Human body rhythm analysis method, human body rhythm adjustment method and human body rhythm analysis system based on behavior recognition

Abstract

The invention discloses a human body rhythm analysis method, an adjustment method and an analysis system based on behavior recognition, and the analysis method comprises the following steps: collecting a moving image of a to-be-analyzed person to form an image sequence, and recognizing the image sequence to obtain behavior record data; processing the behavior record data to obtain a behavior curve; and based on a neural network model for predicting the activity-rest state transition point, inputting the behavior curve into the neural network model for calculation to obtain the activity-rest state transition point and a transition point time sequence curve. The method comprises the following steps: acquiring an activity image based on a visual system, performing behavior recognition on the activity image to obtain behavior record data, converting the behavior record data into a behavior curve capable of deep learning, and predicting the behavior curve by adopting a neural network model through optimizing a data processing algorithm to obtain the human body rhythm of a to-be-analyzed person. The efficiency of processing a large amount of behavior record data and calculating the activity-rest state transition point is ensured, and the method has good accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of human rhythm analysis, and in particular to a human rhythm analysis method, a regulation method and an analysis system based on behavior recognition. Background Art

[0002] Studies have revealed that it is core genes and proteins, such as CLOCK and BMAL1, that control the human circadian rhythm. These genes and proteins regulate the biological clock through a transcription-translation negative feedback loop, among which light is considered to be the strongest external synchronization factor, which transmits retinal signals to the suprachiasmatic nucleus, affecting the biological clock. By changing the intensity and time of light, light therapy can help subjects reset their biological clocks, control human rhythms, and improve the quality of sleep and daily activities.

[0003] In order to regulate human rhythm, it is necessary to record human rhythm first. There are many methods currently, such as using wearable device data (such as wrist action recorders and photometers) to predict the feasibility and accuracy of circadian rhythm misalignment of night shift workers. The accuracy of the human rhythm prediction model of this method is heavily dependent on the data quality of wearable devices. Due to the influence of multiple factors such as wearing methods, device calibration, and usage habits, the data may have errors, affecting the accuracy of model predictions, and the errors will be greater when predicting circadian rhythms over a longer period of time. The existing literature discusses methods for integrating wearable device data into circadian rhythm models, exploring the integration of data from wearable devices (such as activity trackers, smart watches, etc.) into circadian rhythm models to better understand and predict changes in the human body clock. However, the data quality and consistency of wearable devices may be affected by multiple factors, such as device wearing conditions, technical failures, etc., which may affect the accuracy and reliability of the data. In addition, there are significant physiological and behavioral differences between individuals. The general model may perform poorly when adapting to different individuals, requiring a lot of personalized adjustments.

[0004] Existing methods for analyzing human rhythms also use fast Fourier transform (FFT) to identify dominant periodicity in data and characterize daily rhythms. However, the accuracy of the PML (perfectly matched layer) algorithm is heavily dependent on the quality and consistency of wearable device data, which may be affected by the wearing conditions and technical issues of the device, and the complexity of the PML algorithm may limit its application in real-time analysis or resource-constrained environments. There are also analysis methods that want to use machine learning to integrate circadian rhythm models to better understand and predict changes in the human body clock. In existing literature, different machine learning algorithms and statistical models have been developed and applied using long-term continuous data collected by wearable devices to analyze and interpret these data and form a detailed understanding of personal circadian rhythms. However, how to efficiently process and analyze these massive amounts of data remains a challenge. The model of the article is not sufficient to establish long-term dependence and cannot complete long-term predictions.

[0005] From the above, we can see that the existing human rhythm analysis methods have significant limitations. It is difficult to identify and record human behavior in real time, and it is impossible to make personalized adjustments based on individual behavior data. This highlights the urgency and importance of developing a method that can accurately record and analyze human rhythms in real time without relying on wearable devices. Summary of the invention

[0006] The purpose of the present invention is to overcome the above-mentioned deficiencies in the prior art and to provide a human rhythm analysis method, a regulation method and an analysis system based on behavior recognition.

[0007] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions: The human rhythm analysis method based on behavior recognition includes the following steps: Collect activity images of the person to be analyzed to form an image sequence, and identify the image sequence to obtain behavior record data; Processing the behavior record data to obtain a behavior curve; Based on the neural network model for predicting the activity-rest state transition point, the behavior curve is input into the neural network model for calculation to obtain the activity-rest state transition point and the transition point timing curve.

[0008] In the technical solution of the present invention, the state of the person to be analyzed is collected as an active image, and the active images at different times are collected to form an image sequence, and then the person to be analyzed is identified as active or resting according to the image sequence, and the behavior record data at different time points are formed, and then the behavior record data at different time points are processed to obtain a behavior curve, which is used as the data basis for subsequent neural network model calculation, and a neural network model suitable for predicting the activity-rest state transition point is adopted, and the behavior curve is input into the neural network model for calculation, and the activity-rest state transition point and the transition point timing curve are predicted to record the human body rhythm of the person to be analyzed. The method of the present invention collects images based on a visual system, recognizes behaviors based on images into behavior record data, and converts the behavior record data into a behavior curve that can be deeply learned and calculated, and optimizes the data processing algorithm, and then uses the neural network model to predict and obtain the human body rhythm of the person to be analyzed, ensuring the efficiency when processing a large amount of behavior record data and calculating the activity-rest state transition point, with good accuracy, and solving the problem that the prior art relies on wearable devices and cannot recognize and record human behavior in real time.

[0009] As a preferred solution of the present invention, the specific method of identifying the image sequence to obtain the behavior record data is: S11, collecting activity images of the person to be analyzed in real time to obtain an image sequence; S12, assigning an identity code ID to the person to be analyzed according to the image sequence; S13. For the person to be analyzed assigned to the ID, use the logistic regression model to determine whether the person to be analyzed is in a resting state or an active state at different time points; S14. When it is determined that the person to be analyzed is in an active state, the activity behavior is identified and classified to obtain behavior record data.

[0010] As a more preferred embodiment of the present invention, a specific method for assigning an identity code ID to a person to be analyzed according to an image sequence is as follows: performing feature extraction on the image sequence using a convolutional network, generating candidate regions through a region proposal network using the extracted feature map, performing RoI alignment on the candidate regions, and generating bounding boxes and category predictions through a fully connected layer, generating a segmentation mask through full convolution, extracting human features from the segmentation mask, and comparing them with the identity library to assign an identity code ID to the person to be analyzed.

[0011] As a more preferred embodiment of the present invention, in step S13, a logistic regression model is used to determine whether the person to be analyzed is in a resting state or an active state, and the decision rule is: In the formula Indicates the activity status. represents the resting state, and τ is the decision threshold.

[0012] As a more preferred embodiment of the present invention, activity behaviors include standing, sitting, and lying down. The activity behaviors are identified and classified based on a behavior recognition model based on visual-text joint embedding. The specific method is: based on the image sequence of the person to be analyzed, it is converted into a visual feature vector through a visual encoder, the preset different activity behavior texts are text-encoded, and activity behavior classification decisions are made based on the visual feature vectors and the different activity behavior text encodings to obtain the activity behavior category of the person to be analyzed.

[0013] As a preferred embodiment of the present invention, the specific method for processing behavior record data to obtain a behavior curve is: extracting and preprocessing the behavior record data, the preprocessing includes moving average processing and smoothing processing, and then extracting behavior features from the smoothed data, and drawing a behavior curve based on the behavior features.

[0014] In the above technical solution, the behavior record data is converted into a behavior curve through extraction and preprocessing. The converted behavior curve is convenient for decomposition, and different behavioral characteristics are more clearly identified, providing a data basis for predicting the activity-rest state transition point and the transition point timing curve.

[0015] As a preferred embodiment of the present invention, the behavior characteristics include activity intensity, and the activity intensity is the ratio of the frequency and duration of the activity.

[0016] As a preferred embodiment of the present invention, the neural network model includes a decomposition layer, a CTCL Block module and a KP Block module. The decomposition layer is used to decompose the behavior curve into a Trend part and a Season part. The CTCL Block module is used to process the Trend part in the behavior curve to achieve the prediction of the activity-rest state transition point. The KP Block module is used to fuse the Trend part and the Season part to achieve the prediction of the transition point timing curve.

[0017] As a more preferred embodiment of the present invention, the CTCL Block module includes at least one KANLinear layer and a linear projection layer. The KANLinear layer can extract high-dimensional features of the input behavior curve to capture the complex nonlinear relationship in the data, which helps to accurately identify the transition points of the behavior. Its output will pass through a linear projection layer to map the features to a low-dimensional space for predicting the activity-rest state transition points; the KANLinear layer includes a KAN layer and a linear layer, which is used to make up for the limitations of the linear layer in processing complex time series data.

[0018] As a more preferred embodiment of the present invention, the KP Block module includes at least two KANLinear layers, a CAT layer and an LSTM module, the KANLinear layer includes a KAN layer and a linear layer, the CAT layer is used for splicing, and the LSTM module is used as a gating mechanism.

[0019] As a preferred embodiment of the present invention, the activity-rest state transition point includes the rest-activity demarcation time point CT0 and the activity-rest demarcation time point CT12, and the transition point timing curve includes the CT0 timing curve and the CT12 timing curve.

[0020] Another aspect of the present invention provides a human rhythm regulation method based on behavior recognition, the regulation method comprising: using the above-mentioned human rhythm analysis method based on behavior recognition to predict the activity-rest state transition point and the transition point timing curve of the person to be analyzed, setting the standard activity-rest state transition point, comparing the activity-rest state transition point with the standard activity-rest state transition point, and designing a lighting control scheme for human rhythm regulation of the person to be analyzed based on the comparison result, so that the activity-rest state transition point reaches the standard activity-rest state transition point.

[0021] In the above technical solution, human rhythm is predicted in real time based on behavior recognition, without relying on any wearable device. According to the changes in the rhythm of the person to be analyzed, the lighting control solution can be designed and output to achieve real-time detection and real-time regulation, eliminating the inconvenience of the person to be analyzed having to wear the device continuously, providing a more natural and interference-free user experience, and being able to automatically adjust according to the individual's rhythm changes, thereby improving the regulation efficiency and quality of life.

[0022] The present invention also provides a human rhythm analysis system based on behavior recognition, the system comprising: The recognition module is used to collect the activity images of the person to be analyzed to form an image sequence, and recognize the image sequence to obtain behavior record data; A processing module, used for processing the behavior record data to obtain a behavior curve; The prediction module is used to input the behavior curve into the neural network model for calculation based on the neural network model predicting the activity-rest state transition point, and obtain the activity-rest state transition point and the transition point timing curve.

[0023] The present invention also provides an electronic device, comprising at least one processor and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned human rhythm analysis method based on behavior recognition.

[0024] The present invention also provides a computer-readable storage medium, which stores at least one instruction, at least one program, a code set or an instruction set. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the above-mentioned human rhythm analysis method based on behavior recognition.

[0025] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention provides a human rhythm analysis method based on behavior recognition, which collects images based on a visual system, performs behavior recognition based on the images into behavior record data, and converts the behavior record data into a behavior curve that can be deeply learned and calculated. By optimizing the data processing algorithm, the human rhythm of the person to be analyzed is obtained after prediction using a neural network model, ensuring efficiency in processing a large amount of behavior record data and calculating the activity-rest state transition point, with good accuracy.

[0026] 2. The neural network model in the analysis method of the present invention includes a decomposition layer, a CTCL Block module and a KP Block module. The CTCL Block module is used to process the Trend part in the behavior curve to predict the transition point of the activity-rest state. The KPBlock module is used to fuse the Trend part and the Season part to predict the transition point time series curve. The CTCL Block module includes at least two KANLinear layers and a linear projection layer. The KANLinear layer can extract the high-dimensional features of the input behavior curve, capture the complex nonlinear relationship in the data, and help to accurately identify the transition point of the behavior. The KPBlock module includes at least two KANLinear layers, a CAT layer and an LSTM module. The LSTM module is used as a gating mechanism to capture complex behavior patterns to predict the time series. The KP Block module generates a more accurate time series prediction result by expert weighted selection. The model can capture complex time series patterns, and through multi-scale processing, it can process multi-level features, thereby improving the prediction accuracy of the model.

[0027] 3. The present invention provides a human rhythm regulation method based on behavior recognition, which predicts human rhythm in real time based on behavior recognition, does not rely on any wearable device, avoids interference with daily activities, and can perform personalized lighting control according to changes in the rhythm of the person to be analyzed, so as to achieve real-time detection and regulation, and provide a more natural and interference-free user experience, thereby improving regulation efficiency and quality of life.

[0028] 4. The present invention is suitable for analyzing the human rhythm of personnel in confined spaces (such as space stations). It analyzes the human rhythm based on visual behavior recognition and adjusts the lighting control according to the human rhythm. It uses a pure visual system to accurately adjust the physiological rhythm of the human body in a confined space, thereby improving its health and comfort. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow chart of the human rhythm analysis method based on behavior recognition of the present invention; Figure 2 Behavior curves drawn according to activity intensity; Figure 3 It is a behavior curve composed of Season part and Trend part; Figure 4 The behavior curve is decomposed into Season and Trend curves; Figure 5 A structural diagram of the network structure of the neural network model for predicting the activity-rest state transition point of the present invention; Figure 6 This is the structural diagram of the KANLinear layer. DETAILED DESCRIPTION

[0030] In order to more clearly describe the invention purpose, technical scheme and technical effect advantages in the specific implementation case of the present invention, the scheme in the specific embodiment will be described in detail in combination with the drawings of the specification of the present invention. The specific technical scheme involved in the following specific embodiments is only for a clear and complete description of the innovative technical scheme of the present invention. It is only a part of the specific implementation scheme that can be adopted by the present invention, not all embodiments, and should not be understood as a limitation on the innovative scheme of the present invention. Any scheme adopting the same inventive concept of the present invention should be included in the protection scope of the present invention.

[0031] Secondly, the description of the drawings in the specific embodiments of the present invention is only for the convenience of technical personnel to understand the solution of the present invention. The partial details in the drawings are for the convenience of clearly presenting the technical solution. It should not be considered that all technical features in the drawings must be included in the specific implementation cases, and the detailed features in the drawings should not be identified as additional limitations on the innovative technical solution of the present invention. The components in the various embodiments described and shown in the drawings can be combined and arranged according to different configurations, and these changes in combination and arrangement should be identified as part of the entire embodiment of the innovative solution of the present invention and included in the scope of protection of the present invention.

[0032] In summary, the schemes or descriptions presented in the specific embodiments and drawings of the present invention are not intended to limit the scope of protection claimed, but are merely selected embodiments / cases to help technicians understand the relevant innovative solutions. Based on these embodiments, all other equivalent or parallel embodiments obtained by those skilled in the art without creative work are within the scope of protection claimed by the present invention.

[0033] Example 1 This embodiment provides a human rhythm analysis method based on behavior recognition, such as Figure 1 As shown, the following steps are included: S1. Collect activity images of the person to be analyzed to form an image sequence, and identify the image sequence to obtain behavior record data; The specific method for identifying the image sequence to obtain the behavior record data is: S11, collect the activity images of the person to be analyzed in real time through the image acquisition mechanism to obtain an image sequence ; S12, assigning an identity code ID to the person to be analyzed according to the image sequence; S13. For the person to be analyzed assigned to the ID, use the logistic regression model to determine whether the person to be analyzed is in a resting state or an active state at different time points; S14. When it is determined that the person to be analyzed is in an active state, the activity behavior is identified and classified to obtain behavior record data.

[0034] In step S12, the specific method of assigning an identity code ID to the person to be analyzed according to the image sequence is as follows: extracting features from the image sequence using a convolutional network (CNN), generating candidate regions through a region proposal network (RPN) using the extracted feature map, aligning the candidate regions with RoIs (Regions of Interest), generating bounding boxes and category predictions through a fully connected layer, and generating segmentation masks through a fully convolutional network (FCN). , extract human features from the segmentation mask, compare them with the identity database, and assign them to the identity code ID of the person to be analyzed.

[0035] The formula for generating bounding box and category prediction using the fully connected layer is: in is the boundary prediction at time t, is the category prediction at time t, RoIAlign represents the region alignment operation, Indicates the generation of candidate regions at time t, is the extracted feature map.

[0036] The formula for generating the segmentation mask is: in is the segmentation mask, and FCN stands for fully convolutional operation.

[0037] Extracting human features from segmentation masks refers to extracting identity information from segmentation masks. The formula used is: in represents the activation function, usually ReLu, is the linear layer weight, is the linear layer bias term.

[0038] The extracted identity information is compared with the identity information in the pre-established identity database, and an ID is assigned to the qualified personnel to be analyzed. Each person to be analyzed can be assigned a unique ID, and the formula used is: in represents the unique ID assigned at time t, It refers to the value of the independent variable (argument) that makes a function achieve the maximum value in a specific set or interval. cos() represents cosine similarity. F j The feature vector representing the jth identity in the identity library.

[0039] In step S13, a logistic regression model is used to determine whether the person to be analyzed is in a resting state or an active state, and the decision rule is: In the formula Indicates the activity status. represents the rest state, τ is the decision threshold, The calculation formula is: Where θ, θ0 are the weights and biases of logistic regression, T represents transposition, is the change in the Euclidean distance between the bounding boxes of two consecutive frames, The calculation formula is: In order to extract features from image sequences using a convolutional network (CNN), the extracted feature map is passed through a region proposal network (RPN) to generate candidate regions, the candidate regions are aligned with RoIs, and a bounding box is generated through a fully connected layer.

[0040] In this embodiment, τ is set to 0.5, and the output of the logistic regression model is a probability, which indicates that when the input features are given, the probability of the person to be analyzed being in a resting state or an active state is obtained, and the result analysis is used to determine whether the person to be analyzed is in a resting state or an active state at different time points.

[0041] In step S14, the activity behaviors include standing, sitting, and laying. When it is determined that the person to be analyzed is in an activity state, it is necessary to determine which specific activity behavior is.

[0042] This embodiment is based on a behavior recognition model jointly embedded with vision and text to classify the activity behaviors of the person to be analyzed. The specific method is as follows: based on the image sequence of the person to be analyzed, it is converted into a visual feature vector through a visual encoder, the preset different activity behavior texts are text-encoded, and the activity behavior classification decision is made based on the visual feature vector and the different activity behavior text encodings to obtain the activity behavior category of the person to be analyzed.

[0043] Specifically, the image sequence of the person to be analyzed is used , converted into a visual feature vector through a visual encoder: Classify the preset different activity behavior texts into "standing", "sitting" and "laying" and encode them: After calculating the cosine similarity, the activity behavior classification decision is made: Select the text feature vector T with the highest similarity, that is, the highest confidence k The corresponding activity behavior category is taken as the recognition result.

[0044] S2. Processing the behavior record data to obtain a behavior curve; specifically: extracting and preprocessing the behavior record data, the preprocessing includes moving average processing and smoothing processing, and then extracting behavior features from the smoothed data, the behavior features include activity intensity, and drawing a behavior curve based on the behavior features.

[0045] Assume that the dataset of behavior record data is ,in t i Indicates the timestamp, b i Indicates at time t i The behavior category is rest or activity, and the activity behavior is specifically classified. Take appropriate time intervals for sampling: Perform moving average processing on the sampled data to generate a smooth curve for observation: Where k is the moving window size.

[0046] Extract behavioral features from the smoothed data, such as activity intensity. Activity intensity is the ratio of the frequency and duration of an activity, that is, the activity intensity is the number of behavioral changes per unit time. The calculation formula is as follows: The following are obtained according to the intensity of the activity: Figure 2 The behavior curve shown in the figure is composed of the Season part, the Trend part and random noise. Figure 3 As shown in the figure, the Season part reflects the periodic fluctuations of behavior, which is usually related to the natural rhythm of the human body (such as the circadian rhythm). The periodic peaks and troughs reflect the activity and rest state of the human body at different time periods in a day. This part is usually regular and the fluctuation range is relatively stable. The Trend part shows the long-term change trend in the behavior curve. This part can reflect the overall direction of the behavior pattern, such as the phenomenon that the activity gradually decreases or increases over time during the day. The trend part is usually smoother.

[0047] In order to better process these data, the data are broken down as follows Figure 4 As shown in the figure, since the Season part reflects periodic fluctuations, it has stable and repeatable characteristics, and periodic data is relatively predictable within a certain range because these fluctuations occur repeatedly. Therefore, using the seasonal part as input for time series forecasting can effectively capture future periodic fluctuations. The Trend part is a smooth long-term change in the behavior curve, which can more clearly show the overall change direction and important turning points of the behavior. Therefore, using the rate of change of the trend part (i.e., the derivative of the curve) to locate the CT0 point can accurately identify the boundary between the resting and active states of the behavior.

[0048] Fusion of periodicity and trend, as shown in the figure, since the Season part reflects periodic fluctuations, it has stable and repeatable characteristics, while the Trend part shows the long-term trend of behavior, the combination of the two can more comprehensively predict future behavior. Therefore, modeling the Season part and the Trend part separately and fusing their outputs can capture the combined impact of periodicity and long-term trends, thereby improving the accuracy of time series prediction. Therefore, the Season part and the Trend part are processed separately to more clearly identify different behavioral characteristics (periodicity and trend), and they are used to predict the transition point. This method allows the model to better handle complex behavioral data and obtain more accurate results in practical applications.

[0049] S3. Based on the neural network model for predicting the activity-rest state transition point, the behavior curve is input into the neural network model for calculation to obtain the activity-rest state transition point and the transition point timing curve.

[0050] The neural network model includes a decomposition layer, a CTCL Block module and a KP Block module. The decomposition layer is used to decompose the input behavior curve into a Trend part and a Season part. The CTCL Block module is used to process the Trend part in the behavior curve to realize the prediction of the activity-rest state transition point. The KP Block module is used to fuse the Trend part and the Season part to realize the prediction of the transition point timing curve.

[0051] The CTCL Block module includes at least one KANLinear (KAL) layer and a linear projection layer. The KANLinear layer can extract high-dimensional features of the behavior curve, capture complex nonlinear relationships in the data, and help accurately identify the transition points of the behavior. Its output will pass through a linear projection layer to map the features to a low-dimensional space for predicting CT0 or CT12.

[0052] In this embodiment, the KANLinear layer is a new structure. The KANLinear layer includes the KAN layer (Kolmogorov-Arnold Networks) and the linear layer, which is used to make up for the limitations of the linear layer when processing complex time series data. Figure 6 As shown, the final output of KANLinear is the weighted sum of the linear part and the B-spline interpolation part: Where α is the corresponding weight of the linear part, β is the corresponding weight of the spline part, Output the result for the linear part, Output results for the B-spline interpolation part.

[0053] The calculation of the base linear mapping in the KANLinear layer is: In the formula is the activation function, is the linear weight, X is the input, is the linear bias term.

[0054] The calculation of the B-spline interpolation part in the KANLinear layer is: In the formula is the spline function, is the weight corresponding to the spline function. The B-spline basis function calculates the interpolation of the input in different intervals through the control point grid and the input x: where k is the spline order, is a B-spline basis function whose calculation result is based on the input and the mesh.

[0055] Assume that the input data in the CTCL Block module is represented as ,in Represents the data at each input time point. Feature extraction is implemented by at least one KANLinear layer. Each layer can be represented as a nonlinear mapping. Features are extracted from input data. Let the input feature of the i-th layer be The output features are ,as follows: in is the linear transformation part, This is the B-spline interpolation part. It should be noted that , yes and The output of is used to capture nonlinear relationships. The KAN layer can be represented as a nonlinear transformation of each layer until the required features are finally extracted: After extracting features through at least one KANLinear layer, the features are mapped to a low-dimensional space through the output of the linear projection layer. The calculation formula is: Where W, b are the weights and biases of the linear projection layer. The essence of detecting the CT0 point is a regression task, so the output of the linear projection layer can be used to directly predict the index or time value of the activity-rest state transition point in the behavior curve, thereby obtaining the CT0 point or CT12 point.

[0056] The KP Block module includes at least two KANLinear layers, a CAT layer, and an LSTM module. As mentioned above, the KANLinear layer includes a KAN layer and a linear layer. After at least two KANLinear layers, the CAT layer is used to splice the results together, and the LSTM module is used as a gating mechanism. In some embodiments, a linear projection layer is provided after the CAT layer, such as Figure 5 As shown in the KP Block part of Figure 1, the KP Block module integrates the Trend part and the Season part, and combines multiple network components by using LSTM gates, KANLinearexpert and linear projection layers to capture complex behavior patterns to achieve time series prediction.

[0057] This example innovatively uses LSTM as a gating mechanism. LSTM is very effective in processing time series data because it can capture long-term dependencies in the data. The LSTM here takes trend and seasonal features as input and generates gated outputs based on dependencies on time steps. gate_out , to adjust the final prediction results: in Represent the encoding vector and the hidden state of LSTM respectively, and through softmax, the gated output is converted into a selection weight: The score determines which network has the greatest impact on the final prediction. The weighted combination of KANLinear expert is: Final prediction: The linear layer here maps the predicted multi-dimensional features to the time series forecast output.

[0058] The prediction mechanism of the KP Block module combines LSTM gating, KANLinear and linear projection layers to generate more accurate time series prediction results through expert weighted selection. The model can capture complex time series patterns and process multi-level features through multi-scale processing, thereby improving prediction accuracy.

[0059] The activity-rest state transition points include the rest-activity demarcation time point CT0 and the activity-rest demarcation time point CT12, and the transition point timing curves include the CT0 timing curve and the CT12 timing curve. In this embodiment, the behavior curve neural network model is predicted to output the CT0 and CT0 timing curves of the person to be analyzed based on the real-time image sequence.

[0060] Example 2 The present embodiment provides a human rhythm regulation method based on behavior recognition, the regulation method comprising: using the human rhythm analysis method based on behavior recognition of Example 1, predicting the activity-rest state transition point and the transition point timing curve of the person to be analyzed, setting the standard activity-rest state transition point, comparing the activity-rest state transition point with the standard activity-rest state transition point, and designing a lighting control scheme for human rhythm regulation of the person to be analyzed based on the comparison result, so that the activity-rest state transition point reaches the standard activity-rest state transition point.

[0061] By detecting and analyzing the behavior data of the person to be analyzed, the CT0 point of the real-time image sequence is predicted and compared with the standard CT0 point. According to different results (early, delayed, same), corresponding lighting control solutions are adopted to optimize the work and rest environment of the person to be analyzed and improve the quality of life. Specifically, a standard CT0 point needs to be preset first, that is, the CT0 point of the person to be analyzed needs to be , using the CT0 obtained by analysis and For comparison: according to According to the result, take corresponding lighting control plan.

[0062] when , that is, the CT0 point of the person to be analyzed and When they are the same or very similar, it means that the user's biological rhythm is normal and the current work and rest state needs to be maintained. From CT0 to CT12, a white light with a cosine change from 600 lumens to 2000 lumens is executed, and the brightness of the white light is gradually increased to help the user stay awake. From CT12 to CT24, a white light with a cosine change from 600 lumens to 0 lumens is executed, and the brightness of the white light is gradually reduced to help the user gradually fall asleep and maintain a normal biological clock. In the present invention, the time difference between CT0 and CT12 is specified to be 12h.

[0063] when , that is, when the CT0 point of the person to be analyzed is advanced, it means that the user's biological rhythm is early and needs to be gradually delayed through light adjustment to achieve . One hour before CT0, a cosine-changing blue light from 800 lumens to 1200 lumens is used to gradually wake up the user and delay their biological clock by increasing the brightness of the blue light. From CT0 to CT12, a cosine-changing white light from 600 lumens to 2000 lumens is used to simulate the changes in natural daylight and help users stay awake. From CT12 to CT23 (CT12 plus 11h), a cosine-changing red light from 600 lumens to 0 lumens is used to reduce the suppression of melatonin and help users gradually fall asleep.

[0064] when , that is, when the CT0 point of the person to be analyzed is delayed, it means that the user's biological rhythm is late, and it is necessary to gradually advance it through light adjustment to achieve . From CT0 to CT12, cosine-changing white light from 600 lumens to 2000 lumens is used to gradually wake up the user by increasing the brightness of the white light, so that the biological clock is gradually advanced. From CT12 to CT13, cosine-changing yellow light from 600 lumens to 1500 lumens is used to simulate the gradual warming of daylight and reduce the impact on the awake state. From CT13 to CT15, cosine-changing white light from 400 lumens to 0 lumens is used, and from CT15 to CT24, cosine-changing white light from 60 lumens to 0 lumens is used to simulate the gradual weakening of natural daylight and help users gradually fall asleep.

[0065] Through the behavior recognition technology based on computer vision system, the behavior curve of the person to be analyzed is adaptively recorded and analyzed, and the activity-rest state transition point (CT0, CT12) is calculated; at the same time, the recursive feasibility and stability of rhythm regulation are established, so as to optimize the lighting control strategy and achieve precise regulation of the biological rhythm of the person to be analyzed. This method greatly reduces manual intervention, improves regulation efficiency, saves energy resources, ensures control performance, and reduces the risk of rhythm disorder. It also proves the recursive feasibility and system stability under the constraints of the activity state of the person to be analyzed and the input of lighting control. Numerical simulation and experiments prove the effectiveness and advancement of this method.

[0066] Example 3 This embodiment provides a human rhythm analysis system based on behavior recognition, the system comprising: The recognition module is used to collect the activity images of the person to be analyzed to form an image sequence, and recognize the image sequence to obtain behavior record data; A processing module, used for processing the behavior record data to obtain a behavior curve; The prediction module is used to input the behavior curve into the neural network model for calculation based on the neural network model predicting the activity-rest state transition point, and obtain the activity-rest state transition point and the transition point timing curve.

[0067] The systems or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described in various modules according to their functions. Of course, when implementing this application, the functions of each module can be implemented in the same or multiple software and / or hardware, or the module that implements the same function can be implemented by a combination of multiple sub-modules.

[0068] This embodiment also provides an electronic device, including at least one processor, a memory connected to the at least one processor, and at least one input / output interface connected to the at least one processor; the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the human rhythm analysis method based on behavior recognition of the aforementioned embodiment. The input / output interface may include a display, a keyboard, a mouse, and a USB interface for inputting and outputting data.

[0069] The electronic device may be an electronic device for a client, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the human rhythm analysis method based on behavior recognition of Example 1.

[0070] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM), disks or optical disks, etc. Various media that can store program codes.

[0071] When the above-mentioned integrated unit of the present invention is implemented in the form of a software functional unit and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0072] This embodiment also provides a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set. The at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the above-mentioned human rhythm analysis method based on behavior recognition.

[0073] Computer readable storage media can be tangible devices that hold and store instructions used by instruction execution devices. Computer readable storage media can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any combination thereof.

[0074] For those skilled in the art, when understanding the solutions described in the specific embodiments of the present invention, they can refer to the conventional technical manuals in the field. At the same time, for the places where the above-mentioned terms appear, they can make appropriate understandings or adjustments for reference, and deduce the implementation of the same or similar technical solutions without paying any creative work.

[0075] The above embodiments describe only the basic principles, main features and / or advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and the invention content of the specification only describe the principles or specific cases of the present invention. Without departing from the essence of the innovative idea of ​​the present invention, the innovative scheme of the present invention may be subject to various changes and improvements, and these changes and improvements all fall within the scope of protection required by the present invention.

Claims

1. A human rhythm analysis method based on behavior recognition, characterized in that: The following steps are involved: Collect activity images of the person to be analyzed to form an image sequence, and identify the image sequence to obtain behavior record data; Processing the behavior record data to obtain a behavior curve; Based on the neural network model for predicting the activity-rest state transition point, the behavior curve is input into the neural network model for calculation to obtain the activity-rest state transition point and the transition point timing curve.

2. The human body rhythm analysis method based on behavior recognition according to claim 1, characterized in that: The specific method for identifying the image sequence to obtain the behavior record data is: S11, collecting activity images of the person to be analyzed in real time to obtain an image sequence; S12, assigning an identity code ID to the person to be analyzed according to the image sequence; S13. Using a logistic regression model, determine whether the person to be analyzed is in a resting state or an active state at different time points; S14. When it is determined that the person to be analyzed is in an active state, the activity behavior is identified and classified to obtain behavior record data.

3. The human rhythm analysis method based on behavior recognition according to claim 1, characterized in that: The specific method of processing the behavior record data to obtain the behavior curve is: extracting and preprocessing the behavior record data, the preprocessing includes moving average processing and smoothing processing, then extracting the behavior features from the smoothed data, and drawing the behavior curve according to the behavior features.

4. The human rhythm analysis method based on behavior recognition according to claim 3 is characterized in that: Behavioral characteristics include activity intensity, which is the ratio of activity frequency to duration.

5. The human rhythm analysis method based on behavior recognition according to any one of claims 1 to 4, characterized in that: The neural network model includes a decomposition layer, a CTCL Block module and a KP Block module. The decomposition layer is used to decompose the behavior curve into a Trend part and a Season part. The CTCL Block module is used to process the Trend part in the behavior curve to realize the prediction of the activity-rest state transition point. The KP Block module is used to fuse the Trend part and the Season part to realize the prediction of the transition point timing curve.

6. The human body rhythm analysis method based on behavior recognition according to claim 5, characterized in that: The CTCLBlock module includes at least one KANLinear layer and a linear projection layer. The KANLinear layer can extract high-dimensional features of the behavior curve; The KANLinear layer includes a KAN layer and a linear layer.

7. The human body rhythm analysis method based on behavior recognition according to claim 5, characterized in that: The KP Block module includes at least two KANLinear layers, a CAT layer and an LSTM module. The KANLinear layer includes a KAN layer and a linear layer. The CAT layer is used for splicing, and the LSTM module is used as a gating mechanism.

8. A human rhythm regulation method based on behavior recognition, characterized in that: The adjustment method comprises: Using the human rhythm analysis method based on behavior recognition as described in any one of claims 1 to 7, the activity-rest state transition point and the transition point timing curve of the person to be analyzed are predicted; Set standard activity-rest state transition points; The activity-rest state transition point is compared with the standard activity-rest state transition point. Based on the comparison results, a lighting control scheme for regulating the human body rhythm of the person to be analyzed is designed so that the activity-rest state transition point reaches the standard activity-rest state transition point.

9. The human rhythm analysis system based on behavior recognition is characterized by: The system comprises: The recognition module is used to collect the activity images of the person to be analyzed to form an image sequence, and recognize the image sequence to obtain behavior record data; A processing module, used for processing the behavior record data to obtain a behavior curve; The prediction module is used to input the behavior curve into the neural network model for calculation based on the neural network model predicting the activity-rest state transition point, and obtain the activity-rest state transition point and the transition point timing curve.

10. An electronic device comprising at least one processor and a memory in communication with the at least one processor; the memory stores instructions executed by the at least one processor, characterized in that: The instruction is executed by the at least one processor so that the at least one processor can execute the human rhythm analysis method based on behavior recognition as described in any one of claims 1-7.

Images