CN118038548A - Abnormal behavior detection method, device, electronic equipment and storage medium - Google Patents

Abnormal behavior detection method, device, electronic equipment and storage medium Download PDF

Info

Publication number
CN118038548A
CN118038548A CN202410139881.9A CN202410139881A CN118038548A CN 118038548 A CN118038548 A CN 118038548A CN 202410139881 A CN202410139881 A CN 202410139881A CN 118038548 A CN118038548 A CN 118038548A
Authority
CN
China
Prior art keywords
model
behavior
data
sequence
target user
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202410139881.9A
Other languages
Chinese (zh)
Inventor
胡赟
余小燕
王冬青
童亚平
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Telecom Corp Ltd
Original Assignee
China Telecom Corp Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Telecom Corp Ltd filed Critical China Telecom Corp Ltd
Priority to CN202410139881.9A priority Critical patent/CN118038548A/en
Publication of CN118038548A publication Critical patent/CN118038548A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
    • G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
    • G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
    • G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/62—Extraction of image or video features relating to a temporal dimension, e.g. time-based feature extraction; Pattern tracking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02—Alarms for ensuring the safety of persons
    • G08B21/04—Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons
    • G08B21/0407—Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons based on behaviour analysis
    • G08B21/043—Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons based on behaviour analysis detecting an emergency event, e.g. a fall
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02—Alarms for ensuring the safety of persons
    • G08B21/04—Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons
    • G08B21/0438—Sensor means for detecting
    • G08B21/0476—Cameras to detect unsafe condition, e.g. video cameras
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Emergency Management (AREA)
  • Psychiatry (AREA)
  • Business, Economics & Management (AREA)
  • Gerontology & Geriatric Medicine (AREA)
  • Social Psychology (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Psychology (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Human Computer Interaction (AREA)
  • Alarm Systems (AREA)

Abstract

The application discloses an abnormal behavior detection method, an abnormal behavior detection device, electronic equipment and a storage medium, wherein the abnormal behavior detection method comprises the following steps: acquiring monitoring data for a target user; preprocessing monitoring data to obtain a data sequence; inputting the data sequence into a behavior recognition model, and determining the behavior sequence of the target user according to the output of the behavior recognition model; and inputting the data sequence and the behavior sequence into an anomaly detection model, and determining whether the target user has an anomaly behavior according to the output of the anomaly detection model. By applying the technical scheme provided by the application, the abnormal behavior of the user can be found in time, the intervention is performed in time, the disease development is delayed, and the treatment cost is saved.

Description

Abnormal behavior detection method, device, electronic equipment and storage medium
Technical Field
The present application relates to the field of computer applications, and in particular, to a method and apparatus for detecting abnormal behavior, an electronic device, and a storage medium.
Background
As the population ages, alzheimer's disease patients (Alzheimer' S DISEASE PATIENTS, ADP) increase. Alzheimer's disease is a progressive neurodegenerative disease with hidden onset and is characterized clinically by generalized dementia such as dysmnesia, aphasia, disuse, disrecognition, impairment of visual space skills, dysfunctions of execution, and personality and behavioral changes.
Alzheimer's disease usually has three stages of development, mild, moderate and severe. Most patients with Alzheimer's disease are neglected in the mild stage, and the disease is usually developed to the moderate or severe stage by hospital examination, the moderate and severe stages are irreversible stages, and the treatment cost is high. If the patient is diagnosed at the stage of mild cognitive impairment, the patient intervenes in time, so that the disease development can be delayed, and the treatment cost is saved.
How to find the abnormal behavior of the elderly in time is a technical problem which needs to be solved at present.
Disclosure of Invention
The application aims to provide an abnormal behavior detection method, an abnormal behavior detection device, electronic equipment and a storage medium, and the abnormal behavior of a user is detected in time.
In order to solve the technical problems, the application provides the following technical scheme:
In a first aspect, there is provided an abnormal behavior detection method, including:
Acquiring monitoring data for a target user;
Preprocessing the monitoring data to obtain a data sequence;
inputting the data sequence into a behavior recognition model, and determining the behavior sequence of the target user according to the output of the behavior recognition model;
and inputting the data sequence and the behavior sequence into an abnormality detection model, and determining whether the target user has abnormal behaviors according to the output of the abnormality detection model.
Optionally, the behavior recognition model and/or the anomaly detection model comprises at least one of the following model structures:
A multi-head convolutional neural network-long-short-term memory network model structure;
attention mechanism-convolutional neural network-long-short-term memory network model structure.
Optionally, in the multi-head-convolutional neural network-long-short-term memory network model structure, a plurality of convolutional neural networks are connected in parallel, each convolutional neural network receives an input sequence and processes the input sequence, outputs of the plurality of convolutional neural networks enter a full-connection layer, and the full-connection layer is input into the long-term memory network after merging and interpretation, and a prediction result is output.
Optionally, in the structure of the attention mechanism-convolutional neural network-long-short-term memory network model, the convolutional neural network comprises a convolutional layer, a pooling layer and a full-connection layer which are sequentially connected, the convolutional neural network receives an input sequence and performs feature extraction on the input sequence, the long-term memory network learns a data change rule according to the features extracted by the convolutional neural network, weights of different features are iterated through an attention part, weights are reapplied to an output result of the long-term memory network, and a prediction result is output.
Optionally, the preprocessing the monitoring data to obtain a data sequence includes:
carrying out standardized processing on the monitoring data;
and dividing the standardized monitoring data by utilizing a sliding window to obtain a data sequence.
Optionally, the obtaining monitoring data for the target user includes:
acquiring monitoring original data aiming at a target user;
and carrying out missing value filling processing and/or noise reduction processing on the monitoring original data to obtain the monitoring data aiming at the target user.
Optionally, the behavior recognition model and the abnormality detection model are obtained in advance by:
obtaining a sample data set, wherein the sample data set comprises a plurality of groups of monitoring sample data, and each group of monitoring sample data corresponds to a user sample respectively;
Dividing the sample dataset into a training dataset and a test dataset;
Training an initial recognition model by using the training data set, and adjusting model parameters of the initial recognition model, wherein the initial recognition model is an initial model of the behavior recognition model;
Training an initial detection model by utilizing the training data set and the output result of the initial recognition model, and adjusting model parameters of the initial detection model, wherein the initial detection model is an initial model of the abnormal detection model;
And testing the trained initial recognition model and the trained initial detection model by using the test data set, and determining the trained initial recognition model as the behavior recognition model and determining the trained initial detection model as the abnormal detection model under the condition that a test result meets a preset requirement.
In a second aspect, there is provided an abnormal behavior detection apparatus including:
the first acquisition module is used for acquiring monitoring data aiming at a target user;
the second acquisition module is used for preprocessing the monitoring data to acquire a data sequence;
The first determining module is used for inputting the data sequence into a behavior recognition model and determining the behavior sequence of the target user according to the output of the behavior recognition model;
And the second determining module is used for inputting the data sequence and the behavior sequence into an abnormality detection model, and determining whether the target user has abnormal behaviors or not according to the output of the abnormality detection model.
In a third aspect, an electronic device is provided, comprising:
a memory for storing a computer program;
A processor for implementing the steps of the abnormal behavior detection method according to the first aspect when executing the computer program.
In a fourth aspect, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the abnormal behavior detection method according to the first aspect.
In a fifth aspect, there is provided a computer program product comprising computer instructions stored in a computer readable storage medium and adapted to be read and executed by a processor to cause a computer device having the processor to perform the steps of the abnormal behaviour detection method as described in the first aspect.
By applying the technical scheme provided by the embodiment of the application, after the monitoring data aiming at the target user are obtained, the monitoring data are preprocessed to obtain the data sequence, the data sequence is identified by utilizing the behavior identification model, the behavior sequence of the target user is determined, the data sequence and the behavior sequence are input into the abnormality detection model, whether the target user has abnormal behaviors can be determined according to the output of the abnormality detection model, the abnormal behaviors of the user can be found in time through the combined detection of the behavior identification model and the abnormality detection model, the intervention is performed in time, the disease development is helped to be delayed, and the treatment cost is saved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application as claimed.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of an application scenario according to an embodiment of the present application;
FIG. 2 is a flowchart of an abnormal behavior detection method according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a convolutional neural network feature extraction process in an embodiment of the present application;
FIG. 4 is a schematic diagram of a model structure according to an embodiment of the present application;
FIG. 5 is a schematic diagram of data segmentation in an embodiment of the present application;
FIG. 6 is a schematic diagram of a model acquisition process in accordance with an embodiment of the present application;
FIG. 7 is a schematic diagram showing the performance of a conventional CNN-LSTM model in a model comparison according to an embodiment of the present application;
FIG. 8 is a schematic diagram showing the performance of the MH-CNN-LSTM model in a model comparison according to an embodiment of the present application;
FIG. 9 is another schematic diagram of the performance of a conventional CNN-LSTM model in a model comparison of an embodiment of the application;
FIG. 10 is another schematic diagram of the performance of the MH-CNN-LSTM model in a model comparison according to an embodiment of the application;
FIG. 11 is a process diagram of an example application in an embodiment of the application;
FIG. 12 is a schematic diagram of an abnormal behavior detection apparatus according to an embodiment of the present application;
fig. 13 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In order to better understand the aspects of the present application, the present application will be described in further detail with reference to the accompanying drawings and detailed description. It will be apparent that the described embodiments are only some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The core of the application is to provide an abnormal behavior detection method which can be applied to scenes of monitoring abnormal behaviors of users, particularly the elderly, or scenes of guaranteeing health of users, particularly the elderly, and the like.
With the aggravation of the aging population problem, old people, especially elderly people living alone, may have various abnormal behaviors at home, such as falling, and if not found in time, severe injury is caused to the old people. In addition, abnormalities in the sequence of daily activities may suggest some chronic diseases, such as Alzheimer's disease, which if found in time can effectively reduce health risks.
After the monitoring data aiming at the target user are obtained, the monitoring data are preprocessed to obtain a data sequence, the data sequence is identified by utilizing a behavior identification model, the behavior sequence of the target user is determined, the data sequence and the behavior sequence are input into an anomaly detection model, whether the target user has the anomaly can be determined according to the output of the anomaly detection model, the anomaly of the user can be found in time through the joint detection of the behavior identification model and the anomaly detection model, the intervention is timely, the disease development is delayed, and the treatment cost is saved.
In order to facilitate understanding, an application scenario of the technical scheme of the present application is described below. Referring to fig. 1, a monitoring device such as a wearable device and a camera is deployed in a daily living environment of a target user, and the monitoring device can communicate with an application server to transmit monitoring data to the application server. The application server can obtain monitoring data aiming at a target user through the monitoring equipment, preprocesses the monitoring data, can obtain a data sequence, inputs the data sequence into a behavior recognition model, can recognize human behaviors, such as behavior 1, behavior 2, … …, behavior N and the like, can determine the behavior sequence of the target user according to the output of the behavior recognition model, then inputs the data sequence and the behavior sequence into an abnormality detection model, and can determine whether the target user has abnormal behaviors according to the output of the abnormality detection model.
It should be noted that the application server is described as a single independent server, but it is understood that in practical application, the application server may be replaced by an application server cluster or a distributed cluster formed by a plurality of application servers. Accordingly, in fig. 1, the application server may be replaced with an application platform composed of a plurality of application servers.
Referring to fig. 2, a flowchart of an implementation of a method for detecting abnormal behavior according to an embodiment of the present application may include the following steps:
s210: monitoring data for a target user is obtained.
In the embodiment of the application, the target user is any user to be subjected to abnormal behavior detection, and can be an elderly user or other group users. The target user may carry with him a non-invasive daily wearable device in which inertial sensors such as magnetometers, tri-axial accelerometers, gyroscopes, indoor positioning systems, etc. may be embedded. Image acquisition equipment such as a camera can be deployed in the living environment of the target user. These may be referred to as monitoring devices. The monitoring device can continuously monitor various states of the target user to obtain monitoring data.
Monitoring data for the target user can be obtained by the monitoring device.
S220: and preprocessing the monitoring data to obtain a data sequence.
It can be understood that the monitoring of the state of the target user by the monitoring device is a continuous process, corresponding monitoring data is continuously generated, and after the monitoring data for the target user is obtained, the monitoring data can be preprocessed to obtain a data sequence. The data sequence includes sets of data that are correlated in time.
S230: and inputting the data sequence into a behavior recognition model, and determining the behavior sequence of the target user according to the output of the behavior recognition model.
In the embodiment of the application, a behavior recognition model can be obtained through pre-training, the behavior recognition model is used for recognizing human behaviors, the input of the behavior recognition model can be a data sequence, and the output of the behavior recognition model can be the probability of each recognized behavior.
The method comprises the steps of obtaining monitoring data aiming at a target user, preprocessing the monitoring data, inputting the data sequence into a behavior recognition model after obtaining the data sequence, and obtaining the output of the behavior recognition model through the processing of the behavior recognition model, wherein the probability of each behavior corresponding to each group of data contained in the data sequence can be obtained.
From the output of the behavior recognition model, a behavior sequence of the target user, such as behavior 1, behavior 2, … …, behavior N, may be determined.
Optionally, according to the output of the behavior recognition model, the behavior with the highest probability corresponding to each group of data included in the data sequence can be obtained, and the sequence formed by the behaviors with the highest probability is determined as the behavior sequence of the target user.
Optionally, the probability of each behavior corresponding to each set of data included in the data sequence may be obtained according to the output of the behavior recognition model, for each behavior corresponding to a set of data, if the probability of the behavior is maximum and is greater than or equal to a set probability threshold, determining that the set of data actually corresponds to the behavior, and if the probability of the behavior is maximum but is less than the probability threshold, considering that the set of data does not actually correspond to any behavior, and in the determined behavior sequence of the target user, not including the behavior corresponding to the set of data.
S240: and inputting the data sequence and the behavior sequence into an anomaly detection model, and determining whether the target user has an anomaly behavior according to the output of the anomaly detection model.
In the embodiment of the application, an abnormality detection model may be trained in advance, where the abnormality detection model is used for identifying abnormal behaviors, the inputs of the abnormality detection model may be a data sequence and a behavior sequence (the behavior sequence includes a behavior tag), and the output of the abnormality detection model may be the probability of a detected normal behavior or an abnormal behavior or the determination of a detected normal behavior or an abnormal behavior. The anomaly detection model can detect whether the sequence of behaviors in the behavior sequence is normal or not, and can detect whether the specific behaviors in the behavior sequence are normal or not. For example, if there is a behavior in the behavior sequence that does not conform to the normal behavior order, the behavior may be considered an abnormal behavior, or if there is a behavior in the behavior sequence that does not conform to the normal behavior, the behavior may be considered an abnormal behavior.
After the behavior recognition model is utilized to determine the behavior sequence of the target user, the data sequence and the behavior sequence can be further input into the abnormality detection model, and the output of the abnormality detection model can be obtained through the processing of the abnormality detection model. Based on the output of the anomaly detection model, it can be determined whether the target user has an anomalous behavior.
The combination of a behavior recognition model and an anomaly detection model can be understood as a chain of classifiers, the behavior recognition model being a first classifier and the anomaly detection model being a second classifier. The identified behavior sequence and data sequence may be used as inputs to an anomaly detection model. The output of the behavior recognition model is added as an input to the anomaly detection model. The input space of the anomaly detection model is (x, y, z, label), wherein (x, y, z) is a data sequence corresponding to the monitoring data obtained from the monitoring device, namely the actual input feature, and is also the input space of the first classifier. "label" represents the behavior label, which is the output of the first classifier. For the second classifier, "label" is taken as input along with (x, y, z). That is, for a behavior recognition model, there are three instances of each input. For the anomaly detection model, there are four instances of each input. The advantage of using a chain of classifiers is that the correlation between the output labels of the different classifiers is preserved.
After the method provided by the embodiment of the application is applied, the monitoring data aiming at the target user is obtained, the monitoring data is preprocessed to obtain the data sequence, the behavior recognition model is utilized to recognize the data sequence, the behavior sequence of the target user is determined, the data sequence and the behavior sequence are input into the abnormality detection model, whether the target user has abnormal behaviors can be determined according to the output of the abnormality detection model, the abnormal behaviors of the user can be found in time through the combined detection of the behavior recognition model and the abnormality detection model, the intervention is performed in time, the disease development is helped to be delayed, and the treatment cost is saved.
In some embodiments of the application, the behavior recognition model and/or the anomaly detection model may include at least one of the following model structures:
A multi-head convolutional neural network-long-short-term memory network model structure;
attention mechanism-convolutional neural network-long-short-term memory network model structure.
In an embodiment of the present application, the behavior recognition model and/or the anomaly detection model may be a multi-head (MH) -convolutional neural network (Convolutional Neural Networks, CNN) -Long Short-Term Memory (LSTM) model structure.
Or the behavior recognition model and/or the anomaly detection model may be an Attention mechanism (Attention) -convolutional neural network-long-term memory network model structure.
Or the behavior recognition model and/or the abnormality detection model may be a combination of a multi-head-convolutional neural network-long-short-term memory network model structure and an attention mechanism-convolutional neural network-long-term memory network model structure.
The convolutional neural network is effective in automatically extracting and learning features from time series data, and as shown in fig. 3, after different inputs of the convolutional neural network generate feature graphs through convolution kernels, pooling operation is performed on the feature graphs, so that data quantity is reduced, and calculation complexity of a model is reduced.
Which model structure to use can be selected according to actual requirements. The use of the two model structures helps to improve the accuracy of the behavior recognition model and/or the anomaly detection model.
In some embodiments of the present application, in the multi-head-convolutional neural network-long-short-term memory network model structure, a plurality of convolutional neural networks are connected in parallel, each convolutional neural network receives an input sequence and processes the input sequence, outputs of the plurality of convolutional neural networks enter a full connector, and after being combined and interpreted by a full connection layer, the outputs are input into the long-term memory network and output a prediction result.
In an embodiment of the present application, the multi-head convolutional neural network-long-short term memory network model structure may comprise four parts: an input part, a convolutional neural network part, a long-term memory network part and an output part.
The convolutional neural network portion may include a plurality of convolutional neural networks connected in parallel, such as three convolutional neural networks connected in parallel. Each convolutional neural network receives and processes the input sequence, so that the multi-head architecture can read and understand the sequence data under different inputs, and output of a plurality of convolutional neural networks is obtained, and obvious behavior feature vectors are extracted. The output of the convolution neural networks enters a full-connection layer, the full-connection layer performs merging and interpretation and then inputs the merged output into a long-period memory network, namely the merged output is used as the input of a rear-end long-period memory network, and the long-period memory network outputs a prediction result.
It should be noted that, if the behavior recognition model includes a multi-head convolutional neural network-long-short-term memory network model structure, the input sequence is a data sequence, and the prediction result is the probability of each recognized behavior. If the anomaly detection model comprises a multi-head-convolution neural network-long-short-term memory network model structure, the input sequence is an index sequence and a behavior sequence, and the prediction result is the probability of the abnormal behavior or the judgment of the abnormal behavior.
The accuracy of behavior recognition or abnormal behavior detection can be improved through the multi-head-convolution neural network-long-short-term memory network model structure.
In some embodiments of the present application, in the structure of the attention mechanism-convolutional neural network-long-short-term memory network model, the convolutional neural network may include a convolutional layer, a pooling layer and a full-connection layer that are sequentially connected, the convolutional neural network receives an input sequence and performs feature extraction on the input sequence, the long-term memory network learns a data change rule according to features extracted by the convolutional neural network, weights of different features are not iterated through attention, weights are re-assigned to an output result of the long-term memory network, and a prediction result is output.
In the embodiment of the application, the attention mechanism-convolutional neural network-long-short-term memory network model structure comprises five parts: an input section, a convolutional neural network CNN section, a long short term memory network LSTM section, an attention section, and an output section, as shown in FIG. 4.
The convolutional neural network comprises a convolutional layer, a pooling layer and a full-connection layer which are sequentially connected. The convolutional neural network receives an input sequence and performs feature extraction on the input sequence. Specifically, features are extracted through a convolution layer, the input sequence is subjected to convolution operation through a one-dimensional convolution check, then the features are subjected to dimension reduction through a pooling layer, and finally one-dimensional structural transformation of the features is completed through a full-connection layer to obtain output features. The long-term memory network learns the data change rule according to the features extracted by the convolutional neural network in the last step, then iterates out weights of different features through the attention part, re-assigns weights to the output result of the long-term memory network, and finally outputs the prediction result. The long-term memory network part can comprise a plurality of LSTM units which are connected in sequence, the output of the LSTM units is used as the input of the attention part, the normalized exponential function of the attention part is a SoftMax function, and the output is attention weight alpha i.
It should be noted that, if the behavior recognition model includes a structure of an attention mechanism-convolutional neural network-long-short-term memory network model, the input sequence is a data sequence, and the prediction result is a probability of each recognized behavior. If the abnormality detection model comprises a structure of an attention mechanism-convolutional neural network-long-short-term memory network model, the input sequence is a data sequence and a behavior sequence, and the prediction result is the probability of abnormal behavior or the judgment of the abnormal behavior.
The accuracy of behavior recognition or abnormal behavior detection can be improved through the structure of the attention mechanism-convolutional neural network-long-short-term memory network model.
In some embodiments of the present application, preprocessing the monitoring data to obtain a data sequence may include the following steps:
step one: carrying out standardized processing on the monitoring data;
step two: and dividing the standardized monitoring data by utilizing a sliding window to obtain a data sequence.
For convenience of description, the above two steps are described in combination.
In the embodiment of the application, after the monitoring data aiming at the target user is obtained, proper preprocessing can be performed based on the adopted behavior recognition model and the abnormality detection model.
The monitoring data may be normalized, for example, by a center scaling method. This is because the monitoring data of the wearable device, camera, etc. monitoring devices have different scales or different types, so the central scaling method can be used to rescale the distribution of values. Here, the values of the monitored data are adapted to an appropriate mean (mean) and standard deviation (Standard Deviation, SD).
The standardized processing procedure for the monitoring data is as follows:
Wherein the average value is
Standard deviation of
X represents monitoring data before normalization;
y represents standardized monitoring data;
X i represents the ith monitored data;
n represents the total number of monitored data.
The monitoring data is continuously collected and recorded. After the monitoring data is standardized, the standardized monitoring data can be segmented by utilizing a sliding window to obtain data sequences, and the time relation among the monitoring data is maintained among the data sequences. As shown in fig. 5, the monitored data divided using the sliding window has N matrices, each matrix having a size of t×i, where T is the length of the time window, and I is the input instance, and can be understood as the data within the time window.
After the monitoring data aiming at the target user are obtained, the monitoring data are standardized, and then the standardized monitoring data are segmented by utilizing a sliding window, so that the obtained data sequence can be better suitable for being used as the input of a behavior recognition model and an abnormality detection model.
In some embodiments of the present application, obtaining monitoring data for a target user may include the steps of:
The first step: acquiring monitoring original data aiming at a target user;
and a second step of: and carrying out missing value filling processing and/or noise reduction processing on the monitoring original data to obtain the monitoring data aiming at the target user.
For convenience of description, the above two steps are described in combination.
In the embodiment of the application, the monitoring equipment such as the wearable equipment and the camera can transmit the acquired original data to the application server through wireless communication, so that the application server can acquire the monitoring original data aiming at the target user. Since some noise information is generally included in the monitoring original data, and data loss may occur in the communication process, the missing value filling process and/or the noise reduction process may be performed on the monitoring original data first, and the data after the missing value filling process and/or the noise reduction process is determined as the monitoring data for the target user.
Alternatively, the missing value filling process is performed on the monitored original data, and the missing data is represented by NaN (Not a Number) or "0", so that the missing data is inferred.
Optionally, the monitoring raw data obtained from the accelerometer, magnetometer, gyroscope, etc. sensors may be filtered to eliminate noise. Different noise data processing methods may be selected for different types of sensors. For example, for sensors such as accelerometers, an inertial filter may be used to filter the data and remove some of the noise, as follows:
Where y (k) is the kth output of the filter;
x (k) is the kth input of the filter;
a is a filter coefficient, a=t/(T f +t);
T is the time constant of the filtering link;
T f is the sampling period.
In general, the sampling period T f is much smaller than the time constant T, T and T f of the filtering segment, and can be chosen according to the particular situation, so long as the filtered signal does not produce significant ripple.
After the monitoring original data aiming at the target user is obtained, the missing value filling processing and/or the noise reduction processing are carried out on the monitoring original data, so that the obtained monitoring data is more complete and has better performance.
In some embodiments of the present application, in the case where it is determined that the target user has abnormal behavior, the method may further include the steps of:
and outputting alarm information, wherein the alarm information comprises abnormal behavior information.
Alternatively, the alarm information may be output through a short message, social software, telephone, alert tone, etc., and the recipient of the alarm information may be the target user, or may be an associated user of the target user, such as a family member, a medical staff, etc. of the target user. The alert information may include abnormal behavior information such as, in particular, which abnormal behavior, or in what circumstances, the abnormal behavior. Thus, the target user or the associated user of the target user can know the state of the target user in time and intervene in time.
In some embodiments of the present application, the behavior recognition model and the abnormality detection model may be obtained in advance by:
step one: obtaining a sample data set, wherein the sample data set comprises a plurality of groups of monitoring sample data, and each group of monitoring sample data corresponds to a user sample respectively;
step two: dividing the sample data set into a training data set and a test data set;
Step three: training an initial recognition model by using a training data set, and adjusting model parameters of the initial recognition model, wherein the initial recognition model is an initial model of a behavior recognition model;
Step four: training an initial detection model by utilizing the training data set and the output result of the initial recognition model, and adjusting model parameters of the initial detection model, wherein the initial detection model is an initial model of an abnormal detection model;
Step five: and testing the trained initial recognition model and the initial detection model by using a test data set, and determining the trained initial recognition model as a behavior recognition model and determining the trained initial detection model as an abnormal detection model under the condition that a test result meets a preset requirement.
For convenience of description, the above five steps are described in combination.
In the embodiment of the application, a sample data set can be obtained, wherein the sample data set comprises a plurality of groups of monitoring sample data, and each group of monitoring sample data corresponds to a user sample respectively. For example, there are ten user samples, each of which has corresponding monitoring sample data, for a total of ten sets of monitoring sample data.
The sample data included in the sample data set may be data obtained by performing a missing value filling process and/or a noise reduction process on the original data of the user sample, and performing a normalization process and a data division process.
After obtaining the sample data set, the sample data set may be divided into a training data set and a test data set, e.g., may be divided in a ratio of 8:2. The data in the training data set is only used for training, the data in the test data set is not used for training, the data in the test data set is used for testing the trained model, and the capability of the model is analyzed and evaluated according to the relevant evaluation indexes.
An initial model of the behavior recognition model, that is, an initial recognition model, may be previously constructed, the initial recognition model and the behavior recognition model having the same structure, and an initial model of the abnormality detection model, that is, an initial detection model, may be previously constructed, the initial detection model and the abnormality detection model having the same structure.
The initial recognition model can be trained by utilizing the training data set, model parameters of the initial recognition model can be adjusted in the training process, then the initial detection model is trained by utilizing the training data set and the output result of the initial recognition model, and the model parameters of the initial detection model are adjusted in the training process.
And under the condition that the training times reach a set time threshold or the model converges to a preset condition, stopping training to obtain the initial recognition model and the initial detection model after training.
And testing the trained initial recognition model and the initial detection model by using a test data set, and if the test result meets the preset requirement, determining the trained initial recognition model as a behavior recognition model and determining the trained initial detection model as an abnormal detection model. If the test result does not meet the preset requirement, the initial recognition model and the initial detection model can be continuously trained, and if the training data set can be added, the training is continued. The preset requirements may be accuracy requirements, precision requirements, recall requirements, etc.
One possible model acquisition procedure is shown in fig. 6, after the training dataset is obtained, the training dataset may be used to train a multi-headed model for human behavior recognition (Human Activity Recogition, HAR), and after the training is completed, a multi-headed model for HAR is obtained, forming the first flow. And (3) applying the multi-head model for the HAR to a training data set to obtain output of the multi-head model for the HAR to the training data set, training the multi-head model for abnormal behavior recognition (Abnormal Activity Recogition, AAR) by using the output and the training data set, and obtaining the multi-head model for the AAR after training is finished to form a second flow. This forms a classifier chain, each flow consisting of multiple head models, each model having its own inputs and outputs. These models improve the accuracy and generalization ability of the classifier by learning different features and patterns.
The behavior recognition model and the abnormality detection model can be obtained through the process, and the behavior recognition model and the abnormality detection model are applied to an actual scene to perform human behavior recognition and abnormal behavior detection.
The embodiment of the application utilizes the linkage of the behavior recognition model and the abnormality detection model to detect the abnormal behavior, can reduce the false alarm rate, improves the reliability, and is more suitable for monitoring the health and the behavior of the old. In addition, a remote behavior monitoring system employing non-invasive wearable devices, deep learning information technology, and broadcast notifications may provide an efficient, reliable, and low cost solution for assessing user daily behavior.
In order to facilitate understanding, the technical solutions provided by the embodiments of the present application are described again by specific examples.
1) Firstly, data acquisition and acquisition are carried out, and data preprocessing is carried out.
1.1 This example data is from 10 volunteers aged 60 years, 10 volunteers perform 9 daily actions and 3 different kinds of falling actions, respectively, the sensor collects at 20Hz rate, these data are generated by the accelerometer every 50ms, and the human behavior gesture dataset WISDM can be combined to obtain behaviors of about 24000 to 36000 sample users.
1.2 Preprocessing the sample data, and selecting a center scaling method to obtain standardized data after filling the missing values and reducing noise.
1.3 Sample dataset as 8:2 into training data sets and test data sets as shown in table 1:
TABLE 1
2) The present example mainly adopts a MH-CNN-LSTM model architecture, and Keras (an open source artificial neural network library) can be used to construct the MH architecture. Keras is an application programming interface (Application Programming Interface, API) of an efficient deep neural network architecture. The MH model is implemented as a supervision model. This work uses a cross entropy loss function that calculates the error between the predicted and true labels. The optimization is performed by adopting a gradient-based algorithm Adam, and the learning rate of the individual can be found out according to the parameters. In this example, a sliding window method is employed to segment the sensor data. The window size is 200, the step size is 20, and other parameters are shown in Table 2:
TABLE 2
And performing model training by using the training data set, and performing model test by using the test data set to obtain a behavior recognition model and an anomaly detection model.
The accuracy and loss values of the behavior recognition obtained by testing and comparing the conventional CNN-LSTM model with the MH-CNN-LSTM model after 15 iterations are shown in FIG. 7 and FIG. 8. For anomaly detection, using the MH-CNN-LSTM model and the conventional CNN-LSTM model, the results of the models were either normal or abnormal in detection behavior, and the F1-score (a weighted harmonic mean of model accuracy and recall) results are shown in FIGS. 9 and 10. From the above, it can be seen that the MH-CNN-LSTM model can show higher accuracy and precision in behavior recognition and anomaly detection.
As shown in fig. 11, the entire abnormal behavior detection process may include a raw data layer, an observed state layer, and an abnormal detection layer. The original monitoring data aiming at the target user can be acquired through the accelerometer, the gyroscope, the azimuth angle and other sensors, and the indoor position of the target user can be determined through the indoor positioning system, such as bedrooms, toilets, living rooms, restaurants, kitchens and the like.
And after the missing value filling processing and/or the noise reduction processing are carried out on the monitoring original data, the monitoring data aiming at the target user can be obtained. The monitoring data is preprocessed to obtain the data sequence. The MH-CNN-LSTM algorithm can be used for identifying the actions of the data sequence to obtain real-time actions.
The motion can be classified into a static motion, a dynamic motion, a falling motion, and the like. Static actions may include sitting, standing, lying, etc., dynamic actions may include walking, running, sitting down, standing up, going upstairs, going downstairs, etc., and falling actions may include forward falling, backward falling, sideways falling, etc.
The state can be observed according to the real-time actions and the action occurrence positions, and whether specific behavior anomalies exist is determined. If so, the guardian or healthcare worker is notified. If the specific behavior abnormality does not exist, a behavior sequence or a behavior activity chain can be formed, the abnormal behavior of the old is detected through an MH-CNN-LSTM algorithm, if the behavior sequence is abnormal, such as repeated eating, forgetting to eat, and the like, a guardian or a medical care personnel is notified, and the notification information can contain abnormality description information. This allows continuous monitoring of the daily activities of the elderly in the place of daily living.
For example, the abnormal behavior determined by the behavior sequence is shown in table 3:
TABLE 3 Table 3
In general, the embodiment of the application introduces a deep learning technology, integrates a plurality of sensors and the like to perform multi-level sequential abnormal behavior recognition, has higher accuracy than the traditional machine learning method, and provides a low-cost and high-efficiency solution for real-time monitoring of the behaviors of the daily old and abnormal behavior detection. The problem that the senile people can only find the Alzheimer's disease in a high-cost clinical environment is solved, the Alzheimer's disease patient is identified in advance, and the medical cost of the senile people is reduced; remote real-time monitoring and abnormal behavior identification of the old in daily life are realized; the method based on deep learning is provided for identifying the abnormal behaviors and the abnormal behavior sequence of the old; on the basis of the prior art, the abnormality of the behavior sequence is more focused, the abnormal behavior of the old is more accurately identified, and the accuracy of abnormality identification is improved.
Corresponding to the above method embodiment, the embodiment of the present application further provides an abnormal behavior detection device, where the abnormal behavior detection device described below and the abnormal behavior detection method described above may be referred to correspondingly to each other.
Referring to fig. 12, the abnormal behavior detection apparatus 1200 may include the following modules:
A first obtaining module 1210, configured to obtain monitoring data for a target user;
a second obtaining module 1220, configured to pre-process the monitoring data to obtain a data sequence;
A first determining module 1230, configured to input the data sequence into a behavior recognition model, and determine a behavior sequence of the target user according to an output of the behavior recognition model;
A second determining module 1240, configured to input the data sequence and the behavior sequence into the anomaly detection model, and determine whether the target user has an anomaly behavior according to the output of the anomaly detection model.
By applying the device provided by the embodiment of the application, after the monitoring data aiming at the target user are obtained, the monitoring data are preprocessed to obtain the data sequence, the behavior recognition model is utilized to recognize the data sequence, the behavior sequence of the target user is determined, the data sequence and the behavior sequence are input into the abnormality detection model, whether the target user has abnormal behaviors can be determined according to the output of the abnormality detection model, the abnormal behaviors of the user can be found in time through the combined detection of the behavior recognition model and the abnormality detection model, the intervention is performed in time, the disease development is helped to be delayed, and the treatment cost is saved.
In some embodiments of the application, the behavior recognition model and/or the anomaly detection model comprises at least one of the following model structures:
A multi-head convolutional neural network-long-short-term memory network model structure;
attention mechanism-convolutional neural network-long-short-term memory network model structure.
In some embodiments of the present application, in the multi-head-convolutional neural network-long-short-term memory network model structure, a plurality of convolutional neural networks are connected in parallel, each convolutional neural network receives an input sequence and processes the input sequence, outputs of the plurality of convolutional neural networks enter a full connection layer, and the full connection layer performs merging interpretation and then inputs the full connection layer into the long-term memory network, and outputs a prediction result.
In some embodiments of the present application, in the structure of the attention mechanism-convolutional neural network-long-short-term memory network model, the convolutional neural network includes a convolutional layer, a pooling layer and a full-connection layer which are sequentially connected, the convolutional neural network receives an input sequence and performs feature extraction on the input sequence, the long-term memory network learns a data change rule according to the features extracted by the convolutional neural network, weights of different features are iterated through an attention part, weights are re-assigned to an output result of the long-term memory network, and a prediction result is output.
In some embodiments of the present application, the second obtaining module 1220 is specifically configured to:
Carrying out standardized processing on the monitoring data;
and dividing the standardized monitoring data by utilizing a sliding window to obtain a data sequence.
In some embodiments of the present application, the first obtaining module 1210 is specifically configured to:
acquiring monitoring original data aiming at a target user;
and carrying out missing value filling processing and/or noise reduction processing on the monitoring original data to obtain the monitoring data aiming at the target user.
In some embodiments of the present application, the abnormal behavior detection apparatus 1200 further includes a third obtaining module for obtaining the behavior recognition model and the abnormal detection model in advance by:
Obtaining a sample data set, wherein the sample data set comprises a plurality of groups of monitoring sample data, and each group of monitoring sample data corresponds to a user sample respectively;
dividing the sample data set into a training data set and a test data set;
Training an initial recognition model by using a training data set, and adjusting model parameters of the initial recognition model, wherein the initial recognition model is an initial model of a behavior recognition model;
training an initial detection model by utilizing the training data set and the output result of the initial recognition model, and adjusting model parameters of the initial detection model, wherein the initial detection model is an initial model of an abnormal detection model;
And testing the trained initial recognition model and the initial detection model by using a test data set, and determining the trained initial recognition model as a behavior recognition model and determining the trained initial detection model as an abnormal detection model under the condition that a test result meets a preset requirement.
The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.
Corresponding to the above method embodiment, the embodiment of the present application further provides an electronic device, including:
a memory for storing a computer program;
And the processor is used for realizing the steps of the abnormal behavior detection method when executing the computer program.
As shown in fig. 13, which is a schematic diagram of a composition structure of an electronic device, the electronic device may include: a processor 10, a memory 11, a communication interface 12 and a communication bus 13. The processor 10, the memory 11 and the communication interface 12 all complete communication with each other through a communication bus 13.
In an embodiment of the present application, the processor 10 may be a central processing unit (Central Processing Unit, CPU), an asic, a dsp, a field programmable gate array, or other programmable logic device, etc.
The processor 10 may call a program stored in the memory 11, and in particular, the processor 10 may perform operations in an embodiment of the abnormal behavior detection method.
The memory 11 is used for storing one or more programs, and the programs may include program codes including computer operation instructions, and in the embodiment of the present application, at least the programs for implementing the following functions are stored in the memory 11:
Acquiring monitoring data for a target user;
preprocessing monitoring data to obtain a data sequence;
Inputting the data sequence into a behavior recognition model, and determining the behavior sequence of the target user according to the output of the behavior recognition model;
and inputting the data sequence and the behavior sequence into an anomaly detection model, and determining whether the target user has an anomaly behavior according to the output of the anomaly detection model.
In one possible implementation, the memory 11 may include a storage program area and a storage data area, where the storage program area may store an operating system, and at least one application program required for functions, etc.; the storage data area may store data created during use.
In addition, the memory 11 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device or other volatile solid-state storage device.
The communication interface 12 may be an interface of a communication module for interfacing with other devices or systems.
Of course, it should be noted that the structure shown in fig. 13 is not limited to the electronic device in the embodiment of the present application, and the electronic device may include more or less components than those shown in fig. 13 or may be combined with some components in practical applications.
Corresponding to the above method embodiments, the present application further provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps of the above abnormal behavior detection method.
In addition, it should be noted that: embodiments of the present application also provide a computer program product or computer program that may include computer instructions that may be stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor may execute the computer instructions, so that the computer device performs the description of the abnormal behavior detection method in the foregoing corresponding embodiment, and therefore, a detailed description will not be given here. In addition, the description of the beneficial effects of the same method is omitted. For technical details not disclosed in the computer program product or the computer program embodiments according to the present application, reference is made to the description of the method embodiments according to the present application.
In this specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, so that the same or similar parts between the embodiments are referred to each other.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of functionality in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may be disposed in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The principles and embodiments of the present application have been described herein with reference to specific examples, but the description of the examples above is only for aiding in understanding the technical solution of the present application and its core ideas. It should be noted that it will be apparent to those skilled in the art that various modifications and adaptations of the application can be made without departing from the principles of the application and these modifications and adaptations are intended to be within the scope of the application as defined in the following claims.

Claims (10)

1. An abnormal behavior detection method, comprising:
Acquiring monitoring data for a target user;
Preprocessing the monitoring data to obtain a data sequence;
inputting the data sequence into a behavior recognition model, and determining the behavior sequence of the target user according to the output of the behavior recognition model;
and inputting the data sequence and the behavior sequence into an abnormality detection model, and determining whether the target user has abnormal behaviors according to the output of the abnormality detection model.
2. The method according to claim 1, wherein the behavior recognition model and/or the anomaly detection model comprises at least one of the following model structures:
A multi-head convolutional neural network-long-short-term memory network model structure;
attention mechanism-convolutional neural network-long-short-term memory network model structure.
3. The method according to claim 2, wherein in the multi-head-convolutional neural network-long-short-term memory network model structure, a plurality of convolutional neural networks are connected in parallel, each convolutional neural network receives an input sequence and processes the input sequence, outputs of the plurality of convolutional neural networks enter a full connection layer, and the full connection layer performs merging interpretation and then inputs the merged result into the long-term memory network to output a prediction result.
4. The method according to claim 2, wherein in the attention mechanism-convolutional neural network-long-short-term memory network model structure, the convolutional neural network comprises a convolutional layer, a pooling layer and a full-connection layer which are sequentially connected, the convolutional neural network receives an input sequence and performs feature extraction on the input sequence, the long-term memory network learns a data change rule according to features extracted by the convolutional neural network, weights of different features are iterated through an attention part, weights are re-assigned to an output result of the long-term memory network, and a prediction result is output.
5. The method of claim 1, wherein preprocessing the monitoring data to obtain a data sequence comprises:
carrying out standardized processing on the monitoring data;
and dividing the standardized monitoring data by utilizing a sliding window to obtain a data sequence.
6. The method of claim 1, wherein the obtaining monitoring data for the target user comprises:
acquiring monitoring original data aiming at a target user;
and carrying out missing value filling processing and/or noise reduction processing on the monitoring original data to obtain the monitoring data aiming at the target user.
7. The method according to any one of claims 1 to 6, characterized in that the behavior recognition model and the abnormality detection model are obtained in advance by:
obtaining a sample data set, wherein the sample data set comprises a plurality of groups of monitoring sample data, and each group of monitoring sample data corresponds to a user sample respectively;
Dividing the sample dataset into a training dataset and a test dataset;
Training an initial recognition model by using the training data set, and adjusting model parameters of the initial recognition model, wherein the initial recognition model is an initial model of the behavior recognition model;
Training an initial detection model by utilizing the training data set and the output result of the initial recognition model, and adjusting model parameters of the initial detection model, wherein the initial detection model is an initial model of the abnormal detection model;
And testing the trained initial recognition model and the trained initial detection model by using the test data set, and determining the trained initial recognition model as the behavior recognition model and determining the trained initial detection model as the abnormal detection model under the condition that a test result meets a preset requirement.
8. An abnormal behavior detection apparatus, comprising:
the first acquisition module is used for acquiring monitoring data aiming at a target user;
the second acquisition module is used for preprocessing the monitoring data to acquire a data sequence;
The first determining module is used for inputting the data sequence into a behavior recognition model and determining the behavior sequence of the target user according to the output of the behavior recognition model;
And the second determining module is used for inputting the data sequence and the behavior sequence into an abnormality detection model, and determining whether the target user has abnormal behaviors or not according to the output of the abnormality detection model.
9. An electronic device, comprising:
a memory for storing a computer program;
A processor for implementing the steps of the abnormal behavior detection method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the abnormal behavior detection method according to any one of claims 1 to 7.
CN202410139881.9A 2024-01-31 2024-01-31 Abnormal behavior detection method, device, electronic equipment and storage medium Pending CN118038548A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410139881.9A CN118038548A (en) 2024-01-31 2024-01-31 Abnormal behavior detection method, device, electronic equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410139881.9A CN118038548A (en) 2024-01-31 2024-01-31 Abnormal behavior detection method, device, electronic equipment and storage medium

Publications (1)

Publication Number Publication Date
CN118038548A true CN118038548A (en) 2024-05-14

Family

ID=90994430

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410139881.9A Pending CN118038548A (en) 2024-01-31 2024-01-31 Abnormal behavior detection method, device, electronic equipment and storage medium

Country Status (1)

Country Link
CN (1) CN118038548A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118589695A (en) * 2024-08-06 2024-09-03 创意信息技术股份有限公司 Integrated intelligent analysis monitoring system suitable for intelligent power distribution room

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118589695A (en) * 2024-08-06 2024-09-03 创意信息技术股份有限公司 Integrated intelligent analysis monitoring system suitable for intelligent power distribution room

Similar Documents

Publication Publication Date Title
EP3367883B1 (en) Monitoring activities of daily living of a person
US20200337580A1 (en) Time series data learning and analysis method using artificial intelligence
US10332638B2 (en) Methods and systems for pre-symptomatic detection of exposure to an agent
JP7070255B2 (en) Abnormality discrimination program, abnormality discrimination method and abnormality discrimination device
KR102154763B1 (en) System and method for predicting the development of dementia using smart devices
CN117133464B (en) Intelligent monitoring system and monitoring method for health of old people
CN111248859A (en) Automatic sleep apnea detection method based on convolutional neural network
CN118038548A (en) Abnormal behavior detection method, device, electronic equipment and storage medium
CN115662631A (en) AI intelligence discrimination-based nursing home management system
Kuveskar et al. A hybrid iot-enabled transfer learning model for accurate diagnosis of coronary heart disease
CN114343585B (en) Cognitive and behavioral disorder early warning method, device, equipment and storage medium
CN109119155B (en) ICU death risk assessment system based on deep learning
CN111317458A (en) Blood pressure detection system based on deep learning
CN117224080B (en) Human body data monitoring method and device for big data
Orpa et al. An IoT based healthcare solution with ESP32 using machine learning model
JP2024513618A (en) Methods and systems for personalized prediction of infections and sepsis
Panagiotou et al. A multi: modal decision making system for an ambient assisted living environment
CN117116475A (en) Method, system, terminal and storage medium for predicting risk of ischemic cerebral apoplexy
KR102499912B1 (en) A recurrence prediction system based on deep learning for prostate cancer using time series data of examination
CN111466877B (en) LSTM network-based oxygen reduction state prediction method
CN113208609A (en) Electrocardio information management system
CN114224348A (en) Combined neural network electrocardiogram classification method and system
CN113724853A (en) Intelligent medical system based on deep learning
CN117379009B (en) Early warning and monitoring system for critical cardiovascular and cerebrovascular diseases of old people in community
Javed et al. Cognitive Health Assessment of Decentralized Smart home Activities using Federated Learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination