Animal health state monitoring method and system based on fusion of voiceprint and body feature

Abstract

The application relates to the technical field of animal health monitoring, and specifically provides an animal health state monitoring method based on the fusion of voiceprints and body features, which comprises the following steps: collecting voiceprints and body features of a target animal in a healthy stable state for anti-interference processing; constructing a dynamic baseline library corresponding to the target animal based on the processed voiceprints and body features; collecting real-time voiceprints and body features of the target animal to respectively compare and analyze the dynamic baseline library, obtaining feature deviation degrees, and correlatively analyzing the time sequence correlation of the real-time voiceprints and body features, obtaining time sequence correlation scores; dynamically adjusting the fusion weights of the real-time voiceprints and body features according to environmental noise intensity, and outputting animal health state evaluation results based on the feature deviation degrees and the time sequence correlation scores. The scheme improves the accuracy of the health state judgment of the target animal, so as to realize high precision and strong robustness of animal health monitoring in complex scenes.

Description

Technical Field

[0001] This invention relates to the field of animal health monitoring technology, specifically to a method and system for monitoring animal health status based on the fusion of voiceprint and body features. Background Technology

[0002] In animal husbandry (large-scale pig and dairy farming, etc.) and wildlife conservation management, real-time monitoring of animal health status is a core requirement for ensuring farming efficiency, species protection, and animal welfare. With the increasing scale of farming and the refinement of conservation scenarios, traditional monitoring methods relying on manual inspections have gradually revealed their limitations: manual monitoring not only has limited coverage and low efficiency, but is also easily influenced by subjective experience, making it difficult to detect early, latent health abnormalities in animals (such as mild respiratory infections or digestive discomfort), leading to delayed disease detection and increased farming losses or conservation risks.

[0003] Currently, animal health monitoring faces two major challenges: First, complex environmental interferences such as noise from farm fans, wind and rain in the wild, changes in lighting, and obstruction by objects cause mixed voiceprint signals, loss or misidentification of physical characteristics, making it difficult to obtain high-quality monitoring data. Second, existing solutions often adopt species-uniform health judgment standards (such as using the same normal voiceprint frequency range for all dairy cows), ignoring the differences in age, breed, and growth stage of individual animals, which can easily lead to misjudgment or omission of health status.

[0004] Therefore, this invention provides a method for monitoring animal health status based on the fusion of voiceprint and body features to solve the above-mentioned technical problems. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the present invention provides a method and system for monitoring animal health status based on the fusion of voiceprint and body features, so as to solve the problems in the prior art.

[0006] One embodiment of the present invention provides a method for monitoring animal health status based on the fusion of voiceprint and body features, comprising the following steps: Voiceprint and physical characteristics of the target animal in a healthy and stable state were collected and subjected to anti-interference processing to obtain processed voiceprint and physical characteristics. Based on the processed voiceprint and body features, a dynamic baseline database corresponding to the target animal is constructed, and the dynamic baseline database is dynamically updated as the animal grows. Real-time voiceprint and body features of the target animal are collected. The real-time voiceprint and body features are compared and analyzed with the dynamic baseline database to obtain the feature deviation, which includes voiceprint feature deviation and body feature deviation. Correlation analysis was performed to determine the temporal correlation between real-time voiceprint features and abnormal body posture features, and a temporal correlation score was obtained. The fusion weights of real-time voiceprint features and body features are dynamically adjusted based on the environmental noise intensity, and the animal health status assessment result is output based on the feature deviation and temporal correlation score.

[0007] By adopting the above-mentioned scheme, and using anti-interference processing technology to filter environmental noise and supplement the shortcomings of body feature collection, the problems of mixed voiceprints and loss of body features in complex environments are effectively solved, providing high-quality data support for health monitoring. By constructing an individual dynamic baseline database that is updated with growth stages, the problems of misjudgment and omission caused by ignoring individual differences due to uniform standards for species are avoided, and the scheme adapts to the physiological changes of different animals. Combined with real-time feature comparison, abnormal temporal correlation analysis, and adaptive weight adjustment for environmental noise, the accuracy of health status judgment is further improved, so as to achieve high precision and strong robustness of animal health monitoring in complex scenarios.

[0008] In one embodiment, the step of constructing a dynamic baseline database corresponding to the target animal based on the processed voiceprint and body features, wherein the dynamic baseline database is dynamically updated as the animal grows, specifically includes the following steps: The processed voiceprint features are parametrically extracted to obtain voiceprint baseline parameters, which include the normal voiceprint frequency range, voiceprint energy fluctuation range, and abnormal voiceprint determination threshold of the target animal in a healthy state. The processed body features are parametrically extracted to obtain body baseline parameters, which include the normal range of limb movement, the range of average daily resting time, and the threshold of muscle tremor frequency in the target animal under healthy conditions. The voiceprint baseline parameters and body posture baseline parameters are associated and bound with the unique identification information of the target animal, and the current growth stage is tagged. The associated voiceprint baseline parameters, body shape baseline parameters, unique identification information of the target animal, and current growth stage label are stored according to a preset data structure to obtain a dynamic baseline library that is only adapted to the target animal.

[0009] By adopting the above scheme, the baseline parameters are linked and bound to the unique identification information of the target animal and tagged with growth stage labels, ensuring strong adaptability between the dynamic baseline database and a single individual. This avoids the risk of misjudgment caused by the confusion of baselines of different individuals from the source of construction. At the same time, storing the associated information according to the preset data structure provides structured support for the subsequent updates of the dynamic baseline database with growth stages and for efficient comparison with real-time collected features, further enhancing the feasibility and accuracy of the overall monitoring method.

[0010] In one embodiment, the step of collecting real-time voiceprint and body posture features of the target animal, and comparing and analyzing the real-time collected voiceprint and body posture features with the dynamic baseline database to obtain the feature deviation, specifically includes the following steps: The system acquires voiceprint signals and body images of the target animal in real time and performs anti-interference processing to extract real-time voiceprint features and body features, and obtains information on the current growth stage of the target animal. The voiceprint baseline parameters that match the current growth stage are retrieved from the dynamic baseline library. The real-time voiceprint features are compared with the voiceprint baseline parameters of this stage in different dimensions. The voiceprint dimension deviation results are obtained by combining the physiological weights of each dimension feature in the voiceprint baseline parameters in the current growth stage. The body posture baseline parameters that match the current growth stage are retrieved from the dynamic baseline library. The real-time body posture features are compared with the body posture baseline parameters of this stage in terms of dimensions. The body posture dimension deviation results are obtained by combining the physiological weights of each dimension feature in the body posture baseline parameters in the current growth stage. Cross-validation was performed on the deviation results of voiceprint dimension and body shape dimension, and the deviation results of voiceprint feature and body shape feature were fused to obtain the deviation degree of voiceprint feature and body shape feature that are suitable for the current growth stage.

[0011] By adopting the above scheme, the baseline parameters matched with the real-time collected features are compared dimensionally, avoiding the comparison bias caused by using a single baseline across all growth stages. By combining the physiological weights of each dimension feature in the baseline parameters at the corresponding growth stage, the deviation results are more aligned with the health monitoring priorities of different stages, improving the targeting of feature comparison. At the same time, by cross-validating the deviation results of voiceprint and body posture dimensions, the mutual verification of dual-modal features is strengthened, reducing the risk of misjudgment caused by single feature anomalies, and ultimately improving the accuracy and reliability of feature deviation calculation.

[0012] In one embodiment, the step of analyzing the temporal correlation between real-time voiceprint features and abnormal body posture features to obtain a temporal correlation score specifically includes the following steps: Based on the deviation of the voiceprint feature and the deviation of the body feature, it is determined whether the deviation of the real-time voiceprint feature and the real-time body feature both exceed the preset abnormal threshold. When the deviation of real-time voiceprint features and real-time body posture features both exceed the preset abnormal threshold, abnormal time sequence information of abnormal voiceprints and abnormal body postures is extracted. The abnormal time sequence information includes the time point of abnormal occurrence, duration and trend of change. Based on the physiological correlation patterns of the target animal's current growth stage, abnormal time-series information is assigned correlation weights. Based on the aforementioned correlation weights, the temporal correlation degree between real-time voiceprint feature anomalies and real-time body feature anomalies is calculated, resulting in a correlation score between real-time voiceprint features and real-time body features in the time dimension.

[0013] By adopting the above scheme, we focus on scenarios where both voiceprint and physical characteristics are abnormal to conduct temporal correlation analysis, avoiding interference from single abnormal scenarios and ensuring that the analysis object matches the core requirements of the correlation. By assigning correlation weights based on the physiological correlation patterns of the target animal's current growth stage, the temporal correlation analysis is more in line with the physiological characteristics of animals at different growth stages, improving the pertinence and rationality of the correlation judgment. At the same time, by calculating the temporal correlation degree to obtain a quantitative correlation score, the qualitative temporal relationship is transformed into a measurable indicator, reducing the risk of misjudgment caused by subjective judgment. Ultimately, this improves the accuracy of temporal correlation analysis of abnormal voiceprint and physical characteristics, providing a more reliable temporal correlation basis for the accurate determination of the animal's health status.

[0014] In one embodiment, the step of dynamically adjusting the fusion weights of real-time voiceprint features and body features based on environmental noise intensity, and outputting animal health status assessment results based on the feature deviation and temporal correlation, specifically includes the following steps: The ambient noise intensity is collected in real time, and the current ambient noise level is determined according to the preset noise classification standard. The fusion weights of real-time voiceprint features and real-time body features are dynamically assigned based on noise level. The higher the noise level, the lower the fusion weight of real-time voiceprint features and the higher the fusion weight of real-time body features. The comprehensive health status assessment value is obtained by weighting the voiceprint feature deviation, body posture feature deviation, and temporal correlation score with the fusion weight. Based on the preset range in which the comprehensive health status assessment value falls, the corresponding animal health status assessment result is output.

[0015] By adopting the above scheme, the fusion weights of voiceprint and body features are dynamically adjusted according to the environmental noise intensity. The higher the noise level, the lower the weight of voiceprint features and the higher the weight of body features, effectively avoiding the impact of noise interference on voiceprint feature distortion on the assessment results and improving the adaptability of feature fusion in different environments. At the same time, the deviation degree of voiceprint and body features and the time sequence correlation score are combined with the fusion weight for weighted calculation, realizing the complementary fusion of multi-dimensional health-related information and avoiding the one-sidedness of assessment by a single feature or single indicator. Then, the corresponding health status assessment results are output through preset intervals, making the assessment results clearer and more explicit, and reducing subjective judgment errors. Ultimately, the accuracy and environmental adaptability of animal health status assessment are improved, providing a more reliable basis for subsequent targeted health interventions.

[0016] In one embodiment, before the step of dynamically adjusting the fusion weights of real-time voiceprint features and body features based on environmental noise intensity, and outputting the animal health status assessment result based on the feature deviation and temporal correlation score, the following step is further included: If any one or both of the voiceprint feature deviation and body posture feature deviation exceed the preset anomaly threshold, and the temporal correlation score is lower than the preset correlation threshold, anomaly verification is triggered. By retrospectively examining the fluctuation range of voiceprint and physical characteristics at the current growth stage in the dynamic baseline database, it can be determined whether the current abnormal characteristics are within the normal fluctuation range that is not related to health at this stage. Retrieve environmental parameters from the same period to determine whether abnormal voiceprint characteristics are caused by sudden noise and whether abnormal body posture characteristics are caused by temporary site interference. Based on the judgment results, if it is confirmed that the abnormal feature is caused by non-health factors, the current abnormal feature is marked as invalid, and the real-time voiceprint features and real-time body features are backtracked and re-collected; if it is confirmed that the abnormality is caused by health factors, the current abnormal feature and time-series correlation score are retained, and the process proceeds to the next step.

[0017] By adopting this scheme, an anomaly verification step is added before health status assessment. This step precisely triggers the verification process for questionable scenarios with abnormal features but low temporal correlation. By retrospectively examining the growth stage feature fluctuation range of the dynamic baseline database, the scheme effectively distinguishes between normal physiological fluctuations and health-related anomalies, avoiding misjudging normal feature fluctuations as health problems. Combined with the investigation of environmental parameters during the same period, it clarifies whether voiceprint anomalies are caused by sudden noise and whether body posture anomalies are caused by temporary site interference, thus eliminating the influence of environmental interference on bimodal features. Finally, by distinguishing between anomalies caused by healthy factors and non-healthy factors, invalid anomalies are marked and re-collected, while valid anomalies are retained and included in subsequent assessments. This further leverages the unique advantages of the dynamic baseline database and voiceprint-body posture bimodal fusion, reduces the risk of misjudgment, improves the accuracy of anomaly feature screening, ensures that subsequent health status assessment results are more consistent with the animal's actual health status, and enhances the reliability and practicality of the entire monitoring method.

[0018] This application also relates to an animal health status monitoring system based on the fusion of voiceprint and body features, comprising: The data acquisition module is used to collect the voiceprint and physical characteristics of the target animal in a healthy and stable state, and to perform anti-interference processing on them to obtain the processed voiceprint and physical characteristics. The baseline library construction module is used to construct a dynamic baseline library corresponding to the target animal based on the processed voiceprint features and body features. The dynamic baseline library is dynamically updated as the animal grows. The feature analysis module is used to collect real-time voiceprint features and body features of the target animal, and compare and analyze the real-time collected voiceprint features and body features with the dynamic baseline database to obtain the feature deviation, which includes voiceprint feature deviation and body feature deviation. The correlation scoring module is used to analyze the temporal correlation between real-time voiceprint features and abnormal body posture features, and obtain a temporal correlation score. The output module is used to dynamically adjust the fusion weights of real-time voiceprint features and body features according to the environmental noise intensity, and output the animal health status assessment results based on the feature deviation degree and temporal correlation score.

[0019] In one embodiment, it further includes: The anomaly verification module is used to trigger anomaly verification when any one or both of the voiceprint feature deviation and body posture feature deviation exceed the preset anomaly threshold, and the temporal correlation score is lower than the preset correlation threshold.

[0020] This application also relates to a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for monitoring animal health status based on the fusion of voiceprint and body features.

[0021] This application also relates to a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for monitoring animal health status based on the fusion of voiceprint and body features.

[0022] The animal health status monitoring method and system based on the fusion of voiceprint and body features provided in the above embodiments have the following beneficial effects: By using anti-interference processing technology to filter environmental noise and supplement the shortcomings of body feature collection, the problems of mixed voiceprints and loss of body features in complex environments are effectively solved, providing high-quality data support for health monitoring. By constructing an individual dynamic baseline database that is updated with growth stages, the problems of misjudgment and omission caused by ignoring individual differences due to uniform standards for species are avoided, and the system adapts to the physiological changes of different animals. Combined with real-time feature comparison, abnormal temporal correlation analysis, and adaptive weight adjustment for environmental noise, the accuracy of health status judgment is further improved, so as to achieve high precision and strong robustness of animal health monitoring in complex scenarios. Attached Figure Description

[0023] Figure 1 A flowchart illustrating an animal health status monitoring method based on the fusion of voiceprint and body features, provided in an embodiment of the present invention; Figure 2This is a schematic block diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions in the embodiments of the present invention will now be clearly and completely described in conjunction with the accompanying drawings.

[0025] Reference Figure 1 One embodiment of the present invention provides a method for monitoring animal health status based on the fusion of voiceprint and body features, comprising the following steps: S10. Collect the voiceprint and physical characteristics of the target animal in a healthy and stable state, and perform anti-interference processing on them to obtain the processed voiceprint and physical characteristics.

[0026] In this embodiment, step S10 is used to obtain health baseline features. When the target animal (such as piglets, fattening pigs, etc.) is in a healthy and stable state, voiceprints (such as feeding sounds, resting hums) are collected by a directional microphone array, and body postures (such as activity amplitude, respiratory rate) are collected by millimeter-wave radar. Wavelet threshold noise reduction (for voiceprint features) and Kalman filtering (for body posture features) are used for anti-interference processing, and finally, core features such as Mel-frequency cepstral coefficients (MFCC) and limb movement parameters are extracted to ensure that the signal-to-noise ratio of the baseline data is improved and the integrity of the data is maintained.

[0027] S20. Based on the processed voiceprint and body features, construct a dynamic baseline database corresponding to the target animal. The dynamic baseline database is dynamically updated as the animal grows.

[0028] In this embodiment, step S20 constructs a dynamic baseline library based on the voiceprint and body features processed in S10. The normal fluctuation range of voiceprint and body features (e.g., mean ± 2 standard deviations) is determined through statistical analysis and stored according to "individual ID - growth stage". At the same time, stage nodes are set (e.g., piglet stage 0-30 days old, fattening stage 31-180 days old). Before the end of each stage, new voiceprint and body feature collection is automatically triggered to update the baseline library. When the difference between the newly collected health features and the historical baseline exceeds a preset threshold (e.g., 15%), the baseline library is updated with the new feature distribution to solve the misjudgment problem caused by feature drift during the growth process.

[0029] S30. Collect real-time voiceprint and body features of the target animal, and compare and analyze the real-time voiceprint and body features with the dynamic baseline database to obtain the feature deviation, which includes voiceprint feature deviation and body feature deviation.

[0030] In this embodiment, step S30 is used to compare real-time features with the baseline. The same device as in S10 is used to collect voiceprint and body posture data in real time and preprocess them. The feature deviation (range 0~1) is obtained by calculating "|real-time value - baseline mean| / baseline standard deviation". The voiceprint deviation is based on MFCC feature comparison, and the body posture deviation is based on parameters such as respiratory rate and activity amplitude. The difference between real-time features and health benchmark is quantified (≥1 indicates that it exceeds the normal range).

[0031] S40. Correlation analysis is performed to determine the temporal correlation between real-time voiceprint features and abnormal body posture features, and a temporal correlation score is obtained. In this embodiment, step S40 is used to analyze the temporal correlation when there are abnormalities in bimodal features (voiceprint, body posture). When the deviation of voiceprint / body posture is greater than or equal to a preset threshold (e.g., ≥1), the time point, duration and trend of the abnormality are extracted. If only a single modal feature is abnormal, the temporal correlation score is set to 0. If the bimodal feature is abnormal, the score (range 0~0.8) is calculated based on the time difference (e.g., ≤30 minutes, 0.5 points) and trend consistency (e.g., synchronous increase, 0.3 points) to determine whether the abnormality is caused by the same health problem.

[0032] S50. Dynamically adjust the fusion weights of real-time voiceprint features and body features according to the environmental noise intensity, and output the animal health status assessment results based on the feature deviation degree and temporal correlation score.

[0033] In this embodiment, step S50 combines environmental conditions to output health assessment results. First, the environmental noise is classified by noise acquisition equipment (e.g., low noise ≤50dB, medium noise 51-70dB, high noise >70dB), and weights are dynamically allocated (e.g., low noise: voiceprint 0.6 / body posture 0.4; medium noise: 0.5 / 0.5; high noise: 0.3 / 0.7). Then, the comprehensive assessment value is calculated according to the formula "voiceprint deviation × voiceprint weight + body posture deviation × body posture weight + time series score × 0.2". Finally, the results are output according to the comprehensive assessment value <0.3 (healthy), 0.3-0.7 (mildly abnormal), and ≥0.7 (severely abnormal).

[0034] In this embodiment, the scenario is assumed to be a large-scale pig farm piglet health monitoring scenario. The monitoring system conducts health monitoring on 30-day-old piglets, focusing on the prevention of respiratory and digestive tract diseases, and achieves early detection and early intervention through the fusion of voiceprint and body characteristics. Step S10: When the piglet "Xiaobai" is in a healthy and stable state (normal appetite, daily activity frequency of about 8 times / hour), the monitoring system continuously collects its healthy voiceprint (such as grunting sounds when eating, and soft cries when resting) for 72 hours through a directional microphone array (sampling rate 16kHz). At the same time, it collects body posture data (such as respiratory rate of 25 breaths / minute, limb movement amplitude) using millimeter-wave radar (sampling frequency 20Hz). Wavelet threshold noise reduction is used to eliminate fan background noise in the voiceprint data, and Kalman filtering is used to remove abnormal points caused by collisions with other piglets in the body posture data, resulting in the processed voiceprint core features (Mean Mel-frequency cepstral coefficient MFCC mean 1.2kHz) and body posture core parameters.

[0035] Step S20: Based on the voiceprint and body characteristics processed in S10, the system constructs a dynamic baseline library exclusive to "Xiaobai"; it stores the data according to "Piglet ID (Xiaobai) - Growth Stage (30-day-old piglet period)" and statistically determines the normal fluctuation range of health characteristics: voiceprint MFCC 0.9~1.5kHz (mean ±0.3kHz), body breathing rate 20~30 breaths / minute, and activity amplitude 6~10 times / hour; at the same time, the growth stage node is preset (30 days old is the end of the piglet period), and the baseline update detection is automatically triggered 48 hours before the end of the stage.

[0036] Step S30: The monitoring system uses the same equipment as in S10 to collect "Xiaobai's" voiceprint and body posture data in real time; voiceprint frequency 1.8kHz (sharp), body posture respiratory rate 40 breaths / minute, activity amplitude 3 times / hour; compare the real-time features with the dynamic baseline database, and calculate the deviation according to the formula "|real-time value - baseline mean| / baseline standard deviation": voiceprint deviation = |1.8-1.2| / 0.3 = 2.0, body posture deviation (breathing) = |40-25| / 5 = 3.0, (activity) = |3-8| / 2 = 2.5, the overall body posture deviation is 2.75 (both exceed the preset abnormal threshold of 1).

[0037] Step S40: System correlation analysis of the temporal correlation between "Xiaobai's" voiceprint and abnormal body posture; extract abnormal temporal sequence: voiceprint abnormality begins at 8:00 (frequency continues to rise), body posture abnormality begins at 8:10 (rapid breathing, reduced activity, rising synchronously); because both features are abnormal, the score is calculated according to the rules: time difference 10 minutes ≤ 30 minutes (base score 0.5), consistent trend (add 0.3 points), final temporal correlation score = 0.8 (exceeding the preset correlation threshold of 0.5).

[0038] Step S50: The system first detects the ambient noise as 65dB (medium noise) using a noise sensor, and dynamically assigns fusion weights: voiceprint 0.5 and body posture 0.5; then calculates the comprehensive evaluation value according to the formula "voiceprint deviation × voiceprint weight + body posture deviation × body posture weight + temporal correlation score × 0.2" = (2.0 × 0.5) + (2.75 × 0.5) + (0.8 × 0.2) = 2.535; because the comprehensive evaluation value is ≥ 0.7 (severe abnormality threshold), the system triggers an audible and visual alarm, and synchronously uploads the data to the pig farm management platform, prompting the evaluation result "Piglet Xiaobai may have respiratory disease and needs urgent examination".

[0039] In one embodiment, step S20 specifically includes the following steps: S21. The processed voiceprint features are parametrically extracted to obtain voiceprint baseline parameters, which include the normal voiceprint frequency range, voiceprint energy fluctuation range, and abnormal voiceprint determination threshold of the target animal in a healthy state. S22. Parametrically extract the processed body features to obtain body baseline parameters, which include the normal range of limb movement, the range of average daily resting time, and the threshold of muscle tremor frequency of the target animal in a healthy state. S23. Associate and bind the voiceprint baseline parameters and body posture baseline parameters with the unique identification information of the target animal, and mark the current growth stage tag; S24. Store the associated voiceprint baseline parameters, body shape baseline parameters, unique identification information of the target animal, and current growth stage label according to the preset data structure to obtain a dynamic baseline library that is only adapted to the target animal.

[0040] In this embodiment, the core is to form a dedicated baseline library that is only adapted to the target animal through parameterized extraction, association binding, and structured storage, as detailed below: Step S21: The core of this step is to transform the processed voiceprint features (effective voiceprint data after anti-interference) from S10 into quantifiable and comparable baseline parameters. Parametric extraction refers to extracting representative quantitative indicators from the processed voiceprint using signal processing algorithms (such as feature engineering methods), rather than retaining the original voiceprint signal. The voiceprint baseline parameters are the benchmark standard for voiceprint features in a healthy state of the target animal, used for subsequent comparison with real-time voiceprints to determine whether there are abnormalities. Specifically, the three types of parameters include the following meanings: The normal voiceprint frequency range is the main frequency range of the voiceprint signal (such as vocalizations and breathing sounds) of a healthy animal (e.g., the grunting sound of a healthy piglet is mostly between 0.9 and 1.5 kHz). The range of voiceprint energy fluctuation is the range of energy intensity variation of the voiceprint signal in a healthy state (e.g., the energy value fluctuates within -45dB to -30dB; if it exceeds this range, it may be due to emotional stress or illness causing abnormal voiceprint energy). The threshold for judging abnormal voiceprints is a pre-set critical value used to initially determine whether a voiceprint is abnormal (such as when the voiceprint frequency exceeds the normal range and the energy fluctuation exceeds the range, an abnormal judgment is triggered).

[0041] Step S22: This step is logically consistent with S21. It involves parametrically extracting the body posture features processed in S10 (effective body posture data after anti-interference) to form a health baseline for the body posture dimension. The body posture baseline parameters are the benchmark standards for body posture features in a healthy state of the target animal, used for subsequent comparison with real-time body posture. Specifically, the three types of parameters include the following meanings: The normal range of motion of the limbs is the range or frequency of limb (such as limbs and trunk) when the target animal is healthy (e.g., when piglets are healthy, they stand and walk 6 to 10 times per hour). The average daily resting time range is the time interval during which the target animal rests quietly each day in a healthy state (e.g., 30-day-old piglets rest for an average of 18-20 hours per day; too short a time may cause discomfort and restlessness, while too long a time may cause weakness and lack of energy). The muscle tremor frequency threshold is a pre-set critical value for judging whether the muscles are trembling abnormally (e.g., when piglets are healthy, the muscle tremor frequency is extremely low. If it exceeds 5 times / minute and lasts for 10 minutes, it may be a stress or disease signal).

[0042] Step S23: The core of this step is to establish the correspondence between "parameters-animals-growth stages" to avoid confusion of baseline parameters. Specifically, the unique identifier of the target animal refers to an identifier that can uniquely distinguish the animal (such as an electronic ear tag ID or a unique number within the farm, ensuring that the parameters are not confused with those of other animals); the association and binding refers to logically associating the vocal print baseline parameters from S21 and the body shape baseline parameters from S22 with the animal's unique identifier, making the parameters clearly point to a specific animal; the current growth stage label refers to marking the current growth stage of the target animal for this set of parameters (such as 30-day-old piglet stage, 100-day-old fattening stage), laying the foundation for subsequent baseline updates as the growth stage progresses.

[0043] Step S24: The core of this step is to standardize and store all the associated information to form the final dynamic baseline library. The preset data structure refers to a predefined data storage format (such as JSON format or database table structure, ensuring data can be quickly queried, retrieved, and updated); the associated data includes four types of information: voiceprint baseline parameters, body shape baseline parameters, animal unique identifier, and growth stage tags; "Only compatible with this target animal" emphasizes that this baseline library is built based on the animal's own health data and is not applicable to other animals (to avoid comparison errors caused by individual differences); the dynamic baseline library means that the library is not fixed and can be updated according to changes in the target animal's growth stage (such as entering a new growth period) to maintain the baseline's compatibility with the animal's actual health status.

[0044] In one embodiment, step S30 specifically includes the following steps: S31. Real-time acquisition of voiceprint signals and body images of the target animal and anti-interference processing, extraction of real-time voiceprint features and real-time body features, and acquisition of information on the current growth stage of the target animal. S32. Call the voiceprint baseline parameters that match the current growth stage from the dynamic baseline library, compare the real-time voiceprint features with the voiceprint baseline parameters of this stage in different dimensions, and combine the physiological weights of each dimension feature in the voiceprint baseline parameters in the current growth stage to obtain the voiceprint dimension deviation result. S33. Call the body posture baseline parameters that match the current growth stage from the dynamic baseline library, compare the real-time body posture features with the body posture baseline parameters of this stage in terms of dimensions, and combine the physiological weights of each dimension feature in the body posture baseline parameters in the current growth stage to obtain the body posture dimension deviation results. S34. Cross-validate the deviation results of voiceprint dimension and body shape dimension, and fuse them to obtain the deviation degree of voiceprint feature and body shape feature that are adapted to the current growth stage.

[0045] In this embodiment, the core is to obtain the deviation between voiceprint and body features that are adapted to the current growth stage of the target animal through real-time feature acquisition, multi-dimensional baseline comparison, and cross-validation fusion, as follows: Step S31: The core of this step is to acquire real-time, effective bimodal features and the current growth stage of the target animal, providing basic data for subsequent baseline comparison. Real-time acquisition refers to synchronously acquiring the target animal's voiceprint signals (e.g., vocalizations and breathing sounds acquired via microphone array) and body posture images (e.g., body movements acquired via millimeter-wave radar / camera) at a preset frequency (e.g., once / second). Anti-interference processing follows the same technical logic as S10: wavelet threshold denoising is used to eliminate environmental noise (e.g., pig farm fan noise) in the voiceprint signals, and Kalman filtering is used to remove irrelevant interference (e.g., occlusion by other animals) from the body posture images. Extracting real-time voiceprint / body posture features refers to extracting features corresponding to the baseline database parameters from the anti-interference data (e.g., voiceprint frequency and energy, body posture activity amplitude, and resting duration). Acquiring current growth stage information refers to determining the animal's current stage (e.g., 30-day-old piglet stage, 100-day-old fattening stage) by using pre-stored animal growth cycle records (e.g., calculating from the birth date of the piglet "Xiaobai"), ensuring the subsequent use of matching baseline parameters.

[0046] Step S32: The core of this step is to accurately compare the real-time voiceprint features with the voiceprint baseline parameters of the target animal at its current growth stage to obtain multi-dimensional voiceprint deviation data. Among them, the dimensional comparison refers to comparing the real-time voiceprint features with the baseline standard of the corresponding dimension according to the classification of voiceprint baseline parameters (such as normal voiceprint frequency range, voiceprint energy fluctuation range) (e.g., real-time voiceprint frequency 1.8kHz VS baseline frequency range 0.9~1.5kHz, real-time voiceprint energy -50dB VS baseline energy range -45dB~-30dB); the physiological weight refers to the importance coefficient set for each dimension of voiceprint features according to the physiological characteristics of the current growth stage (e.g., the voiceprint frequency in the piglet period has a stronger correlation with the health status, so the weight is set to 0.6; the voiceprint energy has a weaker correlation, so the weight is set to 0.4); the voiceprint dimension deviation result refers to the calculation result of "difference value between real-time feature and baseline × physiological weight of the dimension" for each dimension (e.g., frequency dimension difference value 0.9kHz × 0.6 = 0.54, energy dimension difference value -5dB × 0.4 = 0.2, initially obtaining the deviation data of each dimension of voiceprint).

[0047] Step S33: This step is logically consistent with S32. The core is to compare the real-time body posture characteristics with the body posture baseline parameters of the target animal at its current growth stage in different dimensions to obtain body posture deviation data in different dimensions. Among them, the dimensional comparison refers to comparing real-time body posture characteristics with the baseline standards of the corresponding dimensions according to the classification of body posture baseline parameters (such as the normal range of limb activity and the range of average daily resting time). For example, real-time activity range of 3 times / hour vs. baseline activity range of 6-10 times / hour, real-time resting time of 22 hours vs. baseline resting time of 18-20 hours. The physiological weight refers to the importance coefficient set for each dimension of body posture characteristics according to the physiological characteristics of the current growth stage. For example, the resting time of fattening animals has a stronger correlation with the health status, so the weight is set to 0.5; the activity range weight is set to 0.5; the activity range weight of piglets can be increased to 0.6. The body posture dimension deviation result refers to the calculation result of "the difference between real-time characteristics and baseline × the physiological weight of the dimension" for each dimension. For example, the difference in activity range -5 times / hour × 0.6 = 0.3, the difference in resting time 2 hours × 0.4 = 0.08, and the preliminary deviation data of each dimension of body posture are obtained.

[0048] Step S34: The core of this step is to eliminate single-dimensional misjudgments through cross-validation and fuse them to obtain the final feature deviation. Cross-validation refers to verifying whether there is logical consistency between the deviations of the voiceprint dimension and the body posture dimension (e.g., when the voiceprint display is abnormal, does the body posture also show abnormalities simultaneously, eliminating interference scenarios of "abnormal voiceprint but normal body posture," such as voiceprint misjudgment caused by sudden noise); the voiceprint feature deviation refers to the weighted summation of the deviation results of each dimension of the voiceprint that passed the validation (e.g., frequency dimension 0.54 + energy dimension 0.2 = 0.74 in S32); the body posture feature deviation refers to the weighted summation of the deviation results of each dimension of the body posture that passed the validation (e.g., activity amplitude 0.3 + resting duration 0.08 = 0.38 in S33), ultimately obtaining a feature deviation that is suitable for the current growth stage and has no single-dimensional misjudgments.

[0049] In one embodiment, step S40 specifically includes the following steps: S41. Based on the deviation of the voiceprint feature and the deviation of the body feature, determine whether the deviation of the real-time voiceprint feature and the real-time body feature both exceed the preset abnormal threshold. S42. When the deviation of real-time voiceprint features and real-time body posture features both exceed the preset abnormal threshold, extract the abnormal time sequence information of abnormal voiceprint and abnormal body posture. The abnormal time sequence information includes the time point of abnormal occurrence, duration and trend of change. S43. Assign association weights to abnormal time-series information based on the physiological correlation patterns of the target animal's current growth stage. S44. Based on the aforementioned correlation weights, calculate the temporal correlation degree between real-time voiceprint feature anomalies and real-time body feature anomalies, and obtain the correlation score between real-time voiceprint features and real-time body features in the time dimension.

[0050] In this embodiment, the core is to quantify the temporal correlation between real-time voiceprints and body posture abnormalities through "anomaly detection - temporal extraction - weight assignment - correlation calculation," thereby avoiding misjudgment based on a single feature anomaly. Specifically, the following is a detailed explanation: Step S41: The core of this step is to initially screen scenarios where both features are abnormal, thus defining the scope for subsequent time-series correlation analysis. The preset anomaly threshold refers to a pre-defined critical value for judging whether a feature is abnormal (e.g., set to 1 based on baseline library characteristics; a deviation ≥1 indicates an anomaly). Separate judgments refer to independently verifying whether the deviation of voiceprint features (e.g., 0.74 obtained in S34) and body posture features (e.g., 0.38 obtained in S34) exceed this threshold. If only one feature exceeds the threshold, no further time-series analysis is required; only if both features exceed the threshold is the next step triggered (e.g., if both voiceprint deviation 2.0 and body posture deviation 2.75 are ≥1, it is judged as a double-feature anomaly).

[0051] Step S42: The core of this step is to extract the temporal dimension information of dual-feature anomalies to provide data support for correlation analysis. The anomaly time-series information refers to key data that reflects the temporal patterns of the anomalies, specifically including three categories: Anomaly occurrence time: The time when the real-time feature deviation first exceeds the preset threshold (e.g., voiceprint anomaly begins at 8:00, body posture anomaly begins at 8:10). Duration: The time during which the feature deviation remains above the threshold (e.g., abnormal voiceprint lasts for 40 minutes, abnormal body posture lasts for 35 minutes). Trends: The increase or decrease of deviation during abnormal periods (e.g., voiceprint deviation gradually increased from 2.0 to 2.5, and body posture deviation increased from 2.75 to 3.0, both showing an upward trend).

[0052] Step S43: The core of this step is to assign differentiated importance to temporal information by combining the physiological characteristics of animal growth stages. Specifically, the physiological correlation pattern of the current growth stage refers to the natural correlation strength between voiceprints and abnormal body postures at different growth stages (e.g., respiratory diseases in piglets often show a sharp voiceprint (abnormality) appearing first, followed by reduced body activity (abnormality); digestive diseases in the fattening stage often show abnormal body postures appearing first, followed by abnormal voiceprints). The correlation weight refers to the importance coefficient set for temporal information based on this pattern (e.g., the correlation weight for abnormalities occurring within a time difference of ≤30 minutes in the piglet stage is set to 0.6, and the correlation weight for the same time difference in the fattening stage is set to 0.4, matching the physiological characteristics of different stages).

[0053] Step S44: The core of this step is to calculate the final temporal correlation score based on the correlation weight, quantifying the temporal correlation of the two-feature anomalies. The temporal correlation refers to the comprehensive value calculated using formulas such as "time difference of anomaly occurrence × correlation weight + consistency of change trend × correlation weight" (e.g., a 10-minute time difference in the piglet period × 0.6 = 0.06, consistent change trend × 0.4 = 0.04, temporal correlation = 0.1, then proportionally converted to a score in the range of 0 to 1). The correlation score refers to the final output result reflecting the strength of the temporal correlation between the two-feature anomalies (e.g., a score of 0.8 obtained from the above calculation; a higher score indicates that the anomalies are more likely to be caused by the same health problem).

[0054] In one embodiment, step S50 specifically includes the following steps: S51. Real-time acquisition of ambient noise intensity, and determination of the current ambient noise level based on preset noise classification standards; S52. Dynamically allocate the fusion weights of real-time voiceprint features and real-time body posture features based on noise level. The higher the noise level, the lower the fusion weight of real-time voiceprint features and the higher the fusion weight of real-time body posture features. S53. The deviation of the voiceprint feature, the deviation of the body posture feature, and the temporal correlation score are weighted and calculated with the fusion weight to obtain a comprehensive health status assessment value; S54. Based on the preset range in which the comprehensive health status assessment value falls, output the corresponding animal health status assessment result.

[0055] In this embodiment, the core process involves "noise acquisition and grading - dynamic weight allocation - multi-index weighted calculation - interval matching output," combined with environmental noise optimization feature fusion, to output accurate health assessment results, as detailed below: Step S51: The core of this step is to obtain the current ambient noise level to provide a basis for weight adjustment. This includes: real-time acquisition of ambient noise intensity (decibels (dB)) using a noise sensor (such as a noise detection module attached to a microphone array); preset noise classification standards (e.g., low noise: ≤50dB, medium noise: 51~70dB, high noise: >70dB); and determining the current ambient noise level by comparing the real-time acquired decibel value with the classification standards to determine the appropriate level (e.g., if 65dB is detected, it is determined to be medium noise).

[0056] Step S52: The core of this step is to dynamically adjust the fusion weight of the two features based on the degree of noise interference with the voiceprint features. The fusion weight refers to the calculated proportion of voiceprint and body posture features in the final evaluation (the total weight is 1). The logic behind lower voiceprint weight for higher noise levels is that noise can distort voiceprint features (e.g., voiceprint signals are easily masked under high noise), while body posture features are less affected by noise. Therefore, it is necessary to reduce the weight of distorted features and increase the weight of reliable features (e.g., low noise: voiceprint 0.6, body posture 0.4; medium noise: voiceprint 0.5, body posture 0.5; high noise: voiceprint 0.3, body posture 0.7).

[0057] Step S53: The core of this step is to integrate multi-dimensional indicators (voiceprint deviation, body posture deviation, and temporal correlation score) to calculate a comprehensive quantitative value of health status. The weighted calculation refers to the calculation using the formula: "voiceprint feature deviation × voiceprint fusion weight + body posture deviation × body posture fusion weight + temporal correlation score × fixed weight (e.g., 0.2, to enhance the reference value of temporal correlation)" (e.g., voiceprint deviation 2.0 × 0.5 + body posture deviation 2.75 × 0.5 + temporal score 0.8 × 0.2 = 1.0 + 1.375 + 0.16 = 2.535). The comprehensive health status assessment value refers to the final output, a quantitative result reflecting the animal's health risk (the higher the value, the higher the health risk).

[0058] Step S54: The core of this step is to transform the comprehensive assessment value into an intuitive conclusion about health status. The preset interval refers to a pre-defined range of assessment values ​​corresponding to different health levels (e.g., <0.3: healthy, 0.3~0.7: mild abnormality, ≥0.7: severe abnormality); the output of the corresponding assessment result means matching the comprehensive assessment value with the preset interval and outputting a clear conclusion (e.g., a comprehensive value of 2.535 ≥ 0.7 indicates severe abnormality, suggesting a possible respiratory disease requiring urgent examination), providing a direct basis for subsequent intervention.

[0059] In one embodiment, before step S50, the following step is further included: S401. If any one or both of the voiceprint feature deviation and body posture feature deviation exceed the preset anomaly threshold, and the temporal correlation score is lower than the preset correlation threshold, anomaly verification is triggered. S402. Backtrack the fluctuation range of voiceprint and body characteristics of the current growth stage in the dynamic baseline database, and determine whether the current abnormal characteristics are in the normal fluctuation range that is not related to health at this stage. S403. Retrieve synchronous environmental parameters to determine whether abnormal voiceprint characteristics are caused by sudden noise and whether abnormal body posture characteristics are caused by temporary site interference. S404. Based on the judgment result, if it is confirmed that the abnormal feature is caused by non-health factors, the current abnormal feature is marked as invalid abnormality, and the real-time voiceprint feature and real-time body posture feature are backtracked and re-collected; if it is confirmed that the abnormality is caused by health factors, the current abnormal feature and time series correlation score are retained, and the process proceeds to the next step.

[0060] In this embodiment, the core process involves "identifying suspicious scenarios → backtracking to verify baselines → investigating environmental interference → classifying and processing anomalies" to eliminate misjudgments caused by non-health factors and ensure that the abnormal features entering S50 are genuine health-related signals. Specifically, the process is as follows: Step S401: The core of this step is to accurately trigger suspicious scenarios for anomaly verification, avoiding redundant processing for clearly healthy / clearly abnormal scenarios. Specifically, any one or both of the voiceprint / body posture deviations exceeding a preset anomaly threshold are the basic conditions for the existence of an abnormal signal (e.g., voiceprint deviation 1.2 ≥ 1, body posture deviation 0.8 < 1, or both feature deviations ≥ 1); a temporal correlation score lower than a preset correlation threshold is a key condition for questionable anomaly source (e.g., if the correlation threshold is set to 0.5, a score of 0.3 < 0.5 indicates that the two feature anomalies have no obvious temporal correlation and may be caused by interference); only when both conditions are met simultaneously will subsequent anomaly verification be triggered (if both features exceed the threshold and the correlation score is ≥ 0.5, proceed directly to S50 without verification).

[0061] Step S402: The core of this step is to determine whether the abnormality is a normal physiological fluctuation in the current growth stage through baseline backtracking, excluding unhealthy characteristic changes. Specifically, backtracking the characteristic fluctuation range of the current growth stage in the dynamic baseline database refers to retrieving the "unhealthy normal range" from the baseline parameters for that stage (e.g., the normal fluctuation of the voiceprint frequency of 30-day-old piglets can be widened to 0.8~1.6kHz, slightly wider than the healthy baseline range, covering occasional physiological stress fluctuations); determining whether it is within the unhealthy normal fluctuation range refers to comparing the real-time abnormal characteristics with this range (e.g., a real-time voiceprint frequency of 1.6kHz in piglets, although exceeding the upper limit of the healthy baseline of 1.5kHz, is within the physiological fluctuation range and is therefore determined to be an unhealthy fluctuation).

[0062] Step S403: The core of this step is to investigate environmental interference factors and distinguish between "environment-induced anomalies" and "health-induced anomalies." Specifically, synchronous environmental parameters refer to environmental data that perfectly matches the time of anomaly feature acquisition (such as sudden noise recorded by a noise sensor or site events recorded by monitoring); determining whether a voiceprint anomaly is caused by sudden noise involves comparing the time of the voiceprint anomaly with the time of the sudden noise (e.g., a voiceprint anomaly at 8:00, with a simultaneous 75dB sudden fan noise at 8:00, indicating the voiceprint anomaly is related to noise); determining whether a physical abnormality is caused by temporary site interference involves comparing the time of the physical abnormality with site events (e.g., a physical abnormality at 8:10, with other piglets chasing "Little White" at 8:09, indicating the physical abnormality is related to site interference).

[0063] Step S404: The core of this step is to classify and process abnormal features based on the verification results of the first two steps to ensure the accuracy of subsequent assessments. "Non-health factors" refers to anomalies verified by S402 and S403 as physiological fluctuations or environmental interference (e.g., abnormal voiceprints due to sudden noise, abnormal body posture due to piglets chasing). In this case, it is "marked as an invalid anomaly" and "features are re-collected" to prevent interfering data from entering S50. "Health factors" refers to anomalies not verified as physiological fluctuations or environmental interference (e.g., voiceprint frequency of 1.8kHz exceeding the physiological fluctuation range, and no simultaneous sudden noise). In this case, "abnormal features and time-series correlation scores are retained," and the anomalies proceed normally to S50 for health assessment, ensuring that no real health risks are overlooked.

[0064] In one embodiment, an animal health status monitoring system based on voiceprint and body feature fusion is provided. This animal health status monitoring system based on voiceprint and body feature fusion corresponds to the animal health status monitoring method based on voiceprint and body feature fusion described in the above embodiment. The animal health status monitoring system based on voiceprint and body feature fusion includes: The data acquisition module is used to collect the voiceprint and physical characteristics of the target animal in a healthy and stable state, and to perform anti-interference processing on them to obtain the processed voiceprint and physical characteristics. The baseline library construction module is used to construct a dynamic baseline library corresponding to the target animal based on the processed voiceprint features and body features. The dynamic baseline library is dynamically updated as the animal grows. The feature analysis module is used to collect real-time voiceprint features and body features of the target animal, and compare and analyze the real-time collected voiceprint features and body features with the dynamic baseline database to obtain the feature deviation, which includes voiceprint feature deviation and body feature deviation. The correlation scoring module is used to analyze the temporal correlation between real-time voiceprint features and abnormal body posture features, and obtain a temporal correlation score. The output module is used to dynamically adjust the fusion weights of real-time voiceprint features and body features according to the environmental noise intensity, and output the animal health status assessment results based on the feature deviation degree and temporal correlation score.

[0065] Furthermore, it also includes: The anomaly verification module is used to trigger anomaly verification if any one or both of the voiceprint feature deviation and body posture feature deviation exceed the preset anomaly threshold, and the temporal correlation score is lower than the preset correlation threshold.

[0066] For specific limitations regarding an animal health status monitoring system based on voiceprint and body feature fusion, please refer to the limitations of an animal health status monitoring method based on voiceprint and body feature fusion described above, which will not be repeated here. Each module in the aforementioned animal health status monitoring system based on voiceprint and body feature fusion can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0067] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows. Figure 2 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database is used for data storage, data processing, and data analysis. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for monitoring animal health status based on the fusion of voiceprint and body features.

[0068] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for monitoring animal health status based on the fusion of voiceprint and body features.

[0069] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements a method for monitoring animal health status based on the fusion of voiceprint and body features.

[0070] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0071] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0072] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for monitoring animal health status based on the fusion of voiceprint and body features, characterized in that, Includes the following steps: Voiceprint and physical characteristics of the target animal in a healthy and stable state were collected and subjected to anti-interference processing to obtain processed voiceprint and physical characteristics. Based on the processed voiceprint and body features, a dynamic baseline database corresponding to the target animal is constructed, and the dynamic baseline database is dynamically updated as the animal grows. Real-time voiceprint and body features of the target animal are collected. The real-time voiceprint and body features are compared and analyzed with the dynamic baseline database to obtain the feature deviation, which includes voiceprint feature deviation and body feature deviation. Correlation analysis was performed to determine the temporal correlation between real-time voiceprint features and abnormal body posture features, and a temporal correlation score was obtained. The fusion weights of real-time voiceprint features and body features are dynamically adjusted based on the environmental noise intensity, and the animal health status assessment result is output based on the feature deviation and temporal correlation score.

2. The animal health status monitoring method based on voiceprint and body feature fusion as described in claim 1, characterized in that, The step of constructing a dynamic baseline database corresponding to the target animal based on the processed voiceprint and body features, and dynamically updating the dynamic baseline database as the animal grows, specifically includes the following steps: The processed voiceprint features are parametrically extracted to obtain voiceprint baseline parameters, which include the normal voiceprint frequency range, voiceprint energy fluctuation range, and abnormal voiceprint determination threshold of the target animal in a healthy state. The processed body features are parametrically extracted to obtain body baseline parameters, which include the normal range of limb movement, the range of average daily resting time, and the threshold of muscle tremor frequency in the target animal under healthy conditions. The voiceprint baseline parameters and body posture baseline parameters are associated and bound with the unique identification information of the target animal, and the current growth stage is tagged. The associated voiceprint baseline parameters, body shape baseline parameters, unique identification information of the target animal, and current growth stage label are stored according to a preset data structure to obtain a dynamic baseline library that is only adapted to the target animal.

3. The animal health status monitoring method based on voiceprint and body feature fusion as described in claim 2, characterized in that, The step of collecting real-time voiceprint and body shape features of the target animal, and comparing and analyzing the real-time collected voiceprint and body shape features with the dynamic baseline database to obtain the feature deviation, specifically includes the following steps: The system acquires voiceprint signals and body images of the target animal in real time and performs anti-interference processing to extract real-time voiceprint features and body features, and obtains information on the current growth stage of the target animal. The voiceprint baseline parameters that match the current growth stage are retrieved from the dynamic baseline library. The real-time voiceprint features are compared with the voiceprint baseline parameters of this stage in different dimensions. The voiceprint dimension deviation results are obtained by combining the physiological weights of each dimension feature in the voiceprint baseline parameters in the current growth stage. The body posture baseline parameters that match the current growth stage are retrieved from the dynamic baseline library. The real-time body posture features are compared with the body posture baseline parameters of this stage in terms of dimensions. The body posture dimension deviation results are obtained by combining the physiological weights of each dimension feature in the body posture baseline parameters in the current growth stage. Cross-validation was performed on the deviation results of voiceprint dimension and body shape dimension, and the deviation results of voiceprint feature and body shape feature were fused to obtain the deviation degree of voiceprint feature and body shape feature that are suitable for the current growth stage.

4. The animal health status monitoring method based on voiceprint and body feature fusion as described in claim 3, characterized in that, The step of analyzing the temporal correlation between real-time voiceprint features and abnormal body posture features to obtain a temporal correlation score specifically includes the following steps: Based on the deviation of the voiceprint feature and the deviation of the body feature, it is determined whether the deviation of the real-time voiceprint feature and the real-time body feature both exceed the preset abnormal threshold. When the deviation of real-time voiceprint features and real-time body posture features both exceed the preset abnormal threshold, abnormal time sequence information of abnormal voiceprints and abnormal body postures is extracted. The abnormal time sequence information includes the time point of abnormal occurrence, duration and trend of change. Based on the physiological correlation patterns of the target animal's current growth stage, abnormal time-series information is assigned correlation weights. Based on the aforementioned correlation weights, the temporal correlation degree between real-time voiceprint feature anomalies and real-time body feature anomalies is calculated, resulting in a correlation score between real-time voiceprint features and real-time body features in the time dimension.

5. The animal health status monitoring method based on voiceprint and body feature fusion as described in claim 1, characterized in that, The step of dynamically adjusting the fusion weights of real-time voiceprint features and body features based on environmental noise intensity, and outputting animal health status assessment results based on the feature deviation and temporal correlation, specifically includes the following steps: The ambient noise intensity is collected in real time, and the current ambient noise level is determined according to the preset noise classification standard. The fusion weights of real-time voiceprint features and real-time body features are dynamically assigned based on noise level. The higher the noise level, the lower the fusion weight of real-time voiceprint features and the higher the fusion weight of real-time body features. The comprehensive health status assessment value is obtained by weighting the voiceprint feature deviation, body posture feature deviation, and temporal correlation score with the fusion weight. Based on the preset range in which the comprehensive health status assessment value falls, the corresponding animal health status assessment result is output.

6. The animal health status monitoring method based on voiceprint and body feature fusion as described in claim 1, characterized in that, Before the step of dynamically adjusting the fusion weights of real-time voiceprint features and body features based on environmental noise intensity, and outputting the animal health status assessment result based on the feature deviation and temporal correlation score, the following steps are also included: If any one or both of the voiceprint feature deviation and body posture feature deviation exceed the preset anomaly threshold, and the temporal correlation score is lower than the preset correlation threshold, anomaly verification is triggered. By retrospectively examining the fluctuation range of voiceprint and physical characteristics at the current growth stage in the dynamic baseline database, it can be determined whether the current abnormal characteristics are within the normal fluctuation range that is not related to health at this stage. Retrieve environmental parameters from the same period to determine whether abnormal voiceprint characteristics are caused by sudden noise and whether abnormal body posture characteristics are caused by temporary site interference. Based on the judgment results, if it is confirmed that the abnormal features are caused by non-health factors, the current abnormal features are marked as invalid abnormalities, and the real-time voiceprint features and real-time body posture features are backtracked and re-collected. If the abnormality is confirmed to be caused by health factors, the current abnormality characteristics and time-series correlation score are retained, and the process proceeds to the next step.

7. An animal health status monitoring system based on voiceprint and body feature fusion, used to implement the steps of the animal health status monitoring method based on voiceprint and body feature fusion as described in any one of claims 1-6, characterized in that, include: The data acquisition module is used to collect the voiceprint and physical characteristics of the target animal in a healthy and stable state, and to perform anti-interference processing on them to obtain the processed voiceprint and physical characteristics. The baseline library construction module is used to construct a dynamic baseline library corresponding to the target animal based on the processed voiceprint features and body features. The dynamic baseline library is dynamically updated as the animal grows. The feature analysis module is used to collect real-time voiceprint features and body features of the target animal, and compare and analyze the real-time collected voiceprint features and body features with the dynamic baseline database to obtain the feature deviation, which includes voiceprint feature deviation and body feature deviation. The correlation scoring module is used to analyze the temporal correlation between real-time voiceprint features and abnormal body posture features, and obtain a temporal correlation score. The output module is used to dynamically adjust the fusion weights of real-time voiceprint features and body features according to the environmental noise intensity, and output the animal health status assessment results based on the feature deviation degree and temporal correlation score.

8. The animal health status monitoring system based on voiceprint and body feature fusion as described in claim 7, characterized in that, Also includes: The anomaly verification module is used to trigger anomaly verification if any one or both of the voiceprint feature deviation and body posture feature deviation exceed the preset anomaly threshold, and the temporal correlation score is lower than the preset correlation threshold.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the animal health status monitoring method based on the fusion of voiceprint and body features as described in any one of claims 1-6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the animal health status monitoring method based on the fusion of voiceprint and body features as described in any one of claims 1-6.

Images