CN-121983276-A - Pediatric nursing intelligent shaking table system based on characteristic data feedback

CN121983276ACN 121983276 ACN121983276 ACN 121983276ACN-121983276-A

Abstract

The invention discloses a pediatric nursing intelligent shaking table system based on characteristic data feedback, which is characterized in that multi-dimensional pediatric data is preprocessed, long-term dependence characteristics in the multi-dimensional pediatric data are extracted by a transducer encoder based on a CLP multi-mode base model, space-time characteristics are extracted by a 3D-CNN, multi-scale acoustic characteristics are generated according to WaveNet, a causal graph of infant state and nursing actions is constructed according to the global state vector by a PC algorithm, causal relations among variables are deduced from historical data, potential results under different nursing interventions are simulated by a BSTS Bayesian structure time sequence, and shaking table action amplitude, temperature control parameters and sound pacifying parameters are dynamically adjusted according to shaking table control strategies. And the nursing efficiency is improved.

Inventors

CHENG SHI
JIANG PENG
SUN XIAOLIANG
ZHANG LIN

Assignees

南通大学

Dates

Publication Date: 20260505
Application Date: 20251208

Claims (10)

1. Pediatric nursing intelligent cradle system based on characteristic data feedback, which is characterized in that the pediatric nursing intelligent cradle system comprises the following modules: The system comprises a multi-dimensional data acquisition module, a data processing module and a data processing module, wherein the multi-dimensional data acquisition module is used for acquiring multi-dimensional pediatric data, and preprocessing the multi-dimensional pediatric data to obtain initial multi-dimensional pediatric data; The multimode feature fusion module is used for extracting long-term dependence features in the multi-dimensional pediatric data by using a transducer encoder based on a CLP multimode base model, extracting space-time features by using a 3D-CNN, and generating multi-scale acoustic features according to WaveNet to obtain multimode feature data; the state vector unifying module is used for aligning the characteristic representations of different modes in the multi-mode characteristic data through multi-mode cross attention to generate a unified global state vector; The cradle strategy generation module is used for constructing a causal graph of infant states and nursing actions according to the global state vector by using a PC algorithm, deducing causal relations among variables from historical data, simulating potential results under different nursing interventions through a BSTS Bayesian structure time sequence, and generating a cradle control strategy; and the dynamic adjusting module of the shaking table is used for dynamically adjusting the action amplitude, the temperature control parameter and the sound pacifying parameter of the shaking table according to the shaking table control strategy.
2. The pediatric care intelligent shaker system based on characteristic data feedback of claim 1, wherein the multi-dimensional data acquisition module comprises the following units: The data acquisition unit is used for acquiring infant physiological data, environment parameters, nursing record data and infant behavior data to obtain multidimensional pediatric data; The data cleaning unit is used for setting a reasonable threshold range of physiological data and environmental parameters and deleting abnormal data points exceeding the threshold range; The interpolation duplicate removal unit is used for checking the missing value and the duplicate value of the nursing record data, wherein the missing value is interpolated by adopting the average value of adjacent time points, and the duplicate value is subjected to duplicate removal treatment; the normalization synchronization unit is used for uniformly converting the data with different dimensions into numerical values within the range of [0,1], and performing time alignment on the physiological data, the environmental parameters and the infant behavior data by taking the operation time in the nursing record data as a reference to obtain initial multidimensional pediatric data.
3. The pediatric care intelligent shaker system based on feature data feedback of claim 1, wherein the multimodal feature fusion module comprises the following elements: The input processing unit is used for arranging the preprocessed initial multidimensional pediatric data in time sequence, and the input vector of each time point is formed by splicing normalized vectors of physiological data, environmental parameters, nursing record data and infant behavior data; The weight calculation unit is used for generating a query vector, a key vector and a value vector of each time point from an input vector through linear transformation in the transducer encoder, calculating the dot product similarity of the query vector and the key vector and obtaining the attention weight; and the attention output unit is used for obtaining a weighted sum by multiplying the value vector after carrying out softmax normalization on the attention weight so as to obtain self-attention output.
4. The pediatric care intelligent shaker system based on feature data feedback of claim 3, wherein the multimodal feature fusion module further comprises the following elements: The 3D convolution operation unit is used for simultaneously carrying out convolution operation on space and time dimensions by utilizing a convolution kernel of the 3D-CNN, and extracting space-time characteristics of video data in the initial multi-dimensional pediatric data; And the characteristic output unit is used for flattening the output of the last layer into a one-dimensional vector after the multi-layer 3D convolution and pooling operation and outputting the space-time characteristic representation of the infant behavior data.
5. The pediatric care intelligent shaker system based on characteristic data feedback of claim 1, wherein the state vector unification module comprises the following elements: The multi-modal feature representation unit is used for unifying the space-time features extracted by the long-term dependence features 3D-CNN extracted by the transducer encoder and the multi-scale acoustic features generated by WaveNet in dimension, and converting the space-time features into feature vectors with the same dimension through linear transformation to obtain multi-modal feature data; the inter-mode attention calculating unit is used for calculating the cross attention between every two modes of the multi-mode feature data, enabling the features of each mode to be focused on feature representation through a multi-layer cross attention mechanism, and obtaining cross attention output; And the global state vector generation unit is used for fusing the cross attention output in a weighted summation mode to generate a unified global state vector.
6. The pediatric care intelligent shaker system based on characteristic data feedback of claim 1, wherein the shaker strategy generation module comprises the following units: the independence test unit is used for carrying out condition independence test through chi-square test, deleting edges which do not meet the condition independence, and constructing a causal graph structure between the infant state and the nursing action; And the causal relation deducing unit is used for deducing causal relation strength among variables from the historical data according to the constructed causal diagram, and estimating the weight of each side in the causal diagram by adopting regression analysis to obtain a target causal diagram structure.
7. The pediatric care intelligent shaker system based on characteristic data feedback of claim 1, wherein the shaker strategy generation module further comprises the following units: the potential result simulation unit is used for simulating the infant state change trend under different nursing interventions through a BSTS model on the basis of the target causal graph structure; the control strategy generation unit is used for screening an optimal nursing intervention scheme from the simulated potential results by using the particle swarm optimization algorithm with the optimization of the infant state as a target, and generating a shaking table control strategy.
8. A method of implementing a pediatric care intelligent shaker system based on characteristic data feedback as defined in claim 1, the method comprising the steps of: collecting multi-dimensional pediatric data, and preprocessing the multi-dimensional pediatric data to obtain initial multi-dimensional pediatric data; based on a CLP multi-mode base model, extracting long-term dependence characteristics in the multi-dimensional pediatric data by using a transducer encoder, extracting space-time characteristics by using a 3D-CNN, and generating multi-scale acoustic characteristics according to WaveNet to obtain multi-mode characteristic data; Aligning the characteristic representations of different modes in the multi-mode characteristic data through multi-mode cross attention to generate a unified global state vector; Constructing a causal graph of infant status and nursing actions according to the global status vector by using a PC algorithm, deducing causal relations among variables from historical data, simulating potential results under different nursing interventions through a BSTS Bayesian structure time sequence, and generating a shaking table control strategy; And dynamically adjusting the movement amplitude, the temperature control parameter and the sound pacifying parameter of the shaking table according to the shaking table control strategy.
9. A method of implementing a pediatric care intelligent shaker system based on characteristic data feedback as defined in claim 1, the method comprising the steps of: the condition independence test is carried out through chi-square test, edges which do not meet the condition independence are deleted, and a causal graph structure between the infant state and the nursing action is constructed; And deducing the causal relation strength among the variables from the historical data according to the constructed causal graph, and estimating the weight of each side in the causal graph by adopting regression analysis to obtain a target causal graph structure.
10. A method of implementing a pediatric care intelligent shaker system based on characteristic data feedback as defined in claim 1, the method comprising the steps of: simulating infant state change trends under different nursing interventions through a BSTS model on the basis of a target causal graph structure; And (3) taking the optimization of the infant state as a target, screening an optimal nursing intervention scheme from the simulated potential results through a particle swarm optimization algorithm, and generating a shaking table control strategy.

Description

Pediatric nursing intelligent shaking table system based on characteristic data feedback Technical Field The invention relates to the technical field of intelligent medical equipment, in particular to an intelligent pediatric nursing shaking table system based on characteristic data feedback. Background Traditional paediatrics nursing mode relies on nursing staff's experience and manual operation more, has nursing inefficiency, precision subalternation problem. When judging the cause of crying of an infant, nursing staff often can only guess through limited observation and experience, cannot quickly and accurately find the root, and further take effective nursing measures, so that pediatric nursing efficiency is low. Disclosure of Invention The invention aims to solve the problems, and designs an intelligent pediatric nursing shaking table system based on characteristic data feedback. The technical scheme of the invention for achieving the purpose is that, in the pediatric nursing intelligent shaking table system based on the characteristic data feedback, the pediatric nursing intelligent shaking table system comprises the following modules: The system comprises a multi-dimensional data acquisition module, a data processing module and a data processing module, wherein the multi-dimensional data acquisition module is used for acquiring multi-dimensional pediatric data, and preprocessing the multi-dimensional pediatric data to obtain initial multi-dimensional pediatric data; The multimode feature fusion module is used for extracting long-term dependence features in the multi-dimensional pediatric data by using a transducer encoder based on a CLP multimode base model, extracting space-time features by using a 3D-CNN, and generating multi-scale acoustic features according to WaveNet to obtain multimode feature data; the state vector unifying module is used for aligning the characteristic representations of different modes in the multi-mode characteristic data through multi-mode cross attention to generate a unified global state vector; The cradle strategy generation module is used for constructing a causal graph of infant states and nursing actions according to the global state vector by using a PC algorithm, deducing causal relations among variables from historical data, simulating potential results under different nursing interventions through a BSTS Bayesian structure time sequence, and generating a cradle control strategy; and the dynamic adjusting module of the shaking table is used for dynamically adjusting the action amplitude, the temperature control parameter and the sound pacifying parameter of the shaking table according to the shaking table control strategy. Further, in the pediatric nursing intelligent cradle system based on the characteristic data feedback, the multidimensional data acquisition module comprises the following units: The data acquisition unit is used for acquiring infant physiological data, environment parameters, nursing record data and infant behavior data to obtain multidimensional pediatric data; The data cleaning unit is used for setting a reasonable threshold range of physiological data and environmental parameters and deleting abnormal data points exceeding the threshold range; The interpolation duplicate removal unit is used for checking the missing value and the duplicate value of the nursing record data, wherein the missing value is interpolated by adopting the average value of adjacent time points, and the duplicate value is subjected to duplicate removal treatment; the normalization synchronization unit is used for uniformly converting the data with different dimensions into numerical values within the range of [0,1], and performing time alignment on the physiological data, the environmental parameters and the infant behavior data by taking the operation time in the nursing record data as a reference to obtain initial multidimensional pediatric data. Further, in the pediatric nursing intelligent cradle system based on the feature data feedback, the multi-mode feature fusion module comprises the following units: The input processing unit is used for arranging the preprocessed initial multidimensional pediatric data in time sequence, and the input vector of each time point is formed by splicing normalized vectors of physiological data, environmental parameters, nursing record data and infant behavior data; The weight calculation unit is used for generating a query vector, a key vector and a value vector of each time point from an input vector through linear transformation in the transducer encoder, calculating the dot product similarity of the query vector and the key vector and obtaining the attention weight; and the attention output unit is used for obtaining a weighted sum by multiplying the value vector after carrying out softmax normalization on the attention weight so as to obtain self-attention output. Further, in the pediatric nursing intelligent cradle system base