Method, system and device for detecting behavior mode anomaly of home elderly and storage medium

Abstract

The invention provides a method, a system, equipment and a medium for detecting an abnormal behavior pattern of an elderly living at home, and belongs to the field of home nursing, and the method comprises the following steps: obtaining a behavior sequence set # imgabs0 # of the elderly living at home in the past m days, and obtaining a set Bi of an ith behavior bi in the past m days from # imgabs1 # according to a behavior type, defining a time distance formula of the behavior bi occurring at the time points x and y; extracting time regularity and frequency regularity of behaviors bi in the set Bi; the method comprises the following steps: acquiring a behavior sequence Bi of an ith day, judging regular behaviors according to time regularity and frequency regularity to obtain a set Breg of all regular behaviors, converting the behavior sequence Bi of the ith day into regular behavior characteristic word strings to obtain a characteristic word string set # imgabs2 #, and identifying an abnormal behavior pattern in the characteristic word string set # imgabs3 # by using an unsupervised learning clustering algorithm. According to the invention, the abnormal behavior pattern can be accurately extracted.

Description

Technical Field

[0001] The present invention belongs to the field of home care, and in particular relates to a method, system, device and storage medium for detecting abnormal behavior patterns of elderly people at home. Background Art

[0002] The increasingly serious problem of population aging has made the medical industry face more and more challenges, such as the increase in health care costs and the shortage of nursing staff resources, forcing people to find ways to deal with the problem of aging. The development of wireless sensor networks and the Internet of Things has promoted the smart medical care of the elderly from hospitals and clinics to homes. At the same time, the elderly prefer to live in their own homes. Therefore, home care has become an active research field, mainly involving a variety of technologies, including monitoring systems, behavior recognition, and violation detection. Monitoring systems are designed to record residents' activities through videos, wearable devices or wireless sensor networks. Wireless sensor networks have attracted much attention due to their privacy and user-friendliness, which do not require the elderly to wear any equipment. In addition, many behavior recognition algorithms have been proposed to identify behaviors by analyzing the relationship between behaviors and sensor data. Irregular behaviors can indicate some unsafe situations, such as changes in the health status of the elderly, deterioration of the elderly's mood or instability. Therefore, irregularity detection can effectively improve the quality of life of elderly people living alone.

[0003] Irregularity detection can be mainly divided into two types: explicit and implicit. Explicit irregularity detection aims to identify irregular behaviors in real time and requires corresponding caregivers to deal with abnormal behaviors immediately, such as accelerated heart rate and fall detection. Usually, some clear irregularity features are known, so some specific sensors can be deployed to capture these known features to identify irregularities. However, the features of implicit irregularities are uncertain, which makes detection difficult. For example, discomfort is a kind of implicit irregularity, which may cause changes in the time or sequence of certain behaviors, such as sleeping and eating, but may not cause drastic changes in sensor values. In this case, in-depth analysis of behaviors or behavior patterns is required to identify irregularities and alert caregivers.

[0004] Since the characteristics of implicit irregularity are difficult to describe, some existing algorithms of this type only consider the irregularity of a single behavior, but do not discuss the irregularity of a combination of behaviors. Some studies also take one-day behavior as the basic analysis element and consider the different effects of different behaviors on implicit irregularity detection. However, the extraction of regular behavior patterns in the above studies is rough and imprecise, resulting in inaccurate results for implicit irregular behavior detection. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method for detecting abnormal behavior patterns of elderly people at home, comprising the following steps:

[0006] Get the behavior sequence set of the elderly at home in the past m days According to the behavior category, from the behavior sequence set Get the i-th behavior b in the past m days i The set B i ;

[0007] Define behavior b i The time distance formula between time point x and time point y is combined with the time distance formula to obtain set B i Behavior in b i The time regularity and frequency regularity of i Whether the start time and end time of each behavior in satisfies the time regularity, and whether the occurrence frequency of a behavior satisfies the frequency regularity, if so, it is determined to be a regular behavior;

[0008] Obtained by the determination method of regular behavior All regular behaviors in constitute the regular behavior set B reg ; B reg The behavior sequence B that occurs on day i i Each regular behavior in is converted into a feature string to obtain a feature string set Use unsupervised learning clustering algorithm to cluster feature strings Classify and identify abnormal behavior patterns based on the classification results.

[0009] Preferably, the unsupervised learning clustering algorithm is a density clustering algorithm DBSCAN, which uses the density clustering algorithm DBSCAN to cluster the feature string set Classify and identify abnormal behavior patterns based on the classification results, specifically: Clustering is performed, regular behavior patterns are extracted based on the clustering results, and feature strings with noise are determined as abnormal behavior patterns.

[0010] Preferably, the time regularity includes the regularity of occurrence time and the regularity of end time, and the judgment set B i Whether the start time and end time of a certain behavior in satisfies the time regularity must satisfy both the occurrence time regularity and the end time regularity.

[0011] Preferably, the time distance formula is:

[0012]

[0013] in, Represents set B i The jth element in Represents set B i The lth element in AD(x, y) represents a function that ignores the date and calculates the absolute difference between time points x and y. express The start time of occurrence, express The start time of the occurrence.

[0014] Preferably, the time regularity is:

[0015]

[0016] In the formula, Indicates regular behavior The mean square error of the start time, Indicates regular behavior The mean square error of the end time, and The regularity thresholds represent the start and end time of regular behavior respectively

[0017] Preferably, the frequency regularity is:

[0018]

[0019] In the formula, regular behavior The mean square error of the frequency of occurrence, Represents the frequency regularity threshold.

[0020] Preferably, the density clustering algorithm DBSCAN is used to obtain the set B i Behavior in b i The temporal regularity of i The frequency regularity.

[0021] The present invention also provides a system for detecting abnormal behavior patterns of elderly people at home, comprising:

[0022] The behavior sequence acquisition module is used to obtain the behavior sequence set of the elderly at home in the past m days According to the behavior category, from the behavior sequence set Get the i-th behavior b in the past m days i The set B i ;

[0023] Regular behavior determination module, used to define behavior b i The time distance formula between time point x and time point u is combined with the time distance formula to obtain set B i Behavior in b i The time regularity and frequency regularity of iWhether the start time and end time of each behavior in satisfies the time regularity, and whether the occurrence frequency of a behavior satisfies the frequency regularity, if so, it is determined to be a regular behavior;

[0024] Behavior pattern anomaly detection module, used to obtain the judgment method based on regular behavior All regular behaviors in constitute the regular behavior set B reg ; B reg The behavior sequence B that occurs on day i i Each regular behavior in is converted into a feature string to obtain a feature string set Use unsupervised learning clustering algorithm to cluster feature strings Classify and identify abnormal behavior patterns based on the classification results.

[0025] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the method for detecting abnormal behavior patterns of elderly people at home.

[0026] The present invention also provides a computer-readable storage medium, which stores a computer program. The computer program is suitable for loading by a processor to execute the method for detecting abnormal behavior patterns of elderly people at home.

[0027] The abnormal behavior pattern detection method, system, device and storage medium provided by the present invention have the following beneficial effects:

[0028] The present invention obtains the behavior set of the past multiple days, and can determine the regular behavior in the behavior set by extracting the time regularity and frequency regularity in the behavior set. This process takes the behavior of multiple days as the analysis factor, and considers the different impacts of different behaviors on the detection of implicit irregularity. The idea of ​​implicit irregularity detection is transformed into the feature of extracting regular behavior patterns, making the detection result more refined. By adopting an unsupervised clustering algorithm to perform multi-category clustering on daily behavior patterns, it is possible to facilitate a more detailed analysis of the behavior patterns, thereby accurately extracting abnormal behavior patterns. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiment of the present invention and its design scheme, the following briefly introduces the drawings required for this embodiment. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 This is a flow chart of a method for detecting abnormal behavior patterns of elderly people at home according to an embodiment of the present invention;

[0031] Figure 2 Extracting flow charts for the regularity of daily behavior sequences;

[0032] Figure 3 A construction method for representing the characteristics of daily behavior sequences;

[0033] Figure 4 The relationship between regular behavior patterns, feature representations, and regular behaviors. DETAILED DESCRIPTION

[0034] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0035] In the description of the present invention, it is to be understood that the terms “center”, “longitudinal”, “lateral”, “length”, “width”, “thickness”, “up”, “down”, “front”, “back”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inside”, “outside”, “axial”, “radial”, “circumferential”, etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the technical solutions of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0036] In addition, the terms "first", "second", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance. In the description of the present invention, it should be noted that, unless otherwise clearly specified or limited, the terms "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. In the description of the present invention, unless otherwise specified, "plurality" means two or more, which will not be described in detail here.

[0037] Example

[0038] The present invention provides a method for detecting abnormal behavior patterns of elderly people at home, specifically Figure 1 As shown, the following steps are included:

[0039] Step 1: Obtain the behavior sequence set of the elderly at home in the past m days According to the behavior category, from the behavior sequence set Get the behavior of the past m days in b i The set B i ; define behavior b i The time distance formula between time point x and time point y is combined with the density clustering algorithm DBSCAN to obtain set B i Behavior in b i The temporal regularity of i The frequency regularity of; judgment set B i Whether the start time and end time of a certain behavior in satisfies the time regularity, and whether the occurrence frequency of a certain behavior satisfies the frequency regularity, if so, it is determined to be a regular behavior.

[0040] Assume that a daily behavior monitoring system is installed in the apartment of an elderly person living alone, and the sensors of the monitoring system can collect the behavior of the elderly person. The present invention denotes the behavior set of interest as B = {b 1 , b 2 , ..., b |B|}, the behavior sequence occurring on day i is recorded as Each behavior i，j ∈B i The specific information can be expressed as (name, start_time, end_time), where name∈B represents the name of the behavior, start_time and end_time represent b respectively. i，j The start time and end time of the occurrence, the behavior sequence set collected in the past m days is recorded as

[0041] Assuming that the elderly have regular behavior patterns, they can be further divided into unrelated behavior pattern categories. For example, the elderly may have a regular behavior pattern on weekdays, such as getting up around 5 o'clock, eating breakfast, lunch and dinner at 6 o'clock, 12 o'clock and 18 o'clock respectively, and going to bed around 20 o'clock. However, the elderly may also have other types of regular behavior patterns, such as taking the children out for dinner on Saturdays and going to the market every Wednesday. Represents a regular behavior pattern, which contains some regular behavior sequences. represents a set of q behavior patterns, which can be obtained from the behavior sequence set If the behavior sequence on a certain day does not belong to any behavior pattern, it can be considered that the behavior on that day is abnormal.

[0042] The regularity of daily behavior sequences is composed of regular behaviors. Therefore, the following will define regular behaviors and then explain the extraction of regularities in detail.

[0043] (1) Definition of regular behavior

[0044] A behavior is regular if it occurs regularly at a certain time of day and over a certain number of days.

[0045] make Represents the i-th regular behavior, which has 8 attributes, namely:

[0046] 1) id indicates regular behavior A unique identifier for a . For example and

[0047] 2) name indicates regular behavior The name of It should be noted that different regular behaviors may have the same behavior name but different ids.

[0048] 3) Indicates regular behavior The statistical mean of the start times.

[0049] 4) Indicates regular behavior The mean square error of the start times.

[0050] 5) Indicates regular behavior The statistical mean of the end time.

[0051] 6) Indicates regular behavior The mean square error of the end times. .

[0052] 7) Indicates regular behavior The statistical mean of the frequency of occurrence.

[0053] 8) Indicates regular behavior The mean square error of the frequencies of occurrence.

[0054] Regular behavior Will regularly Starting nearby, Let μ represent the time unit, for example, μ is set to 1 day. Then, if the regular behavior Occurs regularly every day times, then its frequency of occurrence is For example, in every unit time μ, the behavior of "eating lunch" occurs once, so the frequency of this behavior is 1. If the elderly go to the vegetable market once every Wednesday, then the frequency of "going to the vegetable market" is 1 / 7. Based on the definition of regular behavior, the following will describe in detail how to identify regular behavior.

[0055] (2) Identification of regular behaviors

[0056] The present invention will identify regular behaviors through the constraints of time regularity and frequency regularity. These constraints are mainly composed of The following is an introduction to the regularity recognition process, including time regularity constraint checks and frequency regularity constraint checks. Figure 2 shown.

[0057] The first step is to find a behavior that satisfies the time regularity constraint. The time regularity constraint ensures a regular behavior Regularly starts around a certain time and regularly ends around a certain time. Indicates that the behavior in the past m days is b i For example, b j，k .name = 'eating lunch' and m = 365, then all "eating lunch" in the past year will be put into set B i .make For set B i The jth element in For set B i The lth element in Expressing behavior and behavior The distance from the start time, then the distance It can be calculated by the following formula:

[0058]

[0059] Where AD(x, y) represents a function that can ignore the date and calculate the absolute difference between time point x and time point y. For example, let x = '2022-01-02 08:00' and y = '2022-01-12 08:30', then AD(x, y) will return 30 minutes.

[0060] With the distance calculation formula, we can use the DBSCAN clustering algorithm to extract set B i Behavior in b i After DBSCAN clustering, each behavior instance Will be clustered into a group G with temporal regularity TR or is considered as noise. Indicates all B i G obtained after DBSCAN clustering TR The collection, that is,

[0061]

[0062] gather The number of elements in is affected by the DBSCAN parameters. DBSCAN has two parameters, min_pts and Eps, where min_pts represents the minimum number of points in a group and Eps represents the minimum distance between two data points in the same group. These two parameters will be discussed in the performance analysis below.

[0063] make Representing time regularity group The corresponding behavior name, then the group All behavior instances in a have the same behavior name And they all start at similar times. In other words, the behavior has temporal regularity. They can be calculated according to the following formulas.

[0064]

[0065]

[0066]

[0067]

[0068] Assume that, in the past m days, the behavior The start time and end time of the behavior obey the normal distribution, then the confidence interval of the start time and end time of the behavior is defined as and The size of the variance indicates the degree of discreteness of the data, that is, the degree of regularity of the behavior. and Represents the start time and end time regularity thresholds respectively. If the following formula is satisfied, the behavior is considered to have ended with temporal regularity.

[0069]

[0070] make Represents the set of behaviors that satisfy the temporal regularity constraint.

[0071] Next, the second step is to check the behavior Whether the frequency regularity is satisfied. The present invention uses Fourier transform to find the periodicity of the behavior. Let F s Represents the sampling frequency, which is the behavior Constructing a time series Time series F is collected in a unit time μ s For example, assuming that the time unit μ is set to one day, F s Set to 24*2. This means that if the behavior Every half hour, once it occurs, check it. The result of the check will be 1 or 0, indicating that it occurred or did not occur, respectively. In addition, There are 24*2 elements in a day, and There are 24*2*m elements in the past m days.

[0072] So, The power spectral density nonparametric estimate P i (f) can be calculated by the following formula.

[0073]

[0074] P i (f) is a function of frequency f, where Let P trh represents the periodic threshold, let Behavior The frequency set of Properties Can be calculated by the following formula:

[0075]

[0076]

[0077] Similarly, let Indicates the regularity threshold of frequency, behavior If the following equation is satisfied, then the behavior satisfies the frequency regularity constraint.

[0078]

[0079] If a behavior satisfies both the time regularity and frequency regularity constraints, it is called a regular behavior. Represents the obtained regular behavior-set. Among them, B reg The regularity extracted from daily behavior sequences. Figure 1 As shown, the behavioral dataset collected in the past m days Described as input. In the regularity recognition layer, in the set The data is partitioned by the behavior name and DBSCAN clustered by the start time to obtain the temporal regularity clusters. Then, the time regularity constraint check and frequency regularity constraint check are performed, and the regular behavior set B is output. reg The next stage will represent the daily behavior sequences based on the extracted regularities.

[0080] Step 2: Obtain based on the determination method of regular behavior The set of all regular behaviors in B reg ; Set B reg The behavior sequence B that occurs on day i i Each regular behavior in is converted into a feature string to obtain a feature string set

[0081] Use B reg The regular behaviors extracted from represent each daily behavior sequence That is, keep the regular behaviors in the daily behavior sequence and remove the irregular behaviors. Recall that the symbol represents the sequence of behaviors identified on day i. i，j ∈B i will be checked to see if this is a regular behavior. If so, b i，j will be retained and represented by the corresponding regular behavior. Otherwise, b i，j will be deleted. Therefore, each daily behavior sequence can be represented by regular behaviors extracted from the previous stage. is an example of regular behavior i，j. The following is the characteristic representation of the regular behavior b i，j Instance feature representation The generation method of .

[0082] Let the Boolean variable η i，j，k Indicates behavior i，j ∈B i Is it a regular behavior? An example of

[0083]

[0084] It is possible that multiple regular behaviors have the same name. For example, the "TakeMedicine" behavior may appear regularly around 8:00, 13:00, and 19:00. That is, three regular behaviors with the same name but different time patterns are extracted from the regular extraction behavior sequence. Therefore, behavior b i，j The mapping to regular behavior is not clear. Then, the next step is to resolve this ambiguity. First, define the behavior instance b i，j and regular behavior distance.

[0085]

[0086] Then find b according to the following formula i，j The best feature representation breg .id.

[0087]

[0088] Each behavior instance b i，j ∈B i After the above processing, we can get the behavior sequence B i The feature representation of .

[0089] make Indicates behavior sequence B i Then It is composed only of regular behaviors, which is a regular behavior sequence. There is a frequency of occurrence, which is determined by the regular behavior The minimum number of occurrences of an instance is determined. express The frequency of occurrence is Medium regular behavior The frequency is So Frequency of occurrence It can be calculated by the following formula.

[0090]

[0091] Figure 3 A method for constructing feature representation of daily behavior sequences is given. Figure 3 As shown, the daily behavior sequence Contains multiple behavior instances, such as Figure 3 At the same time, Figure 3 The set B in the right half of reg Contains 6 regular behaviors. Regular behaviors The frequency of is equal to 1, which means that the behavior 'GetUp' occurs once a day, and The frequency of is equal to 1 / 3, which means that the behavior 'G0toMarket' occurs once every 3 days. According to the following formula, we can get B 1 The characteristic representation of each behavior in Figure 3 As shown, behavior instance b 1，1 ∈B 1 The feature representation of is the character 'a' because it maps the regular behavior is 'a'. Example b 1，2 ∈B 1 There is no feature representation because b 1，2 .name does not appear in the regular behavior set B reg Similarly, instance b 1，3 The feature representation of . is the character 'b'. In addition, three regular behaviors, namely and With instance b 1，4 have the same behavior name, then based on the distance calculation, instance b 1，4 The characteristic is denoted as 'e', ​​such as Figure 3 As shown, we can finally get Further, according to the above formula, Frequency is 1 / 3.

[0092] In each behavioral sequence After conversion to feature representation, we can get a set of feature strings In the next stage, the feature string set Unsupervised learning clustering algorithms are applied to discover daily behavior patterns and identify abnormal behavior sequences.

[0093] Step 3: Use unsupervised learning clustering algorithm to identify the characteristic string set Abnormal behavior patterns in .

[0094] For the feature string set The goal of this step is to discover some regular behavior patterns and identify abnormal behavior sequences. Specifically, each feature string is a behavior pattern, each characteristic string Each behavior string is checked to see if it should belong to each regular behavior pattern or abnormal behavior sequence. To achieve this goal, this stage includes three main tasks, including weight assignment, distance calculation, and feature representation clustering tasks. In the weight assignment task, each behavior is assigned a weight to emphasize the importance of its regularity. Then, the distance calculation task aims to define and calculate the distance between any two behavior strings. Finally, in the feature representation clustering task, the DBSCAN processing learning algorithm is applied to discover regular patterns and identify abnormal behavior patterns. The detailed approach for each task is given below.

[0095] (1) Weight distribution

[0096] In calculating string and Before calculating the distance between the two, each regular behavior should be associated with a weight to emphasize their importance to the regularity. The weight of the regular behavior will be based on the time regularity and frequency regularity. Indicates regular behavior The weight of the regular behavior has been defined above. The length of the confidence interval for the start time is Its value indicates the regularity of the behavior. If the value of is small, the temporal regularity is strong. On the contrary, if The larger the value of , the weaker the temporal regularity. Similarly, The value of indicates regular behavior The degree of frequency regularity, A larger value indicates that the behavior The frequency regularity is strong, on the contrary, The smaller the value of is, the weaker the frequency regularity of the behavior is. The regularity of can be determined by both time regularity and frequency regularity. of It can be calculated according to the following formula.

[0097]

[0098] (2) Distance calculation

[0099] make Indicates the characteristic string and The distance between Figure 3 As shown, The feature of is represented as a feature string, so Can also be defined as a string from the characteristics Transformed to Here, basic character operations include inserting, deleting, or modifying.

[0100] Assume the signature string When performing the kth basic operation, the operation rule behavior is Then the cost of the kth basic operation is recorded as Its value is equal to As shown in the following formula.

[0101]

[0102] Assumed signature string After K basic character operations, the characteristic string can be obtained from Convert to The cumulative operation cost on Its value can be calculated according to the following formula.

[0103]

[0104] In fact, the characteristic string After K basic character operations, the characteristic string can be obtained There are multiple combinations of operations. Therefore, the distance It is defined as the minimum cost of multiple combinations of operations and can be calculated according to the following formula.

[0105]

[0106] (3) Feature Representation Clustering

[0107] Finally, DBSCAN is applied to the set Clustering is performed to discover regular behavior patterns and detect abnormal behavior sequences. The two parameters of DBSCAN, namely and Need to be determined. Parameters represents the minimum number of characteristic strings in a category, and It is expressed as the minimum distance between two feature strings in the same category. Represents the result of clustering by DBSCAN. The elements in the set Represents a category, each category Represents a regular behavior pattern, consisting of some similar feature strings, that is,

[0108] make Represents a set of regular behaviors B reg The minimum frequency of occurrence of regular behavior in the algorithm. The following is a restriction on the application of DBSCAN algorithm.

[0109] Criteria for the minimum number of elements in a regular pattern class:

[0110] Parameters of DBSCAN The following should be satisfied:

[0111]

[0112] Here, m represents the number of days of observed data.

[0113] The following will further explain the above formula and derive Calculation of regular behavior patterns The frequency of occurrence is Its value will be represented by the feature corresponding to the pattern for each element in the set Determine. Feature string The frequency of the regular behavior that occurs on day i Decision. Regular behavior The frequency of Similarly, the behavior pattern Frequency Defined as a pattern The minimum occurrence frequency of the Chinese character string is calculated according to the following formula.

[0114]

[0115] From the above formula, we can see that the frequency Composed of All possible regular behaviors The minimum frequency is determined by Figure 4 The relationship between regular behavior patterns, feature representations, and regular behavior is shown.

[0116] make Indicates B reg The regular behavior with the minimum frequency in , then:

[0117]

[0118]

[0119] So, Will not be greater than The frequency of any behavior pattern For any regular behavior pattern Will satisfy The minimum number of strings in a pattern set in the past m days can be expressed as Considering The standard deviation The minimum number of elements in a regular behavior pattern set is:

[0120]

[0121] After DBSCAN clustering, each feature string will be classified into a regular behavior pattern Finally, we can get a set of regular behavior patterns. Abnormal behavior sequences can also be detected.

[0122] The present invention also provides a system for detecting abnormal behavior patterns of elderly people at home, including a behavior sequence acquisition module, a regular behavior determination module and a behavior pattern abnormality detection module. The behavior sequence acquisition module is used to obtain the behavior sequence set of the elderly at home in the past m days. According to the behavior category, from the behavior sequence set Get the i-th behavior b in the past m days i The set B i ; Regular behavior determination module is used to define behavior b i The time distance formula between time point x and time point u is combined with the time distance formula to obtain set Bi Behavior in b i The time regularity and frequency regularity of i The start time and end time of each behavior in the model are used to determine whether they meet the time regularity, and whether the frequency of a behavior meets the frequency regularity. If so, it is determined to be a regular behavior. The behavior pattern anomaly detection module is used to obtain the regular behavior according to the determination method of the regular behavior. All regular behaviors in constitute the regular behavior set B reg ; B reg The behavior sequence B that occurs on day i i Each regular behavior in is converted into a feature string to obtain a feature string set Use unsupervised learning clustering algorithm to cluster feature strings Classify and identify abnormal behavior patterns based on the classification results.

[0123] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute a method for detecting abnormal behavior patterns of elderly people at home.

[0124] The present invention also provides a computer-readable storage medium, which stores a computer program. The computer program is suitable for loading by a processor to execute a method for detecting abnormal behavior patterns of elderly people at home.

[0125] The embodiments described above are only preferred specific implementation modes of the present invention, and the protection scope of the present invention is not limited thereto. Any simple changes or equivalent replacements of the technical solutions that can be obviously obtained by any technician familiar with the field within the technical scope disclosed in the present invention belong to the protection scope of the present invention.

Claims

1. A method for detecting abnormal behavior patterns of elderly people at home, characterized in that: The steps include: Get the behavior sequence set of the elderly at home in the past m days According to the behavior category, from the behavior sequence set Get the i-th behavior b in the past m days i The set B i ; Define behavior b i The time distance formula between time point x and time point y is combined with the time distance formula to obtain set B i Behavior in b i The time regularity and frequency regularity of i Whether the start time and end time of each behavior in satisfies the time regularity, and whether the occurrence frequency of a behavior satisfies the frequency regularity, if so, it is determined to be a regular behavior; Obtained by the determination method of regular behavior All regular behaviors in constitute the regular behavior set B reg ; B reg The behavior sequence B that occurs on day i i Each regular behavior in is converted into a feature string to obtain a feature string set Use unsupervised learning clustering algorithm to cluster feature strings Classify and identify abnormal behavior patterns based on the classification results.

2. The method for detecting abnormal behavior patterns of elderly people at home according to claim 1, characterized in that: The unsupervised learning clustering algorithm is a density clustering algorithm DBSCAN, which is used to cluster the feature string set Classify and identify abnormal behavior patterns based on the classification results, specifically: Clustering is performed, regular behavior patterns are extracted based on the clustering results, and feature strings with noise are determined as abnormal behavior patterns.

3. The method for detecting abnormal behavior patterns of elderly people at home according to claim 1, characterized in that: The time regularity includes the regularity of occurrence time and the regularity of end time. i Whether the start time and end time of a certain behavior in satisfies the time regularity requires satisfying both the occurrence time regularity and the end time regularity.

4. The method for detecting abnormal behavior patterns of elderly people at home according to claim 1, characterized in that: The time distance formula is: in, Represents set B i The jth element in Represents set B i The lth element in AD(x,y) represents a function that ignores the date and calculates the absolute difference between time points x and y. express The start time of occurrence, express The start time of the occurrence.

5. The method for detecting abnormal behavior patterns of elderly people at home according to claim 1, characterized in that: The temporal regularity is: In the formula, Indicates regular behavior The mean square error of the start time, Indicates regular behavior The mean square error of the end time, and The regularity thresholds represent the start and end time of regular behavior respectively.

6. The method for detecting abnormal behavior patterns of elderly people at home according to claim 1, characterized in that: The frequency regularity is: In the formula, regular behavior The mean square error of the frequency of occurrence, Represents the frequency regularity threshold.

7. The method for detecting abnormal behavior patterns of elderly people at home according to claim 1, characterized in that: Use density clustering algorithm DBSCAN to obtain set B i Behavior in b i The temporal regularity of i The frequency regularity.

8. A system for detecting abnormal behavior patterns of elderly people at home, characterized in that: include: The behavior sequence acquisition module is used to obtain the behavior sequence set of the elderly at home in the past m days According to the behavior category, from the behavior sequence set Get the i-th behavior b in the past m days i The set B i ; Regular behavior determination module, used to define behavior b i The time distance formula between time point x and time point y is combined with the time distance formula to obtain set B i Behavior in b i The time regularity and frequency regularity of i Whether the start time and end time of each behavior in satisfies the time regularity, and whether the occurrence frequency of a behavior satisfies the frequency regularity, if so, it is determined to be a regular behavior; Behavior pattern anomaly detection module, used to obtain the judgment method based on regular behavior All regular behaviors in constitute the regular behavior set B reg ; B reg The behavior sequence B that occurs on day i i Each regular behavior in is converted into a feature string to obtain a feature string set Use unsupervised learning clustering algorithm to cluster feature strings Classify and identify abnormal behavior patterns based on the classification results.

9. A computer device, characterized in that: It comprises a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the method for detecting abnormal behavior patterns of elderly people at home as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the method for detecting abnormal behavior patterns of elderly people at home as described in any one of claims 1 to 7.

Images