Pet health monitoring method and system based on real-time fusion of multi-source heterogeneous data

Abstract

The pet health monitoring method and system based on multi-source heterogeneous data real-time fusion of the application belong to the technical field of pet health monitoring, and comprise the following steps: S1, acquiring and processing multi-source heterogeneous data of a wearable device arranged on a pet to generate normalized physiological data, motion data and position data; S2, calculating a basic stress level based on the normalized physiological data; S3, fusing the motion data and the position data to construct a context activity factor; S4, performing spectrum analysis based on the motion data to derive a behavior complexity index; S5, according to whether the basic stress level is higher than a preset stress activation threshold, dynamically compensating the basic stress level or directly taking the basic stress level as an output to calculate a final stress level; and S6, generating a monitoring result reflecting a pet health state based on the final stress level, so that the application solves the pain point that the prior art cannot distinguish activity occurrence scenes and endows the system with scene perception capability.

Description

Technical Field

[0001] This invention relates to the field of pet health monitoring technology, specifically to a pet health monitoring method and system based on real-time fusion of multi-source heterogeneous data. Background Technology

[0002] Current pet health monitoring devices typically rely on physiological indicators such as heart rate and body temperature to assess a pet's stress level. When pets engage in high-intensity play activities such as running and jumping, their physiological indicators naturally and temporarily spike, which can be numerically very similar to the physiological changes caused by adverse stresses such as fear, anxiety, or illness. Traditional monitoring methods lack the ability to perceive the context of the situation, making it difficult to distinguish between benign and mild stress. Therefore, they often misinterpret normal vigorous activity as a negative health alert, leading to inaccurate health assessments. Frequent false alarms also reduce user trust in the system. How to accurately identify the root cause of the spike in physiological indicators is a problem that current technology urgently needs to solve.

[0003] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for pet health monitoring based on real-time fusion of multi-source heterogeneous data, so as to solve the problems mentioned in the background art.

[0005] The technical solution of the present invention includes the following steps:

[0006] S1. Acquire and process multi-source heterogeneous data from wearable devices deployed on pets to generate normalized physiological data, motion data, and location data;

[0007] S2. Calculate the baseline stress level based on normalized physiological data;

[0008] S3. Integrate motion data and location data to construct contextual activity factors;

[0009] S4. Based on motion data, perform spectrum analysis to derive the behavioral complexity index;

[0010] S5. Based on whether the baseline stress level is higher than the preset stress activation threshold, the final stress level is calculated by dynamically compensating for the baseline stress level or directly using it as the output.

[0011] S6. Based on the final stress level, generate monitoring results reflecting the pet's health status.

[0012] Preferably, S1 specifically includes:

[0013] The system collects heart rate, body temperature, three-axis acceleration, and GPS coordinate data from pets using wearable devices.

[0014] Heart rate and body temperature data are normalized to generate normalized physiological data.

[0015] Preferably, S2 specifically includes:

[0016] A weighted linear fusion model was used to calculate the baseline stress level by weighted summation of normalized heart rate and normalized body temperature. The weight parameters of the model were empirical parameters set based on the sensitivity of different physiological indicators to stress response.

[0017] Preferably, S3 specifically includes:

[0018] Environmental context parameters are determined using geofencing technology; a motion intensity index is calculated based on motion data; and the environmental context parameters are multiplied by the motion intensity index to obtain the context activity factor.

[0019] Preferably, the calculation of the exercise intensity index specifically includes:

[0020] A high-pass filter is applied to the triaxial acceleration data to remove the static gravity component and obtain the dynamic acceleration component; the vector magnitude of the dynamic acceleration component is calculated to obtain the dynamic acceleration magnitude; the dynamic acceleration magnitude is normalized relative to a preset reference acceleration threshold to obtain the motion intensity index.

[0021] Preferably, S4 specifically includes:

[0022] Fast Fourier Transform is applied to the time series of dynamic acceleration modulus obtained from the previous steps, and the power spectral density is calculated. The power spectral density is converted into a probability distribution to obtain the power proportion of each frequency bin. Based on the Shannon entropy formula, the normalized entropy of the power proportion is calculated to obtain the behavioral complexity index.

[0023] Preferably, S5 specifically includes:

[0024] When the baseline stress level is higher than the stress activation threshold, the product of the compensation scaling factor, the context activity factor, and the behavioral complexity index is used as the compensation amount. The compensation amount is subtracted from the baseline stress level to obtain the final stress level. When the baseline stress level is not higher than the stress activation threshold, the baseline stress level is directly determined as the final stress level.

[0025] A pet health monitoring system based on real-time fusion of multi-source heterogeneous data includes:

[0026] The data acquisition and preprocessing module is used to acquire and process multi-source heterogeneous data from wearable devices deployed on pets to generate normalized physiological data, motion data, and location data.

[0027] The baseline stress assessment module is used to calculate baseline stress levels based on normalized physiological data.

[0028] The contextual activity factor building module is used to fuse motion data and location data to construct contextual activity factors;

[0029] The behavior complexity index derivation module is used to derive the behavior complexity index based on motion data through spectrum analysis.

[0030] The dynamic stress level compensation module is used to calculate the final stress level based on whether the baseline stress level is higher than the preset stress activation threshold.

[0031] The health status output module is used to generate monitoring results reflecting the pet's health status based on the final stress level.

[0032] This invention provides an improved method and system for pet health monitoring based on real-time fusion of multi-source heterogeneous data, which has the following improvements and advantages compared with the prior art:

[0033] 1. This solution achieves a comprehensive assessment of pet status by deeply integrating multi-source heterogeneous data acquired and processed from wearable devices, providing a benchmark for subsequent analysis; and integrates different physiological indicators into a unified, preliminary stress score.

[0034] 2. This solution innovatively integrates motion data and location data to construct contextual activity factors, introducing the concept of environmental safety into stress assessment, and using geofencing technology to determine the χ² value. e The parameters act like a logic gating system, ensuring that only high-intensity movements within a preset safety zone are considered as benign activities; this solves the problem that existing technologies cannot distinguish the scenarios in which activities occur, and gives the system the ability to perceive scenarios.

[0035] 3. This scheme further introduces spectrum analysis based on motion data to derive the behavioral complexity index, enabling the scheme to more precisely identify the nature of activities based on the inherent complexity of behavioral patterns;

[0036] 4. The outstanding effect of this solution is mainly reflected in whether the baseline stress level is higher than the preset stress activation threshold. This results in monitoring results that reflect the pet's health status with unprecedented high fidelity and reliability compared to the results produced by existing technologies. It greatly reduces the false alarm rate and provides users with truly valuable insights into their health status. Attached Figure Description

[0037] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0038] Figure 1 This is a flowchart of the system of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0040] Example 1:

[0041] Please see Figure 1 This invention provides a pet health monitoring method based on real-time fusion of multi-source heterogeneous data, including the following steps: S1, acquiring and processing multi-source heterogeneous data from wearable devices deployed on pets to generate normalized physiological data, movement data and location data;

[0042] S2. Calculate the baseline stress level based on normalized physiological data;

[0043] S3. Integrate motion data and location data to construct contextual activity factors;

[0044] S4. Based on motion data, perform spectrum analysis to derive the behavioral complexity index;

[0045] S5. Based on whether the baseline stress level is higher than the preset stress activation threshold, the final stress level is calculated by dynamically compensating for the baseline stress level or directly using it as the output.

[0046] S6. Based on the final stress level, generate monitoring results reflecting the pet's health status;

[0047] A pet health monitoring method based on real-time fusion of multi-source heterogeneous data aims to solve the technical problem of inaccurate health assessment caused by the inability of existing technologies to distinguish between a surge in benign physiological indicators and adverse stress responses. This method establishes a dynamic compensation framework that includes contextual activity factors and behavioral complexity index. Through a series of computational steps, it intelligently identifies and compensates for physiological changes caused by vigorous activities such as normal play in pets, and finally outputs monitoring results that are corrected by multi-dimensional data and can accurately reflect the pet's true health status. This complete processing flow, from raw data collection to final result generation, constitutes a logical closed loop, aiming to improve the reliability of monitoring results.

[0048] Example 2

[0049] S1 specifically includes:

[0050] The system collects heart rate, body temperature, three-axis acceleration, and GPS coordinate data from pets using wearable devices.

[0051] Heart rate and body temperature data are normalized to generate normalized physiological data;

[0052] S2 specifically includes:

[0053] A weighted linear fusion model was used to calculate the baseline stress level by weighted summation of normalized heart rate and normalized body temperature. The weight parameters of the model were empirical parameters set based on the sensitivity of different physiological indicators to stress response.

[0054] In this embodiment, the data acquisition and preprocessing module collects real-time heart rate, body temperature, triaxial acceleration, and GPS coordinate data of the pet through sensors integrated into the wearable device. In order to eliminate the dimensional differences between different physiological indicators and establish a unified comparison benchmark, the system performs normalization processing on the heart rate data. The technical motivation for this normalization operation is that it is necessary to convert the raw physical readings into standardized dimensionless values ​​to facilitate subsequent fusion calculations.

[0055] The body temperature data were also normalized using a linear normalization method, and the mathematical expression is:

[0056] t n =(t r -t m ) / (t x -t m )

[0057] Among them, t n To normalize body temperature, t r For real-time body temperature readings, t m t is the preset resting baseline body temperature. x This is the preset physiological limit body temperature;

[0058] In this embodiment, taking a healthy adult beagle as an example, these preset parameters can be obtained by referring to publicly available veterinary literature or by baseline measurements of the pet; for example, the resting baseline heart rate h can be set. m =60 beats / minute, physiological limit heart rate h x =220 beats / minute; set resting baseline body temperature t m =38.0℃, physiological limit body temperature t x =40.0℃; Technicians in the relevant field may also adjust these benchmark values ​​according to the specific circumstances of different pets, such as breed, age, and health status;

[0059] Its composition and parameter explanations are as follows:

[0060] h n =(h r -h m ) / (h x -h m )

[0061] Among them, hn h represents normalized heart rate r For real-time heart rate readings, h m It is a preset resting baseline heart rate based on veterinary statistics of specific breeds and ages of pets. x The corresponding preset physiological limit heart rate is used; this normalization step provides stable and comparable input for subsequent assessments; subsequently, the baseline stress assessment module calculates the baseline stress level S based on the normalized physiological data. b The calculation employs a weighted linear fusion model, which was chosen due to its simple structure, low computational cost, and clear physical meaning. The mathematical expression of this model is as follows:

[0062] S b =ω h h n +ω t t n

[0063] Among them, S b Based on the level of stress, h n For normalized heart rate, t n For normalized body temperature processed using a similar method, ω h With ω t These are the weighting coefficients for heart rate and body temperature, respectively, and they satisfy ω h +ω t =1; This weighting coefficient is an empirical parameter set based on the sensitivity of different physiological indicators to stress response. Given that instantaneous changes in heart rate can more quickly reflect acute stress, its weight ω h It is given a relatively high initial value, such as 0.7; and satisfies ω h +ω t =1; the specific value of the weight can be fine-tuned according to the individual characteristics of the pet; for example, for pets known to have heart-related diseases, the heart rate weight ω can be appropriately increased. h The value of ω can be adjusted, for example, to 0.8; conversely, if the pet is more prone to abnormal body temperature due to environmental factors or illness, the body temperature weight ω can be increased. t The value; in a preferred embodiment, the system may allow the user or veterinarian to select a preset weighting model during initial setup, such as cardiac sensitivity or body temperature sensitivity;

[0064] This model generates a quantified preliminary stress score S. b This score forms the basis of the input parameters for subsequent dynamic compensation decisions;

[0065] This solution achieves a comprehensive assessment of a pet's condition by deeply fusing multi-source heterogeneous data acquired and processed from wearable devices. The solution not only calculates the baseline stress level S based on normalized physiological data, but also... b The level Sb =ω h h n +ω t t n S b :Basal stress level, ω h Heart rate weighting, h n Normalized heart rate, ω t Body temperature weighting, t n Normalized body temperature (NBT) serves as an initial quantitative value for physiological status, providing a benchmark for subsequent analysis. The practical significance of this formula lies in its integration of different physiological indicators into a unified, preliminary stress score.

[0066] Example 3

[0067] S3 specifically includes:

[0068] Environmental context parameters are determined based on geofencing technology; a motion intensity index is calculated based on motion data; and the environmental context parameters are multiplied by the motion intensity index to obtain the context activity factor.

[0069] The calculation of the exercise intensity index specifically includes:

[0070] A high-pass filter is applied to the triaxial acceleration data to remove the static gravity component and obtain the dynamic acceleration component;

[0071] In this embodiment, a second-order Butterworth high-pass filter can be used, with the cutoff frequency set to f. c =0.5Hz; This parameter is selected to effectively filter out the DC component generated by gravity and the low-frequency noise introduced by the pet's slow posture changes, while retaining the dynamic signals generated by walking, running, jumping and other behaviors.

[0072] Calculate the vector magnitude of the dynamic acceleration components to obtain the dynamic acceleration magnitude; normalize the dynamic acceleration magnitude relative to a preset reference acceleration threshold to obtain the motion intensity index;

[0073] In this embodiment, the contextual activity factor construction module is used to quantify the probability of benign activity, and its output determines whether stress compensation is triggered and its basic strength; this construction process integrates two heterogeneous data types: position and motion; the system introduces the environmental context parameter χ. e The technological motivation is to assign geospatial attributes to activities and determine whether an activity occurs within a predefined safe area. The generation of this parameter relies on geofencing technology; if the pet's current GPS coordinates are within the user-defined safe area, χ... e =1; conversely, χ e=0; This parameter, as a dimensionless binary value, logically acts as a safety gating mechanism; To accurately assess the intensity of a pet's movement, the system uses raw triaxial acceleration data (a x ,a y ,a z Calculate the exercise intensity index I in ) m The key technical point of this process lies in separating the dynamic acceleration generated by body movement; its derivation process first uses a high-pass filter to filter out the gravitational component, obtaining the dynamic acceleration component (d). x ,d y ,d z Secondly, calculate the vector magnitude a of the dynamic acceleration. d :

[0074]

[0075] Among them, a d For the dynamic acceleration modulus, d x ,d y ,d z The dynamic acceleration components of each axis are represented; the motion intensity index I is obtained by normalizing the modulus. m :

[0076] I m =min(1,a) d / a r )

[0077] Among them, I m a is the exercise intensity index. d For the dynamic acceleration modulus, a r Let be a reference acceleration threshold representing high-intensity motion, and min be the minimum value; this threshold 'a'... r The setting is based on the 95th percentile value obtained from statistical analysis of a large amount of movement data of similar pets. This setting method is more robust than using the historical maximum value and can effectively avoid random noise interference. In the embodiment, the reference acceleration threshold a r It can be set to 2g, where g is the standard gravitational acceleration, approximately 9.8 m / s². 2 This value represents the typical acceleration level of most dogs during vigorous play such as chasing and jumping. As an alternative, the system offers a calibration mode upon initial use, guiding the user to allow the pet several minutes of calm and playful activity, during which the system automatically records the dynamic acceleration modulus 'a'. d The distribution is determined, and its 95th percentile value is taken as the a value for that individualization. r The context activity factor κ is obtained by multiplying the two parameters mentioned above:

[0078] κ=χ e ·Im

[0079] Where κ is the contextual activity factor, and χ e For environment context parameters, I m This is the exercise intensity index; this formula links the intensity of exercise to the location where it occurs, and the technical effect is to generate a quantified index κ of normalized exercise intensity in a safe environment; when the pet is not in the safe zone, κ is always 0, thus logically shutting down the subsequent stress compensation channel.

[0080] This solution innovatively integrates motion data and location data to construct a contextual activity factor κ; this factor is expressed by the formula κ = χ. e ·I m ;κ: Contextual activity factor,χ e : Environment context parameters, I m The exercise intensity index is derived from this formula; its practical significance lies in the fact that it introduces the concept of environmental safety into stress assessment for the first time, determined through geofencing technology. e The parameters, like a logical gating system, ensure that only high-intensity movements within a preset safety area are considered as benign activities. This directly solves the problem that existing technologies cannot distinguish the scenarios in which activities occur, and gives the system the ability to perceive scenarios.

[0081] Example 4

[0082] S4 specifically includes:

[0083] Fast Fourier Transform is applied to the time series of dynamic acceleration modulus obtained from the previous steps, and the power spectral density is calculated. The power spectral density is converted into a probability distribution to obtain the power proportion of each frequency bin. Based on the Shannon entropy formula, the power proportion is normalized and the entropy is calculated to obtain the behavior complexity index.

[0084] In this embodiment, the behavior complexity index derivation module performs in-depth screening of activities initially judged to be benign. The technical motivation is that healthy play behavior patterns are usually complex and varied, while running or pacing caused by panic or anxiety are relatively monotonous. This embodiment uses information entropy to quantify the complexity of this behavior pattern. The process of deriving the behavior complexity index β begins by applying a fast Fourier transform to the resultant acceleration signal within a short time window to achieve the conversion from the time domain to the frequency domain. Here, the resultant acceleration signal specifically refers to the dynamic acceleration magnitude a calculated in the previous step. dThe time series; in this embodiment, the sampling frequency of data acquisition is 50Hz, and the time window length for fast Fourier transform is 5 seconds, which includes 250 sampling points; for ease of calculation, the sample data can be padded with zeros to 256 points, so the total number of frequency bins N = 256; these parameters are set to capture short-term dynamic characteristics that reflect behavioral patterns; the power spectral density of the signal is calculated and converted into a probability distribution; the calculation of the behavioral complexity index β is based on Shannon entropy and normalized so that its value falls within the [0,1] interval; the formula for calculating this index is:

[0085]

[0086] Where β is the behavioral complexity index, p i β represents the power proportion of the i-th frequency box, where i is the index of the frequency box, N is the total number of frequency boxes, and log is the logarithmic function to the base 2. The technical effect of this formula is that a high β value indicates that the energy of the movement is widely and evenly distributed in the frequency domain, corresponding to complex and varied healthy play behavior; a low β value means that the energy is highly concentrated in a few frequencies, mapping a monotonous and repetitive movement pattern. Therefore, the behavior complexity index β, as a key dynamic adjustment factor, provides a quantitative basis for distinguishing high-intensity activities of different natures.

[0087] This scheme further introduces spectral analysis based on motion data to derive the behavioral complexity index β; this index is derived using the normalized Shannon entropy formula. β: Behavioral complexity index, p i : Power percentage of frequency i, N: Total number of frequency sub-blocks; The deeper physical meaning of this formula lies in quantifying the information entropy of pet movement patterns; A high β value corresponds to complex behaviors with energy widely distributed in the frequency domain, such as healthy play; while a low β value corresponds to monotonous behaviors with concentrated energy, such as anxious pacing; This allows this scheme to more finely distinguish the nature of activities based on the inherent complexity of behavioral patterns.

[0088] Example 5

[0089] S5 specifically includes:

[0090] When the baseline stress level is higher than the stress activation threshold, the product of the compensation scaling factor, the context activity factor, and the behavioral complexity index is used as the compensation amount. The compensation amount is subtracted from the baseline stress level to obtain the final stress level. When the baseline stress level is not higher than the stress activation threshold, the baseline stress level is directly determined as the final stress level.

[0091] The dynamic stress level compensation module of this invention integrates the calculation results of all previous modules, namely the baseline stress level S. bThe contextual activity factor κ and the behavioral complexity index β are used to perform the final stress level calculation. The initiation of this compensation process is controlled by a preset condition. The underlying logic is that the basic physiological stress index S is only activated when the baseline physiological stress index S is within a certain range. b Exceeding a clinically significant stress activation threshold S t The system only initiates the compensation analysis program at this time; this threshold S t It is a lower limit of stress level set based on veterinary statistical data, and its function is to filter out meaningless low-level physiological fluctuations;

[0092] In a preferred embodiment, S t It can be set to 0.6. Due to the baseline stress level S... b It is a normalized value between 0 and 1. This threshold means that the system considers that a stress state that needs attention may have been entered only when the comprehensive physiological indicators exceed 60% of their normal range.

[0093] When S b >S t When the condition is met, the system uses a conditional linear compensation model to calculate the final stress level S. f The mathematical principle of this model is as follows:

[0094] S f =S b -λ·κ·β

[0095] Among them, S f For the final stress level, S b Let S be the base stress level, λ be the compensation scaling factor, κ be the contextual activity factor, and β be the behavioral complexity index; if S b ≤S t Then S f =S b In this formula, the compensation amount is determined by the product of three factors λ, κ, and β. This multiplicative relationship is the core technology for achieving accurate scene recognition. It ensures that a significant compensation effect is only produced when the contextual activity factor κ and the behavioral complexity index β are both at high values. This situation perfectly corresponds to healthy and intense play scenarios. The compensation scaling factor λ, as a key hyperparameter, is determined by calibration and optimization on a large amount of sample data labeled by veterinarians and behavioral experts. It aims to maximize the model's discriminative power between known play scenarios and known adverse stress scenarios.

[0096] In a default implementation, λ can be set to an empirical value, such as λ = 0.8. To further improve individualization accuracy, this invention can also provide a simple calibration method: In a scenario where the pet is confirmed to be healthy, safe, and actively playing, for example, playing a fetch game with its owner within a designated safe area, the user can activate a play calibration mode. In this mode, the system automatically adjusts the value of λ so that the final stress level S calculated in this scenario is optimal. f =S b -λ·κ·β can be stably reduced to the stress activation threshold S t The following process simplifies the complex parameter optimization into a simple interaction under user supervision, thereby making the technical solution highly feasible.

[0097] The final stress level S produced by this method f It is a high-fidelity health status index that has undergone deep correction based on multimodal data, aiming to effectively reduce false alarms caused by normal pet activities and improve the operational reliability of the monitoring system;

[0098] The remarkable effectiveness of this solution lies in its ability to dynamically compensate for and calculate the final stress level S based on whether the baseline stress level exceeds a preset stress activation threshold. f In the steps; when the conditions are met, the final stress level is determined by formula S. f =S b -λ·κ·β;S f Final stress level, S b : Baseline stress level, λ: compensation scaling factor, κ: contextual activity factor, β: behavioral complexity index; the logical derivation and practical significance of this formula are extremely clear: it is determined from the baseline stress level S b In this process, a compensation amount determined by the contextual activity factor κ and the behavioral complexity index β is subtracted. The multiplicative relationship of this compensation amount ensures that the stress level is reduced to the maximum extent only when the pet is in a safe area with a high κ value and is engaged in complex and varied strenuous exercise with a high β value. This design makes the generated monitoring results reflecting the pet's health status have unprecedented high fidelity and reliability compared to the results produced by existing technologies, greatly reducing the false alarm rate and providing users with truly valuable health status insights.

[0099] Example 6

[0100] A pet health monitoring system based on real-time fusion of multi-source heterogeneous data includes: a data acquisition and preprocessing module, used to acquire and process multi-source heterogeneous data from wearable devices deployed on pets to generate normalized physiological data, movement data and location data;

[0101] The baseline stress assessment module is used to calculate baseline stress levels based on normalized physiological data.

[0102] The contextual activity factor building module is used to fuse motion data and location data to construct contextual activity factors;

[0103] The behavior complexity index derivation module is used to derive the behavior complexity index based on motion data through spectrum analysis.

[0104] The dynamic stress level compensation module is used to calculate the final stress level based on whether the baseline stress level is higher than the preset stress activation threshold.

[0105] The health status output module is used to generate monitoring results reflecting the pet's health status based on the final stress level;

[0106] This invention also provides a pet health monitoring system based on real-time fusion of multi-source heterogeneous data, serving as the physical implementation of the aforementioned method. The system integrates multiple functionally coordinated modules, forming a complete data processing and decision-making chain. The system's workflow begins with the data acquisition and preprocessing module, responsible for collecting and standardizing raw data. The processed data is then transmitted in parallel: normalized physiological data is sent to the basic stress assessment module to calculate the initial stress level (S). b Motion and position data are then fed into the context activity factor construction module and the behavior complexity index derivation module, respectively, to calculate the compensation adjustment factors κ and β; all these intermediate results S b ,κ,β ultimately converge in the dynamic stress level compensation module; this module, based on preset triggering conditions and compensation algorithms, adjusts S b After making corrections, the final stress level S is calculated. f The health status output module receives high-precision S... f The system converts data into clearly defined health status monitoring results. Through its modular structure and sophisticated data flow design, the system engineered complex multimodal data fusion algorithms. The overall technical effect is to provide an intelligent pet health monitoring solution that generates more reliable and accurate health status indicators by distinguishing between benign physiological fluctuations and adverse stress responses.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A pet health monitoring method based on real-time fusion of multi-source heterogeneous data, characterized in that, Includes the following steps: S1. Acquire and process multi-source heterogeneous data from wearable devices deployed on pets to generate normalized physiological data, motion data, and location data; S2. Calculate the baseline stress level based on normalized physiological data; S3. Integrate motion data and location data to construct contextual activity factors; specifically, determine the environmental context parameter χe based on geofencing technology; calculate the motion intensity index Im based on motion data; and multiply the environmental context parameter and the motion intensity index to obtain the contextual activity factors. ； Where κ is the context activity factor, χe is the environmental context parameter, and Im is the motion intensity index; S4. Based on motion data, perform spectral analysis to derive the behavioral complexity index; specifically, apply Fast Fourier Transform to the time series of dynamic acceleration moduli obtained from previous steps and calculate the power spectral density; convert the power spectral density into a probability distribution to obtain the power proportion of each frequency bin; calculate the normalized entropy of the power proportion based on the Shannon entropy formula to obtain the behavioral complexity index: ； Where β is the behavior complexity index, pi represents the power proportion of the i-th frequency box, i is the index of the frequency box, N is the total number of frequency boxes, and log is the logarithmic function with base 2. S5. Determine whether the baseline stress level is higher than the preset stress activation threshold St; when the baseline stress level Sb is higher than the stress activation threshold St, use the product of the compensation scaling factor λ, the context activity factor κ, and the behavioral complexity index β as the compensation amount, subtract the compensation amount from the baseline stress level to obtain the final stress level: ； Where Sf is the final stress level, Sb is the basic stress level, λ is the compensation scaling factor, κ is the context activity factor, and β is the behavioral complexity index. When the baseline stress level Sb is not higher than the stress activation threshold St, the baseline stress level is directly determined as the final stress level, that is: ； S6. Based on the final stress level, generate monitoring results reflecting the pet's health status.

2. The pet health monitoring method based on real-time fusion of multi-source heterogeneous data according to claim 1, characterized in that, S1 specifically includes: The system collects heart rate, body temperature, three-axis acceleration, and GPS coordinate data from pets using wearable devices. Heart rate and body temperature data are normalized to generate normalized physiological data.

3. The pet health monitoring method based on real-time fusion of multi-source heterogeneous data according to claim 1, characterized in that, S2 specifically includes: A weighted linear fusion model was used to calculate the baseline stress level by weighted summation of normalized heart rate and normalized body temperature. The weight parameters of the model were empirical parameters set based on the sensitivity of different physiological indicators to stress response.

4. The pet health monitoring method based on real-time fusion of multi-source heterogeneous data according to claim 3, characterized in that, The calculation of the exercise intensity index specifically includes: A high-pass filter is applied to the triaxial acceleration data to remove the static gravity component and obtain the dynamic acceleration component; the vector magnitude of the dynamic acceleration component is calculated to obtain the dynamic acceleration magnitude; the dynamic acceleration magnitude is normalized relative to a preset reference acceleration threshold to obtain the motion intensity index.

5. A pet health monitoring system based on real-time fusion of multi-source heterogeneous data, applied to the pet health monitoring method based on real-time fusion of multi-source heterogeneous data as described in any one of claims 1-4, characterized in that, include: The data acquisition and preprocessing module is used to acquire and process multi-source heterogeneous data from wearable devices deployed on pets to generate normalized physiological data, motion data, and location data. The baseline stress assessment module is used to calculate baseline stress levels based on normalized physiological data. The contextual activity factor building module is used to fuse motion data and location data to construct contextual activity factors; The behavior complexity index derivation module is used to derive the behavior complexity index based on motion data through spectrum analysis. The dynamic stress level compensation module is used to calculate the final stress level based on whether the baseline stress level is higher than the preset stress activation threshold. The health status output module is used to generate monitoring results reflecting the pet's health status based on the final stress level.

Images