Search

CN-121313125-B - Pet-wearing vital sign monitoring method based on flexible sensor

CN121313125BCN 121313125 BCN121313125 BCN 121313125BCN-121313125-B

Abstract

The invention discloses a pet wearing type vital sign monitoring method based on a flexible sensor, which comprises the steps of constructing a flexible multi-mode sensing interface, embedding a sensing layer into a bonding layer and fixing the sensing layer at the neck or chest position of a pet, outputting low-power heat pulse to detect contact characteristics to form a stable bonding interface, collecting multi-mode signals and synchronously correcting the signals, performing filtering and baseline processing to obtain a processed signal set, extracting multi-mode characteristics and constructing physical tensors, calculating vital sign parameters, extracting body temperature heart rate breathing rhythm components, generating a rhythm resonance curve and determining target configuration parameters, adjusting sampling and communication according to the target configuration, triggering alarm when abnormal and uploading the alarm to a remote terminal. The invention realizes high-precision, low-power consumption and long-term stable monitoring of the vital signs of the pets by constructing the flexible multi-mode sensing interface and combining a time sequence reasoning and rhythm energy scheduling mechanism driven by physical consistency.

Inventors

  • ZHENG YANG
  • WU XUEFENG
  • ZHAO YANG

Assignees

  • 合肥吉雅其信息科技有限公司

Dates

Publication Date
20260508
Application Date
20251117

Claims (4)

  1. 1. A method for monitoring vital signs of a pet wearing type based on a flexible sensor, comprising the steps of: constructing a flexible multi-mode sensing interface, embedding the flexible multi-mode sensing layer into the flexible attaching layer, and fixing the flexible attaching layer at the neck or chest position of the pet; on the basis of a flexible multi-mode sensing interface, periodically outputting low-power heat pulses by utilizing a micro heat source point array, calculating contact thermal resistance and a thermal diffusion time constant, and controlling a phase change material microcavity array to be locally heated, softened and slightly expanded to form a stable interface when the contact thermal resistance is increased, the biological impedance change rate is reduced or deformation quantity is unevenly distributed; After a stable interface is formed, acquiring a multi-mode time sequence signal, and performing time synchronization, noise filtering and baseline correction on the multi-mode time sequence signal to obtain a processed multi-mode time sequence signal set; extracting features of the multi-mode time sequence signal set, constructing a physical consistency tensor comprising the temperature change rate, the bioimpedance frequency response, the photoelectric waveform gradient and the deformation rate features, inputting the physical consistency tensor into a multi-mode time sequence reasoning model driven by the physical consistency, outputting a fusion feature vector, and calculating vital sign parameters; based on vital sign parameters, extracting low-frequency rhythm components of body temperature, heart rate and respiratory frequency, generating an individual circadian rhythm resonance curve, determining target configuration of sampling frequency, signal processing rate and wireless communication interval, wherein the circadian rhythm resonance curve is used for representing the intensity of rhythm consistency under different predicted time shifting amounts and determining time shifting positions of a resonance intensity sequence and a main peak; According to target configuration, dynamically adjusting the sampling frequency and the wireless communication interval of the flexible multi-mode sensing layer, updating a multi-mode time sequence reasoning model and target configuration parameters according to rhythm prediction errors and energy consumption indexes in multi-period operation, triggering local alarm when the body temperature, heart rate or respiratory rate exceeds a preset threshold range, and uploading the alarm to a remote terminal through a wireless communication module; the construction of the flexible multi-mode sensing interface refers to the integration of a temperature sensing unit, a bioimpedance sensing unit, a deformation sensing unit, a photoelectric sensing unit, a phase change material microcavity array and a micro-heat source lattice on a flexible substrate material to form a flexible multi-mode sensing layer; The forming a stable interface includes: Partitioning the micro heat source point array in the flexible multi-mode sensing interface, sequentially outputting low-power heat pulses in a space-time staggered mode, respectively recording the heating process and the cooling recovery process of each partition and adjacent measuring points, and generating corresponding records of pulse and temperature response; Based on the corresponding records, respectively determining the partition results of the contact thermal resistance and the thermal diffusion time constant of each partition, and rejecting or retesting abnormal or incomplete temperature response to form a temperature response differential time window; Synchronously acquiring the biological impedance change rate and deformation quantity distribution in a temperature response differential time window to form a contact state vector, and fusing the contact state vector with the contact thermal resistance and the thermal diffusion time constant of each partition to generate an interface laminating scoring graph; selecting a target area according to the interface lamination scoring graph, adaptively determining local heating power and heating duration according to the scoring change trend after the previous round of pulse action, and implementing local heating on the phase change material microcavity array of the target area to trigger softening and micro-expansion, and cooling, solidifying and locking the formed interface form after the completion; After interface locking is completed, re-testing is performed again by using a space-time staggered heat pulse sequence, temperature response acquisition, biological impedance change rate and deformation amount synchronous detection and interface laminating score calculation are sequentially performed, and interface laminating scores are updated according to re-testing results, when interface laminating score lifting reaches a preset threshold, the interface is confirmed to reach a stable state, current contact thermal resistance and a thermal diffusion time constant are output, when interface laminating score lifting does not reach the preset threshold, iterative reconstruction of a local microcavity array is continued or is switched to a candidate area with suboptimal scores, and if continuous decrease of the biological impedance change rate or abnormal deformation amount distribution is detected in the re-testing process, heating power is automatically reduced and heating duration is shortened; the output fusion feature vector calculates vital sign parameters, including: Taking a multi-mode time sequence signal set as input, segmenting a temperature channel, a photoelectric channel, a bioimpedance channel, a deformation quantity channel and an inertia channel according to uniform window length and steps, and respectively extracting change rate, waveform gradient, frequency response amplitude, deformation rate and gesture marks to form a multi-dimensional feature set corresponding to time, frequency and channels; Orderly stacking the multi-dimensional feature set according to a channel dimension, a feature operation dimension, a time dimension and a frequency dimension, constructing a physical consistency tensor, converting a contact thermal resistance and a thermal diffusion time constant into a contact quality score, and taking the contact quality score and a contact indication quantity from the biological impedance change rate and deformation quantity distribution as tensor gating parameters; inputting the physical consistency tensor into a multi-mode time sequence reasoning model driven by the physical consistency, wherein the multi-mode time sequence reasoning model consists of three types of structures: The tensor encoder of contact quality gating weights and screens tensors in a channel dimension and a time dimension according to gating parameters, suppresses fragments caused by poor contact and hair shielding and extracts low-rank potential factors; The time sequence causal router is provided with a forward causal branch and a bidirectional association branch, the time sequence causal router aligns the multi-channel characteristics in time by utilizing event marks provided by an inertia channel and a deformation quantity channel, and the causal and association routing proportion is distributed in a self-adaptive mode according to the stability of front and rear adjacent windows; The physical consistency constraint coordinator loads prior constraint between temperature-impedance-deformation and phase consistency rules between photoelectricity-machinery, performs conflict detection and correction on a contact quality gating result and a time sequence causal routing result, and outputs consistency scores for iterative convergence judgment; 2 rounds of reasoning iteration are carried out in a multi-mode time sequence reasoning model driven by physical consistency according to the sequence of coding, routing and coordination, whether the gating weight and the routing proportion need to be updated is judged according to the consistency score and the contact quality score at the end of each round, if the gating weight and the routing proportion do not reach the stability threshold, the sequence is repeated until the stability threshold meets or reaches the preset iteration times, and a fusion feature vector is obtained; and calculating vital sign parameters including body temperature, heart rate, respiratory rate and pulse conduction time according to the fusion feature vector.
  2. 2. The method for monitoring vital signs of a pet wearing based on a flexible sensor according to claim 1, wherein said obtaining a processed multi-modal set of time-series signals comprises: after a stable interface is formed, synchronously acquiring a temperature channel, a photoelectric channel, a bioimpedance channel, a deformation channel and an inertia channel at a uniform sampling frequency to obtain corresponding multi-mode original time sequence data; aligning the time stamps of the multi-mode original time sequence data, completing resampling and paragraph processing under a unified sampling period, and constructing a continuous data window; performing digital denoising processing on the data of each channel, including bandpass or bandstop filtering, power frequency and direct current component suppression, saturation and shearing segment removal and abnormal pulse repair, so as to obtain denoised multi-mode time sequence data; performing baseline correction and amplitude scale unified processing on the denoised multi-mode time sequence data to ensure that the amplitude of each channel is in a comparable range, and establishing a processing result set corresponding to the data window one by one; generating a contact quality score according to the contact thermal resistance and the thermal diffusion time constant, and carrying out weight adjustment and channel selection on time sequence data of the temperature channel, the photoelectric channel, the bioimpedance channel and the deformation channel according to the contact quality score to obtain a processed multi-mode time sequence signal set.
  3. 3. The method for monitoring vital signs of a pet wearing based on a flexible sensor according to claim 1, wherein the determining the target configuration of the sampling frequency, the signal processing rate and the wireless communication interval comprises: Taking the obtained vital sign parameter sequence as input, wherein the vital sign parameter sequence comprises a body temperature time sequence, a heart rate time sequence, a respiratory frequency time sequence and a pulse conduction time sequence, and carrying out sectional processing on the vital sign parameter sequence in a sliding time window with a fixed length; Extracting low-frequency rhythm components from the body temperature time sequence, the heart rate time sequence and the respiratory frequency time sequence in each sliding time window to obtain corresponding rhythm envelopes and phase marks, and aligning the corresponding rhythm envelopes and phase marks with the pulse transmission time sequence to form a rhythm alignment result; constructing a circadian rhythm resonance curve based on a rhythm alignment result; generating a target configuration parameter set according to the resonance intensity sequence and the time shift position of the main peak, wherein the target configuration parameter set comprises sampling frequency, signal processing rate and wireless communication interval, and weighting the contact quality score and the resonance intensity to obtain target configuration in the current time window; And outputting the target configuration, recording the rhythm prediction error and the energy consumption in unit time as feedback indexes in the running process of the cross-time window, and updating the weighting coefficient and the low-frequency rhythm extraction range according to the feedback indexes.
  4. 4. The method for monitoring vital signs of a pet wearing based on a flexible sensor according to claim 1, wherein triggering the local alarm when the body temperature, heart rate or respiratory rate exceeds a preset threshold range and uploading to the remote terminal through the wireless communication module comprises: receiving an output target configuration, wherein the target configuration comprises a sampling frequency reference value, an edge signal processing rate and a wireless communication interval; Uniformly setting all channels of the flexible multi-mode sensing layer according to the target configuration, and respectively setting sampling frequencies of a temperature channel, a photoelectric channel, a bioimpedance channel and a deformation channel to be fixed proportions of sampling frequency reference values, wherein the proportions are preset in the synchronous constraint of the bandwidth of the device and the channels; According to the edge signal processing rate, the length of a processing window and a stepping walking strategy are adjusted, and the reasoning period and the parameter updating period of a multi-mode time sequence reasoning model driven by physical consistency are synchronously adjusted, so that the processing time delay of the whole machine meets the real-time requirement; when any parameter of body temperature, heart rate or respiratory rate exceeds a preset threshold value set, triggering local alarm immediately and reporting immediately, recording triggering time, alarm level and contact quality score and sending with a message; Summarizing the rhythm prediction error and the energy consumption in unit time as feedback indexes in the multicycle operation process, updating the weighting coefficient and the low-frequency rhythm extraction range according to the feedback indexes, synchronously correcting parameters of a physically consistent driving multi-mode time sequence reasoning model and a threshold value for interface reconstruction triggering, solidifying the current target configuration into a new reference configuration when the stability criterion is met, and continuing iterative adjustment until the preset maximum operation period is reached when the stability criterion is not met.

Description

Pet-wearing vital sign monitoring method based on flexible sensor Technical Field The invention relates to the technical field of animal health monitoring and intelligent wearable, in particular to a pet wearable vital sign monitoring method based on a flexible sensor. Background With the continuous increase of the number of pets and the improvement of the consciousness of people on animal health management, the pet vital sign monitoring technology gradually becomes an important research direction in the fields of animal medical treatment and intelligent pet care. At present, various wearable pet monitoring devices are arranged on the market, physiological and motion data are collected mainly through a temperature sensor, a photoelectric sensor or an acceleration sensor, and monitoring of body temperature, heart rate and activity states is achieved. However, these devices often employ rigid structures or single sensing modules, which are difficult to maintain stable fit during pet activity, resulting in large signal fluctuations and poor data reliability. Because the hair on the surface of the pet is thick, the skin is uneven and the movement is frequent, the existing monitoring device is easy to generate the problems of contact falling, signal drift, false triggering and the like in long-term use, and the continuity and the accuracy of monitoring are affected. The existing signal processing method mainly depends on a fixed algorithm or a simple filtering model, has limited data abnormality recognition capability caused by motion artifact, light interference and poor contact, and cannot realize intelligent fusion of cross-mode physiological signals. Especially under the condition of simultaneous acquisition of multichannel signals, physical consistency constraint is lacking among the sensing units, so that deviation exists among temperature, impedance and photoelectric signals in time and energy distribution, and the real physiological state of the pet is difficult to accurately reflect. The traditional system mostly adopts a fixed sampling frequency and a constant energy consumption mode, and can not dynamically adjust a sampling strategy according to the physiological rhythm and the activity state of the animal, so that energy waste and monitoring efficiency are reduced. Therefore, how to provide a pet-worn vital sign monitoring method based on a flexible sensor is a problem to be solved by those skilled in the art. Disclosure of Invention The invention aims to provide a pet wearing type vital sign monitoring method based on a flexible sensor, which comprehensively utilizes a flexible multi-mode sensing technology, a multi-mode time sequence reasoning algorithm driven by physical consistency and a circadian rhythm resonance energy scheduling mechanism, and details the technical flow for realizing high-precision, low-power consumption and stable monitoring of vital signs in a long-term wearing environment of pets. The invention realizes dynamic fitting and self-adaptive correction of a sensor and skin by constructing a thermal-impedance-deformation three-coupling reconstruction interface, realizes cross-modal fusion and physical association constraint of temperature, photoelectricity, bioimpedance and deformation signals by a multi-modal time sequence reasoning model driven by physical consistency, and establishes a self-evolution scheduling strategy of sampling and energy by combining a circadian rhythm resonance mechanism. The method has the advantages of high laminating stability, strong signal anti-interference capability, high monitoring precision, low energy consumption and long-term continuous operation, and can effectively improve the intelligentization and reliability level of health monitoring of pets. According to the embodiment of the invention, the pet wearing type vital sign monitoring method based on the flexible sensor comprises the following steps: constructing a flexible multi-mode sensing interface, embedding the flexible multi-mode sensing layer into the flexible attaching layer, and fixing the flexible attaching layer at the neck or chest position of the pet; on the basis of a flexible multi-mode sensing interface, periodically outputting low-power heat pulses by utilizing a micro heat source point array, calculating contact thermal resistance and a thermal diffusion time constant, and controlling a phase change material microcavity array to be locally heated, softened and slightly expanded to form a stable interface when the contact thermal resistance is increased, the biological impedance change rate is reduced or deformation quantity is unevenly distributed; After a stable interface is formed, acquiring a multi-mode time sequence signal, and performing time synchronization, noise filtering and baseline correction on the multi-mode time sequence signal to obtain a processed multi-mode time sequence signal set; extracting features of the multi-mode time sequence signal set, constructi