A method and system for monitoring infant behavior based on machine learning

Abstract

A method and system for monitoring infant behavior based on machine learning, including a data acquisition and storage module, a machine learning module and a detection and early warning module. Aiming at the problem of monitoring infant behavior, the present invention uses a multi-dimensional sensor system to collect infant behavior information and uses a machine learning algorithm to analyze the infant's behavior.

Description

Technical Field

[0001] The present invention relates to the field of housekeeping, and in particular to a method and system for monitoring infant behavior based on machine learning. Background Art

[0002] In modern society, caring for and monitoring infants and young children has become an important part of family life. Traditional infant monitoring methods mainly rely on manual observation. However, this method has the problem that the guardian may neglect or fail to monitor around the clock. With the development of science and technology, the infant behavior monitoring method and system based on machine learning has gradually become a solution that has attracted much attention because it can provide all-weather, multi-modal monitoring and early warning functions. The shortcomings of traditional monitoring methods have prompted the demand for more advanced technologies. The infant behavior monitoring system based on machine learning can use sensor technology and advanced data processing algorithms to collect and analyze multi-modal data of infants and young children, such as video, sound and motion trajectory. The comprehensive analysis of these data enables the system to more accurately understand the behavior of infants and young children and detect abnormal or dangerous behaviors in time. The background technology of the system includes a data acquisition and storage module, a machine learning module and a detection and early warning module. The data acquisition and storage module records and saves the multi-modal data of infants and young children by deploying sensors. The machine learning module is responsible for training the model, extracting features from the data and predicting the behavior of infants and young children. The detection and early warning module monitors the status of infants and young children in real time, triggers the early warning mechanism when an abnormality is found, and provides immediate feedback to the guardian through a user-friendly interface. Compared with traditional monitoring methods, the infant behavior monitoring system based on machine learning has many technical advantages. It can not only monitor the behavior of infants and young children around the clock and in all directions, but also continuously optimize the model through machine learning to adapt to the changes in the growth process of infants and young children. At the same time, the system takes into account user-friendliness, personalized predictions, and data privacy and security, providing guardians with a smarter, more convenient and reliable infant monitoring solution. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for monitoring infant behavior and movements based on machine learning to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above purpose, a method and system for monitoring infant behavior based on machine learning are provided, including a data acquisition and storage module, a machine learning module and a detection and early warning module; the specific process is described as follows:

[0005] S1. Deploy sensors in the living environment of infants and young children to obtain multimodal infant behavior information, clean the collected data, and extract multimodal data features;

[0006] S2. Create a labeled dataset of infant behavior based on data features, annotate the categories of different behavioral actions, and use the labeled dataset as the input training set of the algorithm;

[0007] S3. Based on the improved support vector machine machine learning algorithm, an intelligent classification model for infant behavior is constructed to classify different infant behaviors;

[0008] S4. Conduct a comprehensive assessment of the information on the current infant’s behavior, predict the possibility of dangerous behavior, and provide early warning and forecast information to the guardian.

[0009] Furthermore, in S1, sensors are deployed in the living environment of the infant, and the sensors include cameras and sound monitoring sensors.

[0010] Furthermore, in S1, the collected data is cleaned and multimodal data features are extracted. The multimodal data features include posture information and sound spectrum features in the video frames.

[0011] Furthermore, in S2, based on the data features, a label data set of infant behaviors is created, and the categories of different behavioral actions are marked. The data set uses a MySQL database built based on C language. The data annotation categories are divided into dangerous and non-dangerous. The data feature matrix As shown below:

[0012]

[0013] Among them, P represents the baby's posture information, Represents baby sound information, Represents the data label type, Represents dangerous behavior. Represents non-dangerous behavior. represents the baby posture information data and label dataset, Represents the Infant Sounds Information and Labels Dataset.

[0014] Furthermore, in S3, based on the improved support vector machine machine learning algorithm, an intelligent classification model for infant behavior is constructed to classify different infant behaviors. The detailed process is as follows:

[0015] (1) The labeled dataset is used as the training dataset for the infant behavior intelligent classification model, and the posture information and sound spectrum features in the video frames obtained by the camera and sound monitoring sensor are used as input data;

[0016] (2) Define the loss function , the formula is as follows:

[0017]

[0018] in, is the weight vector, represents the transpose of the weight vector, b is the bias, , used to select a good decision boundary and prevent overfitting, It is a logarithmic regularization parameter used to increase the data range of the loss function calculation. is the tolerance parameter, is the input feature vector of the ith sample, is the observed value of the i-th sample;

[0019] (3) Based on the idea of ​​ensemble learning, the feature selection method used in random forest is used to define the decision boundary g(x) of the support vector machine.

[0020]

[0021] N is the number of support vectors, is the corresponding coefficient of the support vector, is the class label of the support vector, is the kernel function used to calculate the similarity between samples, b is the bias term, and then the decision function f(x) of the support vector machine with the composite kernel function is introduced.

[0022]

[0023]

[0024]

[0025]

[0026] in, is the composite kernel function, is the sex kernel function, is the radial basis kernel function, is the weight of the linear kernel function, is the weight of the radial basis kernel function, σ is the bandwidth parameter of the kernel function, which controls the width of the kernel function and affects the similarity between samples;

[0027] (4) Use gradient descent to minimize the loss function to obtain the optimized weights and biases, and then put the parameters into the decision function again to obtain the final infant behavior recognition classification output.

[0028] Furthermore, the information of the reaction to the current infant behavior in S4 is used to propose an infant behavior assessment model, to comprehensively assess the infant's current behavior, and to predict the possibility of dangerous behavior. The detailed process is as follows:

[0029] (1) Based on the classification of infants’ behavior labels into dangerous and non-dangerous, it is defined as a binary probability model with a range of (0,1). Inspired by the Sigmund function, a probability formula is constructed:

[0030]

[0031]

[0032] in, is the probability that the baby's current behavior is dangerous, is the regularization parameter, which controls the weight of the regularization term, which helps prevent the model from overfitting; is the L2 regularization term, which penalizes the parameters of the model to make the coefficients of the model as small as possible, thereby reducing the risk of overfitting; is the coefficient of the model, indicating the degree of influence of each feature on the target variable; is the probability that the infant's current behavior is non-dangerous;

[0033] (2) Evaluate and optimize the probability model based on the recall rate R to make the baby's behavior prediction label more accurate. When the accuracy is less than 90%, it is necessary to continuously iterate the calculation parameters to improve the accuracy:

[0034]

[0035] Among them, TP is the number of positive examples correctly classified as positive examples by the model, and FN is the number of positive examples incorrectly classified as negative examples by the model;

[0036] (3) Iterative calculation of the comprehensive classification evaluation of infant behavior.

[0037] Furthermore, the early warning forecast in S4 feeds back information to the guardian, and the detailed process is as follows:

[0038] Set the baby's posture and sound warning thresholds, where the posture threshold refers to the baby's facial occlusion, especially the mouth and nose; the sound threshold refers to the range of the sound spectrum of the baby's crying; based on the analysis and recognition results of the baby's behavior, based on the baby's behavior data collected in real time by the camera and sound monitoring sensor, when the model determines that the infant's behavior is abnormal, the warning system is triggered, and a message or alarm is sent to the guardian to remind them to take corresponding measures.

[0039] Beneficial effects of the present invention: The present invention describes a method and system for monitoring infant behavior based on machine learning, including a data acquisition and storage module, a machine learning module and a detection and early warning module. Through the data acquisition and storage module, the system can simultaneously acquire and store multiple types of data, including video and audio, providing a rich source of information for the machine learning model. The machine learning module is based on an improved support vector machine machine learning algorithm to construct an intelligent classification model for infant behavior, classify different behaviors of infants, comprehensively analyze and identify multimodal infant behavior data, perform feature extraction and model training, propose an infant behavior assessment model, comprehensively evaluate the current behavior of infants, more comprehensively understand and predict the behavior of infants, and predict the possibility of dangerous behaviors. The detection and early warning module can provide real-time behavior information to guardians, interact with users through mobile phone applications and other means, and promptly convey the status of infants to help guardians better pay attention to and care for infants. The system supports continuous model optimization and updating, and can continuously improve the performance and scope of application of the model based on user feedback and new technological advances. The system provides an efficient infant monitoring solution for infant guardians, highlights the role of data science in infant health, and emphasizes the innovation and practicality of machine learning algorithms in analyzing infant sleep. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The invention is further illustrated by the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0041] Figure 1 It is a schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0042] The present invention is further described in conjunction with the following examples.

[0043] See also Figure 1 The present invention aims to provide a method and system for monitoring infant behavior and movements based on machine learning to solve the problems raised in the above-mentioned background technology.

[0044] To achieve the above purpose, a method and system for monitoring infant behavior based on machine learning are provided, including a data acquisition and storage module, a machine learning module and a detection and early warning module; the specific process is described as follows:

[0045] S1. Deploy sensors in the living environment of infants and young children to obtain multimodal infant and young children's behavior information, clean the collected data, and extract multimodal data features. The sensors include cameras and sound monitoring sensors. The multimodal data features include posture information in video frames and sound spectrum features.

[0046] S2. Based on the data features, create a label dataset of infant behavior, annotate the categories of different behavioral actions, and use the labeled dataset as the input training set of the algorithm. The dataset uses a MySQL database built based on C language. The data annotation categories are divided into dangerous and non-dangerous. The data feature matrix As shown below:

[0047]

[0048] Among them, P represents the baby's posture information, Represents baby sound information, Represents the data label type, Represents dangerous behavior. Represents non-dangerous behavior. represents the baby posture information data and label dataset, Represents the Infant Sounds Information and Labels Dataset.

[0049] S3. Based on the improved support vector machine machine learning algorithm, an intelligent classification model for infant behavior is constructed to adaptively classify different infant behaviors. The detailed process is as follows:

[0050] (1) The labeled dataset is used as the training dataset for the infant behavior intelligent classification model, and the posture information and sound spectrum features in the video frames obtained by the camera and sound monitoring sensor are used as input data;

[0051] (2) Define the loss function , the formula is as follows:

[0052]

[0053] in, is the weight vector, represents the transpose of the weight vector, b is the bias, , used to select a good decision boundary and prevent overfitting, It is a logarithmic regularization parameter used to increase the data range of the loss function calculation. is the tolerance parameter, is the input feature vector of the ith sample, is the observed value of the i-th sample;

[0054] (3) Based on the idea of ​​ensemble learning, the feature selection method used in random forest is used to define the decision boundary g(x) of the support vector machine.

[0055]

[0056] N is the number of support vectors, is the corresponding coefficient of the support vector, is the class label of the support vector, is the kernel function used to calculate the similarity between samples, b is the bias term, and then the decision function f(x) of the support vector machine with the composite kernel function is introduced.

[0057]

[0058]

[0059]

[0060]

[0061] in, is the composite kernel function, is the sex kernel function, is the radial basis kernel function, is the weight of the linear kernel function, is the weight of the radial basis kernel function, σ is the bandwidth parameter of the kernel function, which controls the width of the kernel function and affects the similarity between samples;

[0062] (4) Use gradient descent to minimize the loss function to obtain the optimized weights and biases, and then put the parameters into the decision function again to obtain the final infant behavior recognition classification output.

[0063] S4. Based on the information about the current infant behavior, an infant behavior assessment model is proposed to comprehensively assess the infant's current behavior and predict the possibility of dangerous behavior. The detailed process is as follows:

[0064] (1) Based on the classification of infants’ behavior labels into dangerous and non-dangerous, it is defined as a binary probability model with a range of (0,1). Inspired by the Sigmund function, a probability formula is constructed:

[0065]

[0066]

[0067] in, is the probability that the baby's current behavior is dangerous, is the regularization parameter, which controls the weight of the regularization term, which helps prevent the model from overfitting; is the L2 regularization term, which penalizes the parameters of the model to make the coefficients of the model as small as possible, thereby reducing the risk of overfitting; is the coefficient of the model, indicating the degree of influence of each feature on the target variable; is the probability that the infant’s current behavior is non-dangerous;

[0068] (2) Evaluate and optimize the probability model based on the recall rate R to make the baby's behavior prediction label more accurate. When the accuracy is less than 90%, it is necessary to continuously iterate the calculation parameters to improve the accuracy:

[0069]

[0070] Among them, TP is the number of positive examples correctly classified as positive examples by the model, and FN is the number of positive examples incorrectly classified as negative examples by the model;

[0071] (3) Iterative calculation of the comprehensive classification evaluation of infant behavior.

[0072] (4) Setting the posture and sound warning thresholds for infants, where the posture threshold refers to the coverage of the infant's face, especially the mouth and nose; the sound threshold refers to the range of the sound spectrum of the infant's crying; based on the analysis and recognition results of the infant's behavior and the infant behavior data collected in real time by the camera and sound monitoring sensor, when the model determines that the infant's behavior is abnormal, the warning system is triggered and a message or alarm is sent to the guardian to remind them to take appropriate measures.

[0073] In a specific embodiment, the user first obtains infant behavior information through sensors deployed in the infant's living environment, performs data cleaning and multimodal feature extraction on the behavior information, creates a label data set for the infant's behavior, adaptively classifies different infant behaviors through an infant behavior intelligent classification model, intelligently optimizes the infant behavior intelligent classification model, and determines the infant's behavior category. Afterwards, a comprehensive assessment of the infant's current behavior is performed through an infant behavior assessment model to predict the possibility of dangerous behavior. When the infant's behavior is judged to be abnormal, the early warning system is triggered to send information or an alarm to the guardian.

[0074] The present invention also provides a computer-readable storage medium, in which at least one instruction is stored, and the instruction is loaded and executed by a processor to implement the above method. The computer-readable storage medium may be a ROM, a random access memory, a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc. The instructions stored therein may be loaded by a processor in a terminal to execute the above method.

[0075] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0076] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0077] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A machine learning-based infant behavior monitoring system, characterized in that: It includes data collection and storage module, machine learning module and detection and early warning module; the specific process is described as follows: S1. Deploy sensors in the living environment of infants and young children to obtain multimodal infant behavior information, clean the collected data, and extract multimodal data features; S2. Based on the data features, create a labeled dataset of infant behavior, annotate the categories of different behavioral actions, and use the labeled dataset as the input training set of the algorithm; S3. Based on the improved support vector machine machine learning algorithm, an intelligent classification model for infant behavior is constructed to classify different infant behaviors: (1) The labeled dataset is used as the training dataset for the infant behavior intelligent classification model, and the posture information and sound spectrum features in the video frames obtained by the camera and sound monitoring sensor are used as input data; (2) Define the loss function Q(w,b), the formula is as follows: Among them, w is the weight vector, w T represents the transpose of the weight vector, b is the bias, ||·||2 represents the 2-norm, lnC is the logarithmic-based regularization parameter, ε is the tolerance parameter, and x i is the input feature vector of the i-th sample, y i is the observed value of the i-th sample; (3) Based on the idea of ​​ensemble learning, the feature selection method used in random forest is used to define the decision boundary g(x) of the support vector machine. sign() is the sign function used to determine the positive or negative value of the input value, N is the number of support vectors, α i is the corresponding coefficient of the support vector, y' i is the class label of the support vector, K(x i ,x) is the kernel function used to calculate the similarity between samples, b is the bias, and then the decision function f(x) of the support vector machine with the composite kernel function is introduced. in, is the composite kernel function, is a linear kernel function, is the radial basis kernel function, α is the weight of the linear kernel function, β is the weight of the radial basis kernel function, and σ is the bandwidth parameter of the kernel function, which controls the width of the kernel function and affects the similarity between samples; (4) Using gradient descent to minimize the loss function, we obtain the optimized weights and biases, and then put the parameters into the decision function again to obtain the final infant behavior recognition classification output; S4. Conduct a comprehensive assessment of the information on the current infant’s behavior, predict the possibility of dangerous behavior, and provide early warning and forecast information to the guardian.

2. The infant behavior monitoring system based on machine learning according to claim 1, characterized in that: In S1, sensors are deployed in the living environment of infants and young children, and the sensors include cameras and sound monitoring sensors.

3. The infant behavior monitoring system based on machine learning according to claim 1, characterized in that: In the S1, the collected data is cleaned and multimodal data features are extracted. The multimodal data features include posture information and sound spectrum features in the video frame.

4. The infant behavior monitoring system based on machine learning according to claim 1, characterized in that: In the S2, based on the data features, a label data set of infant behavior is created, and the categories of different behavioral actions are annotated. The data set uses a MySQL database built based on C language, and the data annotated categories are divided into dangerous and non-dangerous.

5. The infant behavior monitoring system based on machine learning according to claim 1, characterized in that: The information in S4 responds to the current infant behavior, proposes an infant behavior assessment model, conducts a comprehensive assessment of the infant's current behavior, and predicts the possibility of dangerous behavior. The detailed process is as follows: (1) Based on the classification of infants’ behavior labels into dangerous and non-dangerous ones, it is defined as a binary probability model with a range of (0,1). Inspired by the Sigmund function, a probability formula is constructed: P(δ=1)=1-P(δ=-1) Among them, x i represents the input feature vector, P(δ=-1) is the probability that the baby's current behavior is dangerous, and γ is the regularization parameter, which controls the weight of the regularization term. is the L2 regularization term, which penalizes the parameters of the model and reduces the coefficients of the model; μ i is the coefficient of the model, indicating the degree of influence of each feature on the target variable; P(δ=1) is the probability that the infant’s current behavior is non-dangerous; (2) Evaluate and optimize the probability model based on the recall rate R to make the baby's behavior prediction label more accurate. When the accuracy is less than 90%, it is necessary to continuously iterate the calculation parameters to improve the accuracy: Among them, TP is the number of positive examples correctly classified as positive examples by the model, and FN is the number of positive examples incorrectly classified as negative examples by the model; (3) Iterative calculation to evaluate the comprehensive classification of infant behavior.

6. The infant behavior monitoring system based on machine learning according to claim 1, characterized in that: The early warning forecast in S4 feeds back information to the guardian, and the detailed process is as follows: Set the baby's posture and sound warning thresholds, where the posture threshold refers to the baby's face occlusion; the sound threshold refers to the range of the sound spectrum of the baby's crying, and the monitoring range is [100Hz, 4000Hz]; based on the analysis and recognition results of the baby's behavior, based on the baby's behavior data collected in real time by the camera and sound monitoring sensor, when the model determines that the infant's behavior is abnormal, the warning system is triggered to send information or alarms to the guardians to remind them to take corresponding measures.

Images