A method, system and device for monitoring blood pressure in a ruminant

Abstract

The application provides a ruminant blood pressure monitoring method, system and device, the method comprising: synchronously collecting tail acceleration signals, flank acceleration signals, pressure signals, impedance signals and angular velocities of the ruminant, and preprocessing to obtain preprocessed sequences; taking the flank acceleration signals in the preprocessed sequences as reference noise, and denoising the pressure signals and impedance signals to obtain pure pressure signals and pure impedance signals; based on the continuity constraints of zero-order, first-order and second-order, reconstructing the pure pressure signals to obtain reconstructed pressure signals, and weighting and fusing the reconstructed pressure signals and the pure impedance signals to obtain fused pressure signals; based on the angular velocities and tail inclination angles, compensating the fused pressure signals to obtain compensated pressure signals; and extracting feature quantities from the compensated pressure signals, and combining a blood pressure prediction model to obtain blood pressure. The application can suppress four types of unique motion artifacts of ruminants and improve the accuracy of blood pressure prediction.

Description

Technical Field

[0001] This application relates to the field of blood pressure measurement, specifically to a method, system, and device for monitoring blood pressure in ruminants. Background Technology

[0002] Cardiovascular health monitoring in ruminants (primarily dairy cows) is a key technological requirement for large-scale farming and precision veterinary management. Perifelative dairy cows are prone to cardiovascular-related diseases such as hypocalcemia, preeclampsia, and insufficient circulating blood volume. If these diseases are not detected in time, they can lead to decreased milk production, reduced reproductive performance, and even death, causing significant economic losses. Continuous, non-invasive blood pressure monitoring is a core method for assessing cardiovascular status and providing early warning of these diseases.

[0003] In the field of human medicine, continuous blood pressure monitoring technology based on photoplethysmography (PPG) is relatively mature. It utilizes changes in light intensity to reflect the pulsation of superficial arteries, combining pulse transit time or pulse wave morphology characteristics to estimate blood pressure. However, directly transferring this technology to ruminants faces three major obstacles:

[0004] (1) Attenuation barriers in sensing due to fur. Ruminants are covered with a thick coat of fur. The hair density of dairy cows can reach 2000-3000 hairs / cm², and the skin thickness at the base of the tail is 4-8 mm. Traditional PPG optical signals are severely attenuated when passing through the fur, making it difficult to stably acquire pulse wave signals from deep arteries. There is currently no mature solution for non-invasive acquisition of deep arterial pulse waves under conditions of thick fur coverage.

[0005] (2) Specific motion artifacts at the signal level. Ruminants produce four types of specific motion artifacts during their daily activities: Type I is the low-frequency rhythmic vibration of 0.05-0.2 Hz generated by rumen peristalsis; Type II is the instantaneous high-energy tail-flicking impact triggered by fly repelling or stress response; Type III is the regular head and body movements that last for several minutes during feeding and drinking; and Type IV is the large changes in body position caused by the transition from lying down to standing. The frequency characteristics, energy characteristics, and time scale of the above artifacts are significantly different from those of human daily activity artifacts. Existing general motion artifact suppression algorithms developed for humans cannot effectively cover these artifact types, resulting in severe degradation of pulse wave signal quality.

[0006] In summary, there is an urgent need for a blood pressure monitoring method specifically designed for ruminants that can penetrate thick fur to achieve non-invasive acquisition of deep arterial pulse waves and effectively suppress the four types of motion artifacts unique to ruminants, thereby establishing a blood pressure estimation model suitable for the cardiovascular characteristics of ruminants. Summary of the Invention

[0007] This application addresses the shortcomings of existing technologies by providing a method, system, and device for monitoring blood pressure in ruminants.

[0008] To solve the above-mentioned technical problems, this application provides the following technical solution: A method for monitoring blood pressure in ruminants, comprising the following steps: Simultaneously, tail acceleration signals, flank acceleration signals, pressure signals, impedance signals, and angular velocities in the middle tail artery region of ruminants were collected and preprocessed to obtain a preprocessed sequence. Using the flank acceleration signal in the preprocessed sequence as reference noise, the pressure signal and impedance signal within the rumen peristalsis time interval in the preprocessed sequence are denoised to obtain a pure pressure signal sequence and a pure impedance signal sequence; wherein, the rumen peristalsis time interval is identified by the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence. Based on the zeroth-order, first-order, and second-order continuity constraints, the pressure signal within the tail-flip impact time interval in the pure pressure signal sequence is subjected to waveform reconstruction processing to obtain a reconstructed pressure signal sequence, which is then weighted and fused with the pure impedance signal sequence to obtain a fused pressure signal sequence; wherein, the tail-flip impact time interval is identified based on the energy density of the tail acceleration signal in the preprocessed sequence. Based on the angular velocity and tail tilt angle in the preprocessed sequence, pressure compensation is performed on the pressure signal within the body position transition time interval in the fused pressure signal sequence to obtain a compensated pressure signal sequence; wherein, the tail tilt angle and the body position transition time interval are obtained through the tail acceleration signal in the preprocessed sequence; Feature quantities are extracted from the compensated pressure signal sequence and combined with a blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant.

[0009] As one possible implementation, the simultaneous acquisition of tail acceleration signals, flank acceleration signals, and pressure signals, impedance signals, and angular velocities along the course of the middle caudal artery of ruminants, followed by preprocessing to obtain a preprocessed sequence, includes the following steps: Based on time series, tail acceleration signals, flank acceleration signals, and pressure signals, impedance signals, and angular velocities in the middle tail artery region of ruminants were collected to obtain the original sequences; The original sequence is subjected to outlier removal and unified timestamp alignment to obtain a preliminary processed sequence; The pressure signal and impedance signal in the preliminary processing sequence are respectively subjected to bandpass filtering based on a first frequency range, and the muscular acceleration in the preliminary processing sequence is subjected to bandpass filtering based on a second frequency range to obtain a preprocessed sequence. The first frequency range is determined based on the heart rate range of the ruminant, and the second frequency range is determined based on the rumen contraction frequency of the ruminant.

[0010] As one possible implementation, the step of using the flank acceleration signal in the preprocessed sequence as reference noise to denoise the pressure signal and impedance signal within the rumen peristalsis time interval in the preprocessed sequence to obtain a pure pressure signal sequence and a pure impedance signal sequence includes the following steps: Based on the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence, the rumen peristalsis time interval is identified. Using the flank acceleration signal in the preprocessed sequence as reference noise, the normalized least mean square adaptive filtering algorithm is used to denoise the pressure signal in the rumen peristalsis time interval in the preprocessed sequence to obtain a clean pressure signal sequence. Using the flank acceleration signal in the preprocessed sequence as reference noise, a normalized least mean square adaptive filtering algorithm is used to denoise the impedance signal within the rumen peristalsis time interval in the preprocessed sequence, resulting in a clean impedance signal sequence.

[0011] As one possible implementation, the waveform reconstruction processing of the pressure signal within the tail-flip impact time interval in the pure pressure signal sequence based on zero-order, first-order, and second-order continuity constraints to obtain the reconstructed pressure signal sequence includes the following steps: Based on the tail acceleration signal in the preprocessed sequence, the tail-flip impact time interval is identified; The tail-flip impact time interval is extended based on a preset third sliding time window to obtain a constraint time interval; Based on the continuity constraint of the zeroth, first and second derivatives at the boundary of the constrained time interval, and based on the one-to-one correspondence between the pure pressure signal sequence and the reconstructed pressure signal sequence in terms of the zeroth, first and second derivatives, a set of pressure signal equations is constructed. The pressure signal equations are solved to obtain the reconstructed pressure signal sequence.

[0012] As one possible implementation, identifying the tail-flip impact time interval based on the tail acceleration signal in the preprocessed sequence includes the following steps: Based on a preset second sliding time window, the values ​​of the tail acceleration signal in the preprocessed sequence are accumulated to obtain the tail acceleration energy density; Based on the tail acceleration energy density, the rate of change of tail acceleration energy density is obtained; If the rate of change of the tail acceleration energy density at a certain time point is higher than a preset rate of change threshold, and the time for the tail acceleration energy density to return to the baseline level is lower than a first preset time threshold, then it is determined that a tail-flipping impact event has occurred at this time point; wherein, the baseline level is the acceleration energy density of the ruminant when it is at rest. Based on the first preset extension time, the timestamp corresponding to the tail-flip impact event is extended forward and backward to obtain the tail-flip impact time interval.

[0013] As one possible implementation, the pressure compensation of the pressure signal within the body position transition time interval in the fused pressure signal sequence based on the angular velocity and tail tilt angle in the preprocessed sequence to obtain a compensated pressure signal sequence includes the following steps: The tail tilt angle sequence is obtained by using the tail acceleration signal in the preprocessed sequence; The body position transition time interval is obtained by comparing the cumulative angle change with a preset angle change threshold; wherein, the cumulative angle change is the difference between the maximum and minimum values ​​of the tail tilt angle sequence within the fifth sliding time window; The angular velocities in the preprocessed sequence are accumulated to obtain the horizontal plane tilt angle sequence; The tail tilt angle sequence and the horizontal plane tilt angle sequence are weighted and summed to obtain the compensated tail tilt angle sequence; Based on the compensated tail tilt angle sequence, and combined with the baseline height from the cardiac reference plane to the sacrococcygeal joint of the ruminant, a vertical height sequence is obtained; Based on the vertical height sequence and combined with the blood density of the ruminant, the pressure signal in the body position transition time interval of the fused pressure signal sequence is compensated to obtain a compensated pressure signal sequence.

[0014] As one possible implementation, the compensated pressure signal sequence is represented as follows:

[0015] in, Represents the timestamp in the compensated pressure signal sequence Pressure signal at time, Represents the timestamp in the fused pressure signal sequence Pressure signal at time, Represents timestamps in the vertical height sequence At that time, This indicates the blood density of the ruminant. It represents the acceleration due to gravity.

[0016] As one possible implementation, extracting feature quantities from the compensated pressure signal sequence and combining them with a blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant includes the following steps: The first feature of each cardiac cycle is extracted from the compensated pressure signal sequence; wherein the first feature includes pulse wave conduction time, enhancement index, heart rate, relative time delay of diabetic notch, and normalized area under the pulse wave of the cardiac cycle. Based on the first characteristic of each cardiac cycle, combined with the blood pressure prediction model, the systolic and diastolic blood pressure of the ruminant in each cardiac cycle are obtained; The blood pressure prediction model is represented as follows:

[0017] in, Indicates the first Systolic blood pressure per cardiac cycle Indicates the first Diastolic blood pressure per cardiac cycle Indicates the first The pulse wave conduction time of one cardiac cycle Indicates the first The enhancement index of each heart cycle, Indicates the first Heart rate per cardiac cycle Indicates the first The relative time delay of the diphtheria wave notch per cardiac cycle Indicates the first The normalized area under the pulse wave for each cardiac cycle. , , , , , All represent characteristic coefficients. , Both represent bias terms.

[0018] As one possible implementation, determining the characteristic coefficients and the bias term includes the following steps: Based on the second characteristic of a benchmark human, and combined with the ratio between the weight of the ruminant and the weight of the benchmark human, the second characteristic of the ruminant is obtained and used as a priori estimate; wherein, the second characteristic includes pulse wave conduction time, total arterial compliance, diastolic time constant, and mean arterial pressure; While the ruminant was at rest, a reference blood pressure value was measured using a standard oscillometric sphygmomanometer. Simultaneously, tail acceleration signals, flank acceleration signals, pressure signals, impedance signals, and angular velocities at the main monitoring sites were acquired. After processing, a reference compensated pressure signal sequence was obtained. The first feature of each cardiac cycle is extracted from the reference compensated pressure signal sequence, and combined with the reference blood pressure value to construct an observation dataset; Using the prior estimate as the center of the prior distribution and the regularization constraint boundary, and the observed dataset as the likelihood function condition, the feature coefficients and bias terms in the blood pressure prediction model are solved by the maximum a posteriori estimation optimization algorithm.

[0019] A blood pressure monitoring system for ruminants, used to implement the method described in any one of the above embodiments, the system comprising: The data acquisition and processing module is used to simultaneously acquire tail acceleration signals, flank acceleration signals, and pressure signals, impedance signals, and angular velocities in the middle tail artery region of ruminants, and perform preprocessing to obtain a preprocessed sequence. A type I artifact suppression module is used to denoise the pressure and impedance signals within the rumen peristalsis time interval of the preprocessed sequence, using the flank acceleration signal in the preprocessed sequence as reference noise, to obtain a pure pressure signal sequence and a pure impedance signal sequence; wherein, the rumen peristalsis time interval is identified by the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence; The Type II and Type III artifact suppression module is used to perform waveform reconstruction processing on the pressure signal within the tail-flip impact time interval in the pure pressure signal sequence based on zero-order, first-order, and second-order continuity constraints, to obtain a reconstructed pressure signal sequence, and then perform weighted fusion with the pure impedance signal sequence to obtain a fused pressure signal sequence; wherein, the tail-flip impact time interval is identified based on the energy density of the tail acceleration signal in the preprocessed sequence. A Type IV artifact suppression module is used to perform pressure compensation on the pressure signal within the body position transition time interval in the fused pressure signal sequence based on the angular velocity and tail tilt angle in the preprocessed sequence, to obtain a compensated pressure signal sequence; wherein, the tail tilt angle and the body position transition time interval are obtained from the tail acceleration signal in the preprocessed sequence; The blood pressure prediction module is used to extract feature quantities from the compensated pressure signal sequence and combine them with the blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant.

[0020] A blood pressure monitoring device for ruminants includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the method described in any one of the above.

[0021] This application, by adopting the above technical solution, has significant technical effects: (1) Reliable pulse wave acquisition under thick fur conditions is achieved using multimodal complementary data. This application adopts a dual-modal synchronous acquisition strategy using a pressure sensor and an impedance electrode. The pressure sensor directly contacts the skin to acquire pulse wave pressure signals, while the impedance electrode penetrates the fur layer and keratin layer to measure the impedance changes of deep tissues. The two complement each other naturally, meaning that when the ruminant's body condition score is high, the pressure signal attenuates but the impedance signal is not affected by the fat layer; when the body condition score is low, the pressure sensor can acquire high-quality signals. Dual-modal complementarity ensures stable acquisition of arterial pulse wave signals under different body condition conditions, fundamentally overcoming the technical bottleneck of the failure of the traditional PPG method under thick fur conditions.

[0022] (2) Suppression of four types of motion artifacts specific to ruminants. This application identifies and suppresses four types of motion artifacts specific to ruminants based on preprocessed data. The resulting compensated pressure signal suppresses these four types of motion artifacts, providing a clean data source for subsequent blood pressure prediction. The graded suppression strategy systematically covers all four types of motion artifacts specific to ruminants for the first time, significantly improving the quality and continuity of pulse wave signals in complex behavioral scenarios, thereby improving the accuracy of blood pressure prediction for ruminants. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating an embodiment of the method of this application; Figure 2 This is an overall schematic diagram of an embodiment of the system in this application. Detailed Implementation

[0025] The present application will be further described in detail below with reference to the embodiments. The following embodiments are explanations of the present application, but the present application is not limited to the following embodiments. Unless otherwise specified, the features in the following embodiments can be combined with each other.

[0026] Example 1: In one embodiment, a method for monitoring blood pressure in ruminants, such as Figure 1 As shown, it includes the following steps: S100: Simultaneously acquire tail acceleration signals, flank acceleration signals, and pressure signals, impedance signals, and angular velocities in the middle tail artery region of ruminants, and perform preprocessing to obtain a preprocessed sequence; S200: Using the flank acceleration signal in the preprocessed sequence as reference noise, the pressure signal and impedance signal within the rumen peristalsis time interval in the preprocessed sequence are denoised to obtain a pure pressure signal sequence and a pure impedance signal sequence; wherein, the rumen peristalsis time interval is identified by the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence. S300: Based on zero-order, first-order, and second-order continuity constraints, the pressure signal within the tail-flip impact time interval in the pure pressure signal sequence is subjected to waveform reconstruction processing to obtain a reconstructed pressure signal sequence, which is then weighted and fused with the pure impedance signal sequence to obtain a fused pressure signal sequence; wherein, the tail-flip impact time interval is identified based on the energy density of the tail acceleration signal in the preprocessed sequence. S400: Based on the angular velocity and tail tilt angle in the preprocessed sequence, pressure compensation is performed on the pressure signal within the body position transition time interval in the fused pressure signal sequence to obtain a compensated pressure signal sequence; wherein, the tail tilt angle and the body position transition time interval are obtained through the tail acceleration signal in the preprocessed sequence; S500: Extract feature quantities from the compensated pressure signal sequence and combine them with the blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant.

[0027] The ruminants include cattle, sheep, giraffes, deer, camels, etc. Ruminants possess a rumen with similar functions, and rumen peristalsis affects the measurement of their blood pressure. In this specific embodiment, a dairy cow is used as an example to illustrate the technical solution of this application.

[0028] In this application, the ventral aspect of the cow's tail, specifically the area along the middle caudal artery (within 5-15 cm of the tail root), was selected as the primary monitoring site. This is because the middle caudal artery is the most direct continuation of the abdominal aorta at the tail, and its pulse wave morphology is most similar to that of the central artery. Furthermore, the skin in this area is relatively thin (4-8 mm), and subcutaneous tissue thickness is positively correlated with body condition scores. The left flank (abdominal region, near the projection area of ​​the rumen dorsal sac) was selected as the motion reference signal acquisition site, as this site exhibits the highest coupling sensitivity to rumen peristalsis.

[0029] The multimodal sensing module is fixed to the main monitoring site. The module uses an elastic fabric fixing band (15-25mm wide) with an inner layer of medical-grade silicone anti-slip pads. The array of micro-protrusions on the surface of the silicone pads penetrates the hair layer, reducing the attenuation effect of hair on sensor-tissue coupling. The protrusions have a diameter of 0.5-1mm, a height of 0.3-0.5mm, and a spacing of 1.5-2mm. The preset contact pressure range for fixing tightness is 20-40mmHg, which ensures the coupling stability between the sensor and tissue without obstructing arterial blood flow. The reference acceleration sensing unit is then fixed to the flank using an elastic abdominal binder.

[0030] In another embodiment, in S100, the tail acceleration signal, flank acceleration signal, pressure signal, impedance signal, and angular velocity of the ruminant are simultaneously acquired and preprocessed to obtain a preprocessed sequence, including the following steps: S110: Based on time series, tail acceleration signals, flank acceleration signals, pressure signals, impedance signals, and angular velocities of ruminants are collected to obtain the original sequence.

[0031] The following signals are simultaneously acquired using a multimodal sensing module, with a sampling rate of no less than 256Hz: (1) Pressure signal : This is the signal from the main measurement channel, which acquires the pulse wave pressure signal in mmHg. It is acquired by a piezoresistive pressure sensor array, which is a 3×1 or 5×1 linear array with a unit spacing of 2mm. The array direction is perpendicular to the course of the caudal artery.

[0032] (2) Impedance signal : The signal collected is a local impedance volumetric plethysmography signal from the tail, which is used as an auxiliary channel. The unit is Ω. It is collected by an injectable impedance electrode pair with a spacing of 1-2 cm, arranged along the course of the middle tail artery. The injectable minimally invasive electrode can penetrate the fur and keratin layer of ruminants to measure the impedance changes of deep tissues. The signal attenuation is not affected by the fur.

[0033] (3) Tail angular velocity : Measured by the gyroscope in the multimodal sensing module.

[0034] (4) Tail acceleration : The tail triaxial acceleration vector, The unit is m / s². Data was collected by a MEMS triaxial accelerometer (range ±16g).

[0035] (5) Traumatic acceleration : This is the triaxial acceleration vector of the flank, in m / s². The rumen participates in the speed measurement and sensing unit to collect data synchronously, which is used to record information on rumen peristalsis and body movement. It is a key reference for distinguishing between Class I artifacts (rumen motility) and Class III artifacts (feeding / drinking motion).

[0036] , , as well as All are time-domain signals.

[0037] When a cow has a high body condition score (≥3.5, thick subcutaneous fat), the pressure sensor signal... The attenuation is significant, but the impedance signal... Unaffected by the fat layer; when the cow's body condition score is low (≤3.0), the pressure sensor can acquire high-quality pulse waves. Therefore, in this application, pressure and impedance data are collected simultaneously to form complementary dual-modal data.

[0038] S120: The original sequence is subjected to outlier removal and unified timestamp alignment to obtain a preliminary processed sequence.

[0039] This improves the authenticity of the data and ensures that data from multiple sources are in the same physical time coordinate system, providing high-quality data for subsequent modeling and decision-making.

[0040] S130: The pressure signal and impedance signal in the preliminary processing sequence are respectively subjected to bandpass filtering based on the first frequency range, and the muscular acceleration in the preliminary processing sequence is subjected to bandpass filtering based on the second frequency range to obtain the preprocessed sequence.

[0041] pressure signal Impedance signal Bandpass filtering is performed based on a first frequency range. The first frequency range of the bandpass filter is determined according to the heart rate range of the ruminant. The bandpass filter frequency range should cover the fundamental frequency and its main harmonic components corresponding to the ruminant's heart rate range. This removes very low-frequency baseline drift and high-frequency noise, thus retaining the desired heartbeat signal. For example, the heart rate range of a dairy cow is 40~100 bpm, corresponding to a fundamental frequency of 0.67~1.67 Hz, so the bandpass filter frequency range can be set to 0.3 Hz~5 Hz.

[0042] Acceleration of the groin Preprocessed scrotal acceleration is obtained based on bandpass filtering in the second frequency range. Acceleration of the groin Filtering is performed to extract characteristic signals of rumen peristalsis, which are then used as a reference noise source for adaptive filtering. The second frequency range is determined based on the rumen contraction frequency of ruminants. For example, the rumen contraction frequency of cattle is extremely low, about 1 to 3 times per minute, or 1 / 60Hz to 0.05Hz. In order to ensure that the extracted acceleration signal is purer and more stable, the second frequency range can be set to 0.05 to 0.2Hz.

[0043] The preprocessing sequence includes: a preprocessed pressure signal sequence. Preprocessed impedance signal sequence Preprocessing the tail angular velocity sequence Preprocessing the tail acceleration sequence Preprocessing of shin acceleration sequences Also recorded as .

[0044] In another embodiment, in S200, using the flank acceleration signal in the preprocessed sequence as reference noise, the pressure signal and impedance signal within the rumen peristalsis time interval in the preprocessed sequence are denoised to obtain a pure pressure signal sequence and a pure impedance signal sequence, including the following steps: S210: Based on the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence, the rumen peristalsis time interval is identified.

[0045] Rumen peristalsis in ruminants is a rhythmic contraction of the rumen, manifested as low-frequency rhythmic movement. For example, the frequency of rumen peristalsis in cattle is 0.05Hz~0.2Hz. The mechanical transmission path of rumen peristalsis is through the abdominal wall → spine → to the tail, and finally coupled to the tail sensors.

[0046] (1) Based on the preset frequency range and the first sliding time window, the frequency domain power spectral density of the tail acceleration and shin acceleration signals in the preprocessed sequence is integrated to obtain the tail acceleration power sequence and the shin acceleration power sequence.

[0047] Tail acceleration power sequence And shin acceleration power sequence They are represented as follows:

[0048]

[0049] in, Represents the preprocessed muscular acceleration sequence First sliding time window Power spectral density within, Represents the preprocessed tail acceleration sequence Power spectral density within the first sliding time window, Indicates a preset frequency range, then for dairy cows, , .

[0050] (2) If the power in the tail acceleration power sequence at a certain time point is higher than the preset tail power threshold, and the power in the rumen acceleration power sequence at the same time point is lower than the preset rumen power threshold, then it is determined that there is rumen peristalsis in the current window.

[0051] if Higher than the preset tail power threshold and Below the preset limb power threshold The value indicates that rumen peristalsis was detected in the flank but no strong, synchronous movement of the same frequency occurred in the tail, thus classifying the current window as a period of rumen peristalsis interference. This criterion utilizes the attenuation effect of rumen peristalsis transmitted to the tail. The amplitude of flank vibration is much greater than that of tail-coupled vibration, therefore, there is a significant asymmetry in the power of the two data in the rumen peristalsis frequency band (0.05~0.2Hz).

[0052] (3) Identify all time windows of rumen peristalsis to obtain the time interval of rumen peristalsis.

[0053] S220: Using the flank acceleration signal in the preprocessed sequence as reference noise, the pressure signal in the rumen peristalsis time interval in the preprocessed sequence is denoised using the normalized least mean square adaptive filtering algorithm to obtain a clean pressure signal sequence.

[0054] S230: Using the flank acceleration signal in the preprocessed sequence as reference noise, the impedance signal in the rumen peristalsis time interval of the preprocessed sequence is denoised using the normalized least mean square adaptive filtering algorithm to obtain a clean impedance signal sequence.

[0055] The preprocessed scrotum acceleration sequence As reference noise, the Normalized Least Mean Squares (NLMS) adaptive filtering algorithm is used to filter the preprocessed pressure signal sequence. Real-time noise cancellation is performed to obtain a clean pressure signal. The pure pressure signal was obtained. The process is represented as follows: Filter output: .

[0056] Error signal calculation: .

[0057] Weight vector update: .

[0058] in, Represents a timestamp. Indicates length is The adaptive filter weights are a column vector. This represents the transpose of a vector. Indicates the filter order, usually The range is 32 to 64. Indicates the step size factor. , Represents the regularization constant. It can be set to . This indicates the estimation error.

[0059] This adaptive noise cancellation suppresses Type I artifacts, and the resulting pure pressure signal sequence and pure impedance signal sequence eliminate the influence of Type I artifacts (rumen peristalsis).

[0060] In another embodiment, in S300, based on zero-order, first-order, and second-order continuity constraints, waveform reconstruction processing is performed on the pressure signal within the tail-flip impact time interval in the pure pressure signal sequence to obtain a reconstructed pressure signal sequence, which is then weighted and fused with the pure impedance signal sequence to obtain a fused pressure signal sequence, including the following steps: S310: Based on the tail acceleration signal in the preprocessed sequence, identify the tail-flip impact time interval.

[0061] Tail flicking is an occasional, high-frequency action in ruminants to drive away flies or express discomfort. It is characterized by its extremely short duration (<0.5 seconds) but extremely high instantaneous energy. It is necessary to identify the tail flicking action and its duration to specifically suppress its influence on blood pressure measurements.

[0062] In another embodiment, in S310, identifying the tail-flip impact time interval based on the tail acceleration signal in the preprocessed sequence includes the following steps: (1) Based on the preset second sliding time window, the value of the tail acceleration signal in the preprocessed sequence is accumulated to obtain the tail acceleration energy density.

[0063] Tail flicking is an occasional, high-frequency action in ruminants / cows to drive away flies or express discomfort. It is characterized by its extremely short duration (<0.5 seconds) but extremely high instantaneous energy.

[0064] In motion artifact suppression algorithms for human bodies, the sliding time window used to calculate acceleration energy density is typically set to 0.5 to 1 second. However, the tail-flicking motion of cattle is more rapid than that of humans, therefore a second sliding time window is needed. It should be significantly lower than 0.5 to 1 second; for example, by setting a second sliding time window. The time is 0.3 seconds. Therefore, the tail acceleration energy density within the second sliding time window is... , means as follows:

[0065] in, Represents a timestamp. , , These represent the accelerations in the preprocessed tail acceleration sequence, respectively. Components on the x, y, and z axes.

[0066] (2) Based on the tail acceleration energy density, the tail acceleration energy density change rate is obtained.

[0067] Calculate the tail acceleration energy density using the difference approximation. The first derivative, i.e., the rate of change of tail acceleration energy density. : .

[0068] (3) If the rate of change of the tail acceleration energy density at a certain time point is higher than a preset rate of change threshold, and the time for the tail acceleration energy density to return to the baseline level is lower than a preset time threshold, then it is determined that a tail-flip impact event has occurred at this time point.

[0069] If the rate of change of energy density of tail acceleration The rate of change exceeds a preset threshold at a certain time point. Furthermore, the duration of the high-energy state is less than a first preset time threshold (e.g., 0.5 seconds), which corresponds to the tail acceleration energy density of the timestamp. If the time window for returning to the baseline level is less than 0.5 seconds, a tail-flipping impact event is determined to have occurred. The baseline level is the acceleration energy density of the ruminant at rest, which is a small positive value greater than 0.

[0070] (4) Based on the first preset extension time, the timestamp corresponding to the tail-flip impact event is extended forward and backward to obtain the tail-flip impact time interval.

[0071] The signal intervals before and after the tail-drift impact event are marked based on a first preset extension time (e.g., 0.3 seconds) to obtain the tail-drift impact time interval. Because of the sliding time window Smaller, then This is the instantaneous differential. This criterion utilizes the amplification effect of the derivative on transient states, i.e. The response to shock is much greater than This can significantly improve detection sensitivity and reduce the false negative rate.

[0072] S320: The tail-flip impact time interval is extended based on a preset third sliding time window to obtain a constraint time interval.

[0073] Based on the third sliding time window (For example, 1 second) for the tail-flip impact time interval Extending this, the constraint time interval becomes: .

[0074] S330: Based on the continuity constraint of the zeroth, first and second derivatives at the boundary of the constrained time interval, and based on the correspondence between the pure pressure signal sequence and the reconstructed pressure signal sequence in terms of the zeroth, first and second derivatives, construct a set of pressure signal equations.

[0075] (1) Calculate the velocity and acceleration of the pressure signal in the pure pressure signal sequence respectively to obtain the pure velocity volume recording sequence and the pure acceleration volume recording sequence.

[0076] The velocity plethysmogram (VPG) is the first derivative of the pressure signal; therefore, the pure velocity plethysmogram... The unit is mmHg / s.

[0077] The acceleration plethysmogram (APG) is the second derivative of the pressure signal; therefore, the pure acceleration plethysmogram... The unit is mmHg / s².

[0078] However, in actual discrete implementations, a five-point central difference scheme is used to calculate the velocity volume profile. and acceleration volume recording To ensure numerical stability (sampling interval) ):

[0079] .

[0080] (2) Based on the continuity constraint at the boundary of the constraint time interval based on the zeroth, first and second derivatives, and based on the pure pressure signal sequence With reconstructed pressure signal sequence Based on the one-to-one correspondence between the zeroth, first, and second derivatives, a system of pressure signal equations is constructed. This system of pressure signal equations is expressed as follows: , , indicating zero-order continuity.

[0081] , , indicating first-order continuity.

[0082] , , indicating second-order continuity.

[0083] The smallest order polynomial satisfying the above zero-order, first-order, and second-order continuity constraints is a fifth-degree polynomial (six coefficients, six constraint equations). The continuity constraints at the boundaries of the constraint time interval ensure the smoothness of the reconstructed waveform. It is well known to those skilled in the art that the reconstructed pressure signal sequence here... It is still unknown and remains to be solved.

[0084] S340: Solve the pressure signal equations to obtain the reconstructed pressure signal sequence.

[0085] The reconstructed waveform within the constrained time interval is obtained by solving a system of linear equations, and this reconstructed waveform is used to replace the pure pressure signal sequence. The corresponding segment in the sequence yields the reconstructed pressure signal sequence. This approach ensures the continuity of signal values, slope, and curvature at the boundaries simultaneously. While linear or spline interpolation can guarantee first-order continuity, it cannot guarantee second-order continuity. (Reconstructing the pressure signal sequence) Type II artifacts were eliminated.

[0086] S350: The reconstructed pressure signal sequence is weighted and fused with the pure impedance signal sequence to obtain a fused pressure signal sequence.

[0087] (1) Based on the tail acceleration energy density within the second sliding time window, the feeding and drinking time intervals are identified.

[0088] Lowering the head to feed or drink is a regular, multi-times behavior in ruminants, occurring 4-6 times daily, each lasting 3-8 minutes. During this time, the body remains relatively still but exhibits continuous, moderate-amplitude movements, such as lowering the head, raising the head, and chewing. The key difference between Type III artifacts (feeding or drinking actions) and Type II artifacts (tail-flicking) lies in the fact that Type III artifacts... The amplitude is moderate, far lower than the instantaneous derivative of a tail-flip impact, but The image is consistently above the baseline, meaning it has high energy but is not sharp. This is a typical application of combining integral / cumulative and differential / instantaneous rate of change to distinguish artifact types.

[0089] If the tail acceleration energy density within the sliding time window Higher than the preset energy density threshold If the duration exceeds the second preset time threshold (e.g., 10 seconds), the duration will be recorded as the feeding and drinking time interval.

[0090] (2) Reconstruct the pressure signal sequence of the feeding and drinking time intervals. and pure impedance signal Weighted fusion is performed to fuse the pressure signal sequences.

[0091] For signals during periods of active feeding / drinking, dynamic weighted fusion of pressure and impedance modes is performed, resulting in a fused pressure signal. Represented as: Among them, the weighting coefficient The weighting coefficients are adaptively allocated based on the instantaneous signal-to-noise ratio of the pressure and impedance signals within the local window. Represented as: ,in, and These represent the reconstructed pressure signal sequence. and pure impedance signal Within a sliding time window centered on the timestamp t and with a width of 3 times the cardiac cycle, the ratio of the pulse fundamental frequency component energy to the residual noise energy is calculated.

[0092] This fusion enables soft handover, achieving continuous weight allocation based on information quality between the two channels, rather than using hard handover, thus preventing output jumps. The fused pressure signal sequence eliminates the effects of Type III artifacts (feeding or drinking actions).

[0093] In another embodiment, in S400, based on the angular velocity and tail tilt angle in the preprocessed sequence, pressure compensation is performed on the pressure signal within the body position transition time interval in the fused pressure signal sequence to obtain a compensated pressure signal sequence, including the following steps: S410: Obtain the tail tilt angle sequence through the tail acceleration signal in the preprocessed sequence.

[0094] The lying-up and standing-up movements of ruminants cause significant changes in the height of the tail sensor relative to the heart. In cattle, the height change can reach 50-80 cm, which causes a large drift in the hydrostatic pressure component.

[0095] From the preprocessed tail acceleration sequence The DC component is extracted to obtain the static tail acceleration sequence. : ,in, This represents the preprocessed tail acceleration sequence at timestamp t. In the fourth sliding time window The moving average acceleration over a period of time (e.g., 1 second) can be after the timestamp t. The average over a period of time, or the average over time, or... Average over a period of time.

[0096] Using the projection of gravity onto each axis of the sensor, from the static tail acceleration sequence Extracting the tail tilt angle :

[0097] in, , , These represent the accelerations in the static tail acceleration sequence, respectively. Components on the x, y, and z axes.

[0098] S420: Based on the comparison between the cumulative angle change and the preset angle change threshold, the body position transition time interval is obtained; wherein, the cumulative angle change is the difference between the maximum and minimum values ​​of the tail tilt angle sequence within the fifth sliding time window; The process of changing body position takes approximately 5-15 seconds; therefore, the fifth sliding time window for recognizing body position changes is... It should be much longer than the time required for this process, for example, by setting... The maximum tail tilt angle within the sliding time window is 60 seconds. and minimum tail tilt angle Perform interpolation to obtain the cumulative angle change. :

[0099] If the cumulative angle change Higher than the preset angle change threshold (For example, 30°), then a positional change event is determined to have occurred. The selection is based on behavioral observations of ruminants. For example, the angular change of tail-whipping movements in cattle is typically less than 20°, while the angular change of recumbent-standing transitions is typically greater than 45°. Sufficient discrimination margin should be provided.

[0100] Based on a second preset extension time (e.g., 60 seconds), the timestamps corresponding to the postural change events are extended before and after the events to obtain the postural transition time interval. Blood pressure output is paused during the postural transition time interval, and hydrostatic compensation is activated after the postural position stabilizes.

[0101] S430: Accumulate the angular velocities in the preprocessed sequence to obtain the horizontal plane tilt angle sequence. .

[0102] Horizontal tilt angle sequence Represented as: ,in, This represents the angular velocity measured by the gyroscope in the multimodal sensing module. The initial tilt angle is indicated by the horizontal tilt angle, which is the tilt angle of the tail sensor relative to the horizontal plane.

[0103] S440: The tail tilt angle sequence and the horizontal tilt angle sequence are weighted and summed to obtain the compensated tail tilt angle sequence.

[0104] Compensation for tail camber angle Represented as: ,in, This represents the complementary filter coefficient, which can be set to 0.02~0.05.

[0105] S450: Based on the compensated tail tilt angle vector, and combined with the baseline height from the cardiac reference plane of the ruminant to the sacrococcygeal joint, a vertical height sequence is obtained.

[0106] Vertical height at timestamp t Represented as: ,in, This indicates the baseline height from the cardiac reference plane to the sacrococcygeal joint in ruminants. This indicates the length of the tail segment from the sacrococcygeal joint to the sensor placement point in a ruminant. and Determined based on the specific ruminant. Vertical height This represents the instantaneous vertical height of the sensor relative to the cardiac reference plane at timestamp t.

[0107] S460: Based on the vertical height sequence and combined with the blood density of the ruminant, the pressure signal in the body position transition time interval of the fused pressure signal sequence is compensated to obtain a compensated pressure signal sequence.

[0108] Accelerometers are reliable at low frequencies but have high noise at high frequencies, while gyroscopes are reliable at high frequencies but suffer from integral drift. The two complement each other, ultimately outputting a continuously compensated pressure signal. The compensated pressure signal sequence is then described. It is expressed as follows:

[0109] in, Represents the timestamp in the compensated pressure signal sequence Pressure signal at time, Represents the timestamp in the fused pressure signal sequence Pressure signal at time, Represents timestamps in the vertical height sequence At that time, This represents the blood density of the ruminant animal; for cattle, the blood density can be taken as 1.06. This represents gravitational acceleration. The compensated pressure signal sequence eliminates the effects of Type IV artifacts (postural changes).

[0110] As is known to those skilled in the art, the data in the sequence of this application can be matched one-to-one by timestamps. This application processes the data in the time intervals such as the rumen peristalsis time interval, the tail-flip impact time interval, and the body position transition time interval in the sequence to obtain a new sequence. Data that does not need to be processed remains unchanged and is retained in the new sequence.

[0111] In ruminants, especially cattle, the tail is always higher than the heart reference plane, resulting in a consistently positive and large hydrostatic pressure deviation, meaning compensation must be performed continuously. Furthermore, pressure deviation measured at the human wrist... It can be positive or negative because the arm can be above or below the heart. However, for measurements taken from the tail of ruminants... The hydrostatic pressure correction in ruminants is always positive because the tail is consistently above or near the heart level. Therefore, the hydrostatic pressure correction direction in ruminants is unidirectional, always downward. This structural constraint makes the correction algorithm far more stable and reliable than in human bidirectional correction scenarios. This allows for continuous integral tracking and compensation of hydrostatic pressure, enabling real-time correction even when ruminants slowly change tail position, such as gradually raising or lowering it without violent tail flicking, without waiting for the body position to stabilize.

[0112] In another embodiment, in S500, extracting feature quantities from the compensated pressure signal sequence and combining them with a blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant includes the following steps: S510: Extract a first feature quantity for each cardiac cycle from the compensated pressure signal sequence; wherein the first feature quantity includes pulse wave conduction time, enhancement index, heart rate, relative time delay of dicrotic notch, and normalized area under the pulse wave waveform of the cardiac cycle.

[0113] From the compensation pressure signal sequence Extract compensated velocity volumetric recording Compensated acceleration volumetric recording During each cardiac cycle, from , as well as Extract the following features: (1) Pulse wave conduction time .

[0114] Pulse wave conduction time in the kth cardiac cycle The maximum compensated velocity volumetric recording can be performed within the same cardiac cycle without synchronized ECG acquisition. Instead of estimation, then , This represents the pulse pressure during the k-th cardiac cycle. .

[0115] (2) Enhancement Index .

[0116] Enhancement Index It is a dimensionless number. ,in, Indicates compensation pressure signal The first pressure peak during the k-th cardiac cycle. Indicates compensation pressure signal The pressure value corresponding to the dicrotic notch during the k-th cardiac cycle, where the dicrotic notch refers to the pressure value recorded by the compensating acceleration volumic plethysmography. The position of the C wave (second positive peak) in the middle.

[0117] In ruminants, the dicrotic wave characteristic is weaker than in humans because the reflected wave intensity in cattle is lower. Only 1 / 3 to 1 / 2 the size of a human. When the dicrotic notch within the cardiac cycle is not obvious, i.e., the C wave is unrecognizable ( When this occurs, the cardiac cycle within that time period can be used. Replace the diabetic wave notch at the local minimum location in early diastole.

[0118] (3) Heart rate .

[0119] Identify compensated velocity volumetric recording The zero-crossing point where the pressure changes from negative to positive corresponds to a local minimum, and the corresponding timestamp is marked as [value missing]. ,but This represents the end-diastolic moment of the k-th cardiac cycle.

[0120] The interval of the kth cardiac cycle is: Cycle duration .

[0121] Heart rate , This represents the duration of the k-th cardiac cycle.

[0122] (4) Relative time delay of diphtheria wave notch .

[0123] relative time delay of diphtheria wave notch This refers to the delay of the diphtheria wave notch relative to the contraction peak. ,in, This indicates that the diphtheria wave notch is in compensation for the pressure signal. The time position on It indicates the time position of the systolic peak within the same cardiac cycle.

[0124] Consistent with the previous text, if the diabetic wave notch is not obvious during this cardiac cycle, then utilize... The minimum position replaces the diabetic wave notch.

[0125] (5) Area under the normalized cardiac cycle pulse wave .

[0126] First, compensate for the pressure signal Duration of a single cardiac cycle Compressed to the dimensionless interval [0,1], where the horizontal axis represents the normalized time. This can eliminate the influence of heart rate differences on the waveform area.

[0127] Time-normalized compensated pressure signal Numerical integration of the pressure curve within a single cardiac cycle, such as using the trapezoidal method, yields the normalized area under the pulse wave waveform. , The unit is mmHg·s.

[0128] S520: Based on the first characteristic of each cardiac cycle, combined with the blood pressure prediction model, the systolic and diastolic blood pressure of the ruminant in each cardiac cycle are obtained.

[0129] The blood pressure prediction model is represented as follows:

[0130] in, Indicates the first Systolic blood pressure per cardiac cycle Indicates the first Diastolic blood pressure per cardiac cycle Indicates the first The pulse wave conduction time of one cardiac cycle Indicates the first The enhancement index of each heart cycle, Indicates the first Heart rate per cardiac cycle Indicates the first The relative time delay of the diphtheria wave notch per cardiac cycle Indicates the first The normalized area under the pulse wave for each cardiac cycle. , , , , , All represent characteristic coefficients. , Both represent bias terms.

[0131] In another embodiment, determining the characteristic coefficients and the bias term includes the following steps: (1) Based on the second characteristic of the benchmark human, and combined with the ratio between the weight of the ruminant and the weight of the benchmark human, the second characteristic of the ruminant is obtained and used as a priori estimate.

[0132] The second characteristic quantity includes pulse wave propagation time. Common artery compliance diastolic time constant and mean arterial pressure .

[0133] Based on the Poiseuille flow equation, continuity equation, and fractal branching index in the West-Brown-Enquist fractal network model Assuming a scaling factor of 0.75–0.80, the following scaling relationship is derived for the cardiovascular system of ruminants: vessel radius Thick blood vessel walls Body length / aortic length Heart rate Pulse wave velocity Pulse conduction time Stroke volume Cardiac output Peripheral resistance Common artery compliance time constant Mean arterial pressure (Approximately constant).

[0134] Using the allometric growth power-law relationship of the West-Brown-Enquist (WBE) fractal network model, and taking cardiovascular parameters of benchmark humans as a reference, combined with the body weight of the target ruminants, the second characteristic of ruminants was calculated. They are represented as follows:

[0135]

[0136]

[0137] (Zero-order approximation, independent of body weight) in, This indicates the weight of a ruminant animal, for example, a dairy cow is 650 kilograms. This indicates the weight of a benchmark human, such as 70 kilograms. , , , These represent the pulse wave conduction time, total arterial compliance, diastolic time constant, and mean arterial pressure, respectively, in a baseline human. The second characteristic of this ruminant is... As a priori estimate, then .

[0138] Prior estimate As the initial center point for subsequent Bayesian updates, it provides the model with a reasonable starting point that conforms to biological principles.

[0139] (2) When the ruminant is in a quiet state, the reference blood pressure value is measured using a standard oscillometric sphygmomanometer, and the tail acceleration signal, flank acceleration signal, pressure signal, impedance signal and angular velocity of the main monitoring site are collected simultaneously. After processing, a reference compensation pressure signal sequence is obtained.

[0140] While ruminants are standing or resting quietly, three interval measurements are taken using a standard oscillometric tail sphygmomanometer, and the average is recorded to obtain a set of reference systolic blood pressure readings. and reference diastolic pressure , and All values ​​are observed true values. The multimodal sensing module synchronously collects 5 minutes of data, and the synchronously collected data is processed according to the method described above to obtain a reference compensated pressure signal sequence.

[0141] (3) Extract the first feature of each cardiac cycle from the reference compensated pressure signal sequence, and construct the observation dataset by combining it with the reference blood pressure value.

[0142] The first characteristic of each cardiac cycle extracted from the reference compensated pressure signal sequence, with the average value as the final first characteristic, includes the mean pulse wave conduction time. Average Enhancement Index Average heart rate Relative time delay of mean diphtheria wave notch and the area under the mean normalized cardiac cycle pulse wave. .

[0143] The observed dataset Z is then represented as: .

[0144] (4) Using the prior estimate as the center of the prior distribution and the regularization constraint boundary, and the observation dataset as the likelihood function condition, the feature coefficients and bias terms in the blood pressure prediction model are solved by the maximum a posteriori estimation optimization algorithm.

[0145] Prior estimation The center of the prior distribution and the regularized constraint boundary are used to observe the dataset. As a condition for the likelihood function, the feature coefficients and bias terms of the blood pressure prediction model that best fits the current physiological characteristics of ruminants are obtained by back-solving the maximum a posteriori estimation optimization algorithm. These are the optimal feature coefficients and optimal bias terms, as shown below:

[0146] in, Represents individualized parameters, including characteristic coefficients. , , , , , and bias terms , , This represents the optimal individualization parameters, including the optimal characteristic coefficients and the optimal bias term. To represent the prior distribution, it is based on A multivariate Gaussian distribution centered at the center. This is the likelihood function. This application requires only 2-3 reference measurements to converge to the optimal individualized parameters. In contrast, existing blind calibration methods typically require 20 to 50 reference measurements, which significantly reduces the reliance on large amounts of training data.

[0147] Example 2: A blood pressure monitoring system for ruminants, such as Figure 2 As shown, an embodiment for implementing the method described in any one of the above methods is provided, wherein the system comprises; The data acquisition and processing module 100 is used to simultaneously acquire the tail acceleration signal, flank acceleration signal, and pressure signal, impedance signal, and angular velocity of the tail artery course area of ​​ruminants, and perform preprocessing to obtain a preprocessed sequence. The Type I artifact suppression module 200 is used to denoise the pressure signal and impedance signal within the rumen peristalsis time interval of the preprocessed sequence, using the flank acceleration signal in the preprocessed sequence as reference noise, to obtain a pure pressure signal sequence and a pure impedance signal sequence; wherein, the rumen peristalsis time interval is identified by the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence; The Type II and Type III artifact suppression module 300 is used to perform waveform reconstruction processing on the pressure signal within the tail-flip impact time interval in the pure pressure signal sequence based on the zero-order, first-order and second-order continuity constraints, to obtain the reconstructed pressure signal sequence, and to perform weighted fusion with the pure impedance signal sequence to obtain the fused pressure signal sequence. The Type IV artifact suppression module 400 is used to perform pressure compensation on the pressure signal within the body position transition time interval in the fused pressure signal sequence based on the angular velocity and tail tilt angle in the preprocessed sequence, to obtain a compensated pressure signal sequence; wherein, the tail tilt angle and the body position transition time interval are obtained from the tail acceleration signal in the preprocessed sequence; The blood pressure prediction module 500 is used to extract feature quantities from the compensated pressure signal sequence and combine them with the blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant.

[0148] All changes and modifications made without departing from the spirit and scope of this application, and all equivalent technical solutions, also fall within the scope of this application.

[0149] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0150] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] This application is described with reference to flowchart illustrations and / or block diagrams of the method, terminal device (system), and computer program product according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0152] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0153] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0154] It should be noted that: The phrase "an embodiment" or "an embodiment" used in this specification means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of this application. Therefore, the phrase "an embodiment" or "an embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment.

[0155] Furthermore, it should be noted that the shapes and names of the parts and components described in the specific embodiments described in this specification may differ. All equivalent or simple variations made to the structure, features, and principles described in this application are included within the scope of protection of this application. Those skilled in the art to which this application pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, as long as they do not deviate from the structure of this application or exceed the scope defined by the claims, all of which should fall within the scope of protection of this application.

Claims

1. A method for monitoring blood pressure in ruminants, characterized in that, Includes the following steps: Simultaneously, tail acceleration signals, flank acceleration signals, pressure signals, impedance signals, and angular velocities in the middle tail artery region of ruminants were collected and preprocessed to obtain a preprocessed sequence. Suppressing Type I artifacts: Using the flank acceleration signal in the preprocessed sequence as reference noise, the pressure and impedance signals within the rumen peristalsis time interval in the preprocessed sequence are denoised to obtain a clean pressure signal sequence and a clean impedance signal sequence; wherein, the rumen peristalsis time interval is identified by the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence; the Type I artifacts are caused by rumen peristalsis; Suppressing Type II and Type III artifacts: Based on zero-order, first-order, and second-order continuity constraints, the pressure signal within the tail-flip time interval of the pure pressure signal sequence is subjected to waveform reconstruction processing to obtain a reconstructed pressure signal sequence, which is then weighted and fused with the pure impedance signal sequence to obtain a fused pressure signal sequence; wherein, the tail-flip time interval is identified based on the energy density of the tail acceleration signal in the preprocessed sequence; Type II artifacts are caused by tail-flip impacts, and Type III artifacts are caused by feeding or drinking actions; Suppressing Type IV artifacts: Based on the angular velocity and tail tilt angle in the preprocessed sequence, pressure compensation is performed on the pressure signal within the body position transition time interval in the fused pressure signal sequence to obtain a compensated pressure signal sequence; wherein, the tail tilt angle and the body position transition time interval are obtained through the tail acceleration signal in the preprocessed sequence; the Type IV artifacts are caused by changes in body position; Feature quantities are extracted from the compensated pressure signal sequence and combined with a blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant.

2. The method according to claim 1, characterized in that, The simultaneous acquisition of tail acceleration signals, flank acceleration signals, and pressure, impedance, and angular velocity signals of the middle tail artery region of ruminants, followed by preprocessing to obtain a preprocessed sequence, includes the following steps: Based on time series, tail acceleration signals, flank acceleration signals, and pressure signals, impedance signals, and angular velocities in the middle tail artery region of ruminants were collected to obtain the original sequences; The original sequence is subjected to outlier removal and unified timestamp alignment to obtain a preliminary processed sequence; The pressure signal and impedance signal in the preliminary processing sequence are respectively subjected to bandpass filtering based on a first frequency range, and the muscular acceleration in the preliminary processing sequence is subjected to bandpass filtering based on a second frequency range to obtain a preprocessed sequence. The first frequency range is determined based on the heart rate range of the ruminant, and the second frequency range is determined based on the rumen contraction frequency of the ruminant.

3. The method according to claim 1, characterized in that, The step of using the flank acceleration signal in the preprocessed sequence as reference noise to denoise the pressure signal and impedance signal within the rumen peristalsis time interval in the preprocessed sequence to obtain a pure pressure signal sequence and a pure impedance signal sequence includes the following steps: Based on the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence, the rumen peristalsis time interval is identified. Using the flank acceleration signal in the preprocessed sequence as reference noise, the normalized least mean square adaptive filtering algorithm is used to denoise the pressure signal in the rumen peristalsis time interval in the preprocessed sequence to obtain a clean pressure signal sequence. Using the flank acceleration signal in the preprocessed sequence as reference noise, a normalized least mean square adaptive filtering algorithm is used to denoise the impedance signal within the rumen peristalsis time interval in the preprocessed sequence, resulting in a clean impedance signal sequence.

4. The method according to claim 1, characterized in that, Based on zero-order, first-order, and second-order continuity constraints, waveform reconstruction processing is performed on the pressure signal within the tail-flip impact time interval of the pure pressure signal sequence to obtain a reconstructed pressure signal sequence, including the following steps: Based on the tail acceleration signal in the preprocessed sequence, the tail-flip impact time interval is identified; The tail-flip impact time interval is extended based on a preset third sliding time window to obtain a constraint time interval; Based on the continuity constraint of the zeroth, first and second derivatives at the boundary of the constrained time interval, and based on the one-to-one correspondence between the pure pressure signal sequence and the reconstructed pressure signal sequence in terms of the zeroth, first and second derivatives, a set of pressure signal equations is constructed. The pressure signal equations are solved to obtain the reconstructed pressure signal sequence.

5. The method according to claim 4, characterized in that, The process of identifying the tail-flip impact time interval based on the tail acceleration signal in the preprocessed sequence includes the following steps: Based on a preset second sliding time window, the values ​​of the tail acceleration signal in the preprocessed sequence are accumulated to obtain the tail acceleration energy density; Based on the tail acceleration energy density, the rate of change of tail acceleration energy density is obtained; If the rate of change of the tail acceleration energy density at a certain time point is higher than a preset rate of change threshold, and the time for the tail acceleration energy density to return to the baseline level is lower than a first preset time threshold, then it is determined that a tail-flipping impact event has occurred at this time point; wherein, the baseline level is the acceleration energy density of the ruminant when it is at rest. Based on the first preset extension time, the timestamp corresponding to the tail-flip impact event is extended forward and backward to obtain the tail-flip impact time interval.

6. The method according to claim 1, characterized in that, The step of compensating for pressure signals within the body position transition time interval in the fused pressure signal sequence based on the angular velocity and tail tilt angle in the preprocessed sequence to obtain a compensated pressure signal sequence includes the following steps: The tail tilt angle sequence is obtained by using the tail acceleration signal in the preprocessed sequence; The body position transition time interval is obtained by comparing the cumulative angle change with a preset angle change threshold; wherein, the cumulative angle change is the difference between the maximum and minimum values ​​of the tail tilt angle sequence within the fifth sliding time window; The angular velocities in the preprocessed sequence are accumulated to obtain the horizontal plane tilt angle sequence; The tail tilt angle sequence and the horizontal plane tilt angle sequence are weighted and summed to obtain the compensated tail tilt angle sequence; Based on the compensated tail tilt angle sequence, and combined with the baseline height from the cardiac reference plane to the sacrococcygeal joint of the ruminant, a vertical height sequence is obtained; Based on the vertical height sequence and combined with the blood density of the ruminant, the pressure signal in the body position transition time interval of the fused pressure signal sequence is compensated to obtain a compensated pressure signal sequence.

7. The method according to claim 6, characterized in that, The compensation pressure signal sequence is represented as follows: in, Represents the timestamp in the compensated pressure signal sequence Pressure signal at time, Represents the timestamp in the fused pressure signal sequence Pressure signal at time, Represents timestamps in the vertical height sequence At that time, This indicates the blood density of the ruminant. It represents the acceleration due to gravity.

8. The method according to claim 1, characterized in that, The step of extracting feature quantities from the compensated pressure signal sequence and combining them with a blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant includes the following steps: The first feature of each cardiac cycle is extracted from the compensated pressure signal sequence; wherein the first feature includes pulse wave conduction time, enhancement index, heart rate, relative time delay of diabetic notch, and normalized area under the pulse wave of the cardiac cycle. Based on the first characteristic of each cardiac cycle, combined with the blood pressure prediction model, the systolic and diastolic blood pressure of the ruminant in each cardiac cycle are obtained; The blood pressure prediction model is represented as follows: in, Indicates the first Systolic blood pressure per cardiac cycle Indicates the first Diastolic blood pressure per cardiac cycle Indicates the first The pulse wave conduction time of one cardiac cycle Indicates the first The enhancement index of each heart cycle, Indicates the first Heart rate per cardiac cycle Indicates the first The relative time delay of the diphtheria wave notch per cardiac cycle Indicates the first The normalized area under the pulse wave for each cardiac cycle. , , , , , All represent characteristic coefficients. , Both represent bias terms.

9. The method according to claim 8, characterized in that, Determining the characteristic coefficients and the bias term includes the following steps: Based on the second characteristic of a benchmark human, and combined with the ratio between the weight of the ruminant and the weight of the benchmark human, the second characteristic of the ruminant is obtained and used as a priori estimate; wherein, the second characteristic includes pulse wave conduction time, total arterial compliance, diastolic time constant, and mean arterial pressure; While the ruminant was at rest, a reference blood pressure value was measured using a standard oscillometric sphygmomanometer. Simultaneously, tail acceleration signals, flank acceleration signals, pressure signals, impedance signals, and angular velocities at the main monitoring sites were acquired. After processing, a reference compensated pressure signal sequence was obtained. The first feature of each cardiac cycle is extracted from the reference compensated pressure signal sequence, and combined with the reference blood pressure value to construct an observation dataset; Using the prior estimate as the center of the prior distribution and the regularization constraint boundary, and the observed dataset as the likelihood function condition, the feature coefficients and bias terms in the blood pressure prediction model are solved by the maximum a posteriori estimation optimization algorithm.

10. A blood pressure monitoring system for ruminants, used to implement the method as described in any one of claims 1 to 9, characterized in that, The system includes: The data acquisition and processing module is used to simultaneously acquire tail acceleration signals, flank acceleration signals, and pressure signals, impedance signals, and angular velocities in the middle tail artery region of ruminants, and perform preprocessing to obtain a preprocessed sequence. A type I artifact suppression module is used to denoise the pressure and impedance signals within the rumen peristalsis time interval of the preprocessed sequence, using the flank acceleration signal in the preprocessed sequence as reference noise, to obtain a pure pressure signal sequence and a pure impedance signal sequence; wherein, the rumen peristalsis time interval is identified by the frequency domain power spectral density of the tail acceleration and flank acceleration signals in the preprocessed sequence; The Type II and Type III artifact suppression module is used to perform waveform reconstruction processing on the pressure signal within the tail-flip impact time interval in the pure pressure signal sequence based on zero-order, first-order, and second-order continuity constraints, to obtain a reconstructed pressure signal sequence, and then perform weighted fusion with the pure impedance signal sequence to obtain a fused pressure signal sequence; wherein, the tail-flip impact time interval is identified based on the energy density of the tail acceleration signal in the preprocessed sequence. A Type IV artifact suppression module is used to perform pressure compensation on the pressure signal within the body position transition time interval in the fused pressure signal sequence based on the angular velocity and tail tilt angle in the preprocessed sequence, to obtain a compensated pressure signal sequence; wherein, the tail tilt angle and the body position transition time interval are obtained from the tail acceleration signal in the preprocessed sequence; The blood pressure prediction module is used to extract feature quantities from the compensated pressure signal sequence and combine them with the blood pressure prediction model to obtain the systolic and diastolic blood pressure of the ruminant.

11. A blood pressure monitoring device for ruminants, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 9.

Images