctuation value of the k-th section on the i-th day, C k is the coverage time constant corresponding to the k-th section; The unified time sequence generation sub-module, according to the adjusted effective behavior interval group, performs an equidistant repositioning of the daily behavior segment length, converts it into a proportional position under the standard time span, arranges the behavior segments in chronological order, establishes a time sequence in a unified format, and obtains a behavior time sequence column.

5. The system for analyzing the user behavior of an electric toothbrush based on big data according to claim 4, wherein The rhythm trajectory offset detection module includes: The first brushing time offset sub-module calls the daily first brushing record time in the behavior time sequence column, collects time distribution data, judges whether there is a one-way offset of continuous time values according to the time sequence, and makes record marks according to the consistency of the offset direction to generate a continuous offset time group; The high-frequency point grouping sub-module, according to the daily brushing records in the continuous offset time group, extracts the first three time points with the highest vibration frequency within each day, and assigns them to fixed time period tags according to the hour period where the time points are located, using the formula: Calculate to obtain the time attribution difference value, which is used to identify the time period crossing difference between different days, and judge whether a change feature of time period classification is formed to obtain a grouped difference change value; Among them, S t is the attribution difference value, N f is the number of high-frequency time points extracted daily, H u is the hour period number where the u-th high-frequency time point is located, D u is the corresponding brushing duration at this time point, V u is the corresponding vibration intensity value at this time point, G v is the total count of the time period to which the v-th day belongs, N d is the number of participating days within the cycle, U c is the number of time period types, F v is the total number of brushing behaviors on the v-th day; The time concentration comparison sub-module combines the grouped difference change value to perform a normalized comparison on the repetition of the time period attribution frequency within continuous cycles, judges whether the time distribution is concentrated in a fixed segment aggregation, obtains an index of the concentration degree of the brushing time classification, and establishes a manifestation of the concentrated change of the brushing behavior time.

6. The system for analyzing the user behavior of an electric toothbrush based on big data according to claim 5, wherein The behavior rhythm linkage recognition module includes: The offset cycle screening sub-module, according to the marked cycle time range in the manifestation of the concentrated change of the brushing behavior time, extracts the cycle data with a coincidence degree exceeding two days with the cleaning habit change section, performs a cross comparison on the cycle boundaries, and screens out the eligible time periods to generate a combined offset cycle set; The vibration sequence extraction sub-module calls the cycle data in the combined offset cycle set, extracts the vibration strength alternating numerical sequence of the daily brushing behavior, constructs a vibration change array in sequence and marks the corresponding time interval position to obtain a continuous vibration interval sequence; The jump frequency judgment sub-module, based on the continuous vibration interval sequence, judges the difference between adjacent vibration interval values, and compares them with the stable vibration interval limit value to identify the vibration section with an excessive fluctuation range, using the formula: Where: vector vector ||·||1 represents the 1-norm, that is, the sum of the absolute values of the elements of the vector difference; Calculate to obtain the vibration interval offset value, and judge whether an abnormal jump behavior is formed according to the comparison result between the vibration interval offset value and the offset amplitude threshold, and obtain a jump frequency abnormal flag sequence; Among them, F dev is the vibration interval offset value, N seg is the total number of jump sections, P r is the average vibration intensity of the r-th section, T r is the time difference between the start and end of this section, C r is the number of cleaning areas covered in this section, Θ is the upper threshold of the vibration stable time interval, D r is the behavior duration of the r-th section.

7. The system for analyzing the user behavior of an electric toothbrush based on big data according to claim 6, wherein The abnormal trend health reminder module includes: The abnormal cycle statistics sub-module counts all the marked abnormal records in the jump frequency abnormal flag sequence, extracts the corresponding cycle number information, removes duplicates and summarizes the abnormal cycle numbers, and generates a list of abnormal cycle numbers; The drift feature comparison sub-module calls the list of abnormal cycle numbers, matches the drift direction label data in the concentrated change performance of the brushing behavior time, identifies whether there are marked types of both extended jump intervals and extended drift regions within the cycle, screens and summarizes the cycles with dual markings, and obtains a sequence of healthy abnormal cycles that can be classified; The healthy cycle determination sub-module extracts the start and end time fields of the cycle according to the sequence of healthy abnormal cycles that can be classified, performs annotation integration in chronological order after unifying the format, establishes a classification information record based on the cycle behavior characteristics, and obtains the analysis results of the electric toothbrush user behavior.

Images