Intelligent early warning system and method for cognitive function deterioration in the elderly

Abstract

The present application provides an intelligent early warning system and method for cognitive function deterioration in the elderly, which relates to the field of intelligent data analysis. It uses data analysis technology based on deep learning to embed and semantically associate each data item in the physiological measurement data and the Montreal Cognitive Scale data of the analyzed elderly object, so as to automatically determine whether to generate a cognitive deterioration early warning prompt based on the significant fusion representation features between the physiological state semantic features of the physiological measurement data and the cognitive state semantic features of the cognitive scale data based on principal component query matching. In this way, the semantic relationship between physiology and cognition can be captured and represented more accurately, which helps to enhance the model's ability to identify cognitive deterioration features, thereby improving the accuracy of predicting cognitive impairment.

Description

Technical Field

[0001] The present application relates to the field of intelligent data analysis, and more specifically, to an intelligent early warning system and method for cognitive function deterioration in the elderly. Background Art

[0002] As the global population ages, the importance of health monitoring and management for the elderly becomes increasingly prominent. In particular, early detection of cognitive decline (such as Alzheimer's disease) can help with timely intervention, delay disease progression, and improve patients' quality of life.

[0003] The invention patent with publication number CN111524602B proposes a memory and cognitive function assessment screening and early warning system for the elderly, which collects physiological data and Montreal Cognitive Assessment (MoCA) results of the elderly and integrates these two types of data using an adaptive weighting method. The BP neural network model optimized by genetic algorithms is used to effectively assess the possibility of cognitive impairment in the elderly.

[0004] However, although the genetic algorithm optimized BP neural network model mentioned in the above patent can improve the optimization of model parameters to a certain extent, this method based on traditional machine learning may be difficult to deeply explore the potential complex relationship between human physiological data and cognitive assessment data. That is, although the BP neural network has nonlinear fitting capabilities, when processing high-dimensional and heterogeneous data, its internal feature representation may not be sufficient to reveal the deep interaction between physiological and cognitive states. In addition, the use of adaptive weighting methods to integrate physiological data and cognitive data, although it can adjust the importance of different data sources to a certain extent, is still limited to linear combinations, which may not be enough to fully capture the intricate relationship between the two types of data. Therefore, the fused feature representation may lack a complete reflection of the significant correlation between physiological and cognitive states, resulting in limited accuracy of the model in assessing cognitive impairment. Especially in the detection of early mild cognitive impairment, subtle changes may be ignored, thus affecting the sensitivity and reliability of the early warning system.

[0005] Therefore, an optimized intelligent early warning scheme for cognitive function deterioration in the elderly is desired. Summary of the invention

[0006] In order to solve the above technical problems, this application is proposed.

[0007] According to one aspect of the present application, there is provided an intelligent early warning system for cognitive function deterioration in the elderly, comprising:

[0008] The module for collecting data on the relevant status of the elderly subjects is used to obtain the physiological measurement data and Montreal Cognitive Scale data of the elderly subjects being analyzed;

[0009] A physiological state semantic coding module for elderly subjects, used for embedding and coding each data item in the physiological measurement data to obtain a sequence of physiological measurement data item embedding coding vectors, and then inputting the sequence into a physiological state semantic association encoder to obtain a physiological state semantic coding feature vector for elderly subjects;

[0010] A semantic coding module for cognitive state of elderly subjects, used for embedding and coding each data item in the Montreal Cognitive Scale data to obtain a sequence of cognitive data item embedding coding vectors, and then inputting the sequence into a cognitive state semantic association encoder to obtain a semantic coding feature vector of cognitive state of elderly subjects;

[0011] The elderly object cognitive-physiological state fusion module is used to input the elderly object cognitive state semantic encoding feature vector and the elderly object physiological state semantic encoding feature vector into a significant fusion network based on feature principal component query matching to obtain a sparse significant fusion representation vector of the elderly object cognitive-physiological state;

[0012] The analysis result generation module is used to obtain the analysis result based on the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject, and the analysis result is used to indicate whether a cognitive degradation warning prompt is generated.

[0013] In the above-mentioned intelligent early warning system for cognitive function deterioration in the elderly, the physiological state semantic coding module of the elderly subject includes: an elderly subject physiological measurement data embedding coding unit, which is used to embed and code each data item in the physiological measurement data using a physiological data embedding coding matrix to obtain a sequence of embedded coding vectors of the physiological measurement data items; and an elderly subject physiological state semantic coding feature generation unit, which is used to input the sequence of embedded coding vectors of the physiological measurement data items into a converter-based physiological state semantic association encoder to obtain the physiological state semantic coding feature vector of the elderly subject.

[0014] In the above-mentioned intelligent early warning system for cognitive function deterioration in the elderly, the semantic coding module for the cognitive state of the elderly subjects includes: an elderly subject cognitive data embedding coding unit, which is used to use a cognitive data embedding coding matrix to embed code each data item in the Montreal Cognitive Scale data to obtain a sequence of embedded coding vectors of the cognitive data items; an elderly subject cognitive state semantic coding feature generation unit, which is used to input the sequence of embedded coding vectors of the cognitive data items into a converter-based cognitive state semantic association encoder to obtain the elderly subject cognitive state semantic coding feature vector.

[0015] In the above-mentioned intelligent early warning system for cognitive function degeneration of the elderly, the cognitive-physiological state fusion module of the elderly object includes: a semantic feature standardization unit of cognitive-physiological state of the elderly object, which is used to standardize the semantic coding feature vector of the cognitive state of the elderly object and the semantic coding feature vector of the physiological state of the elderly object to obtain a standardized semantic coding feature vector of the cognitive state of the elderly object and a standardized semantic coding feature vector of the physiological state of the elderly object; a sample covariance matrix calculation unit of cognitive-physiological state of the elderly object, which is used to calculate the sample covariance matrix of the standardized semantic coding feature vector of the cognitive state of the elderly object and the standardized semantic coding feature vector of the physiological state of the elderly object to obtain a sample covariance matrix of the cognitive state of the elderly object and a sample covariance matrix of the physiological state of the elderly object; a principal component feature vector generation unit of cognitive-physiological state of the elderly object, which is used to extract feature vectors based on matrix decomposition on the sample covariance matrix of the cognitive state of the elderly object and the sample covariance matrix of the physiological state of the elderly object to obtain a set of principal component feature vectors of the cognitive state of the elderly object and a set of principal component feature vectors of the cognitive state of the elderly object and a set of principal component feature vectors of the cognitive state of the elderly object. A set of principal component feature vectors of physiological states; a language-behavior maximum approximation query unit, used for inputting the set of principal component feature vectors of cognitive states of elderly subjects and the set of principal component feature vectors of physiological states of elderly subjects into a maximum approximation query matching network to obtain a set of best matching pairs of principal component feature vectors of cognitive states of elderly subjects and principal component feature vectors of physiological states of elderly subjects; a principal component fusion unit for cognitive states of elderly subjects-physiological states, used for inputting the best matching pairs of principal component feature vectors of cognitive states of elderly subjects and principal component feature vectors of physiological states of elderly subjects in the set of best matching pairs of principal component feature vectors of cognitive states of elderly subjects and principal component feature vectors of physiological states of elderly subjects into a semantic fine-grained gated joint module to obtain a set of principal component fusion feature vectors of cognitive states of elderly subjects-physiological states; a sparse significant fusion representation generation unit for cognitive states of elderly subjects-physiological states, used for cascading the set of principal component fusion feature vectors of cognitive states of elderly subjects-physiological states to obtain a sparse significant fusion representation vector of cognitive states of elderly subjects-physiological states.

[0016] In the above-mentioned intelligent early warning system for cognitive function degeneration in the elderly, the semantic feature standardization unit of the cognitive physiological state of the elderly object is used to: respectively calculate the mean and standard deviation of the semantic coding feature vector of the cognitive state of the elderly object to obtain the mean and standard deviation of the semantic coding feature of the cognitive state of the elderly object; after subtracting the semantic coding feature vector of the cognitive state of the elderly object from the mean of the semantic coding feature of the cognitive state of the elderly object by position, the calculated semantic offset vector of the cognitive state of the elderly object and the standard deviation of the semantic coding feature of the cognitive state of the elderly object are divided by position to obtain the standardized semantic coding feature vector of the cognitive state of the elderly object; respectively calculate the mean and standard deviation of the semantic coding feature vector of the physiological state of the elderly object to obtain the mean and standard deviation of the semantic coding feature of the physiological state of the elderly object; after subtracting the semantic coding feature vector of the physiological state of the elderly object from the mean of the semantic coding feature of the physiological state of the elderly object by position, the calculated semantic offset vector of the physiological state of the elderly object and the standard deviation of the semantic coding feature of the physiological state of the elderly object are divided by position to obtain the standardized semantic coding feature vector of the physiological state of the elderly object.

[0017] In the above-mentioned intelligent early warning system for cognitive function deterioration in the elderly, the elderly subject cognitive physiological state sample covariance matrix calculation unit is used to: after multiplying the transpose vector of the standardized elderly subject cognitive state semantic coding feature vector with the standardized elderly subject cognitive state semantic coding feature vector, the obtained standardized elderly subject cognitive state semantic association matrix is ​​divided by the value obtained by subtracting one from the length of the standardized elderly subject cognitive state semantic coding feature vector to obtain the elderly subject cognitive state sample covariance matrix; after multiplying the transpose vector of the standardized elderly subject physiological state semantic coding feature vector with the standardized elderly subject physiological state semantic coding feature vector, the obtained standardized elderly subject physiological state semantic association matrix is ​​divided by the value obtained by subtracting one from the length of the standardized elderly subject physiological state semantic coding feature vector to obtain the elderly subject physiological state sample covariance matrix.

[0018] In the above-mentioned intelligent early warning system for cognitive function deterioration in the elderly, the language-behavior maximum approximation query unit is used to: extract a predetermined principal component feature vector of the cognitive state of the elderly object from the set of principal component feature vectors of the cognitive state of the elderly object; calculate the cosine similarity between the predetermined principal component feature vector of the cognitive state of the elderly object and each principal component feature vector of the physiological state of the elderly object in the set of principal component feature vectors of the physiological state of the elderly object to obtain a set of matching query similarities; and use the principal component feature vector of the physiological state of the elderly object corresponding to the largest matching query similarity in the set of matching query similarities and the predetermined principal component feature vector of the cognitive state of the elderly object as the best matching pair of the predetermined principal component feature vector of the cognitive state of the elderly object and the principal component feature vector of the physiological state of the elderly object.

[0019] In the above-mentioned intelligent early warning system for cognitive function degeneration in the elderly, the principal component fusion unit of the cognitive state-physiological state of the elderly subject is used to: respectively calculate the position difference, position dot product and position addition between the best matching pairs of the principal component feature vector of the cognitive state of the elderly subject and the principal component feature vector of the physiological state of the elderly subject to obtain the principal component difference vector of the cognitive state-physiological state of the elderly subject, the principal component dot product vector of the cognitive state-physiological state of the elderly subject and the principal component sum vector of the cognitive state-physiological state of the elderly subject; cascade the principal component difference vector of the cognitive state-physiological state of the elderly subject, the principal component dot product vector of the cognitive state-physiological state of the elderly subject and the principal component sum vector of the cognitive state-physiological state of the elderly subject, and perform one-dimensional convolution encoding on the principal component difference vector of the cognitive state-physiological state of the elderly subject, so as to obtain the principal component multi-dimensional fusion vector of the cognitive state-physiological state of the elderly subject; perform local window-based maximum pooling processing on the multi-dimensional fusion vector of the principal component of the cognitive state-physiological state of the elderly subject to obtain the principal component fusion feature vector of the cognitive state-physiological state of the elderly subject.

[0020] In the above-mentioned intelligent early warning system for cognitive function deterioration in the elderly, the analysis result generation module is used to: input the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject into the classifier-based cognitive deterioration early warning analysis module to obtain the analysis result, and the analysis result is used to indicate whether a cognitive deterioration early warning prompt is generated.

[0021] According to another aspect of the present application, there is provided an intelligent early warning method for cognitive function deterioration in the elderly, comprising:

[0022] Obtain physiological measurement data and Montreal Cognitive Scale data from the analyzed elderly subjects;

[0023] Embedding and encoding each data item in the physiological measurement data to obtain a sequence of physiological measurement data item embedding encoding vectors, and then inputting the sequence into a physiological state semantic association encoder to obtain a physiological state semantic encoding feature vector of the elderly subject;

[0024] Embedding and encoding each data item in the Montreal Cognitive Scale data to obtain a sequence of cognitive data item embedding encoding vectors, and then inputting the sequence into a cognitive state semantic association encoder to obtain a cognitive state semantic encoding feature vector of the elderly subject;

[0025] Inputting the semantic encoding feature vector of the cognitive state of the elderly subject and the semantic encoding feature vector of the physiological state of the elderly subject into a significant fusion network based on feature principal component query matching to obtain a sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject;

[0026] Based on the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject, an analysis result is obtained, and the analysis result is used to indicate whether a cognitive degradation early warning prompt is generated.

[0027] Compared with the prior art, the intelligent early warning system and method for cognitive function deterioration in the elderly provided by the present application adopts data analysis technology based on deep learning to embed and semantically associate each data item in the physiological measurement data and the Montreal Cognitive Scale data of the analyzed elderly object, so as to automatically judge whether to generate a cognitive deterioration early warning prompt based on the significant fusion representation features between the physiological state semantic features of the physiological measurement data and the cognitive state semantic features of the cognitive scale data based on principal component query matching. In this way, the semantic relationship between physiology and cognition can be captured and represented more accurately, which helps to enhance the model's ability to identify cognitive deterioration features, thereby improving the accuracy of predicting cognitive impairment. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0029] Figure 1 This is a system block diagram of an intelligent early warning system for cognitive function deterioration in the elderly according to an embodiment of the present application.

[0030] Figure 2 This is a data flow diagram of the intelligent early warning system for cognitive function deterioration in the elderly according to an embodiment of the present application.

[0031] Figure 3This is a block diagram of a semantic encoding module for the physiological state of elderly subjects in an intelligent early warning system for cognitive function deterioration in the elderly according to an embodiment of the present application.

[0032] Figure 4 It is a block diagram of the semantic coding module of the cognitive state of the elderly subject in the intelligent early warning system for cognitive function deterioration in the elderly according to the embodiment of the present application.

[0033] Figure 5 This is a block diagram of a cognitive-physiological state fusion module for elderly subjects in an intelligent early warning system for cognitive function deterioration in the elderly according to an embodiment of the present application.

[0034] Figure 6 This is a flow chart of an intelligent early warning method for cognitive function deterioration in the elderly according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0036] In the context of global population aging, health monitoring of the elderly is particularly critical, especially the early identification of cognitive decline. The Chinese patent with publication number CN111524602B introduces an early warning system for evaluating the memory and cognitive function of the elderly. The system analyzes the physiological data of the elderly and the results of the Montreal Cognitive Assessment (MoCA) scale, and combines the adaptive weighting method with the BP neural network model optimized by the genetic algorithm to improve the accuracy of the assessment of cognitive impairment. However, when dealing with complex, high-dimensional heterogeneous data, this method may not fully reveal the deep connection between physiological and cognitive states, and the adaptive weighting method may not be able to fully capture the complex relationship between the data, which may limit the sensitivity and reliability of the model in the detection of early mild cognitive impairment.

[0037] Therefore, in view of the above technical problems, the technical concept of the present application is to obtain the physiological measurement data and Montreal Cognitive Scale data of the analyzed elderly object, and use the data analysis and processing algorithm based on deep learning to embed the data items in the physiological measurement data and the Montreal Cognitive Scale data and perform semantic association, so as to automatically judge whether to generate cognitive deterioration warning prompts based on the significant fusion representation features between the physiological state semantic features of the physiological measurement data and the cognitive state semantic features of the cognitive scale data based on principal component query matching. In this way, the semantic relationship between physiology and cognition can be captured and represented more accurately, which helps to reveal the deeper connection and complex interaction between the data, thereby enhancing the model's ability to identify cognitive deterioration features and improve the accuracy of predicting cognitive impairment, and thus providing a more intelligent cognitive function deterioration warning solution for the elderly.

[0038] Figure 1 This is a system block diagram of an intelligent early warning system for cognitive function deterioration in the elderly according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the intelligent early warning system for cognitive function deterioration in the elderly according to an embodiment of the present application. Figure 1 and Figure 2 As shown, in the intelligent early warning system 100 for cognitive function deterioration of the elderly, it includes: an elderly subject related state data acquisition module 110, which is used to obtain physiological measurement data and Montreal cognitive scale data of the elderly subject to be analyzed; an elderly subject physiological state semantic coding module 120, which is used to embed code each data item in the physiological measurement data to obtain a sequence of physiological measurement data item embedding coding vectors, and then input the sequence into the physiological state semantic association encoder to obtain the elderly subject physiological state semantic coding feature vector; an elderly subject cognitive state semantic coding module 130, which is used to embed code each data item in the Montreal cognitive scale data to obtain cognitive data The sequence of the data item embedded coding vectors is input into the cognitive state semantic association encoder to obtain the semantic coding feature vector of the cognitive state of the elderly object; the cognitive-physiological state fusion module 140 of the elderly object is used to input the semantic coding feature vector of the cognitive state of the elderly object and the semantic coding feature vector of the physiological state of the elderly object into the significant fusion network based on feature principal component query matching to obtain the sparse significant fusion representation vector of the cognitive-physiological state of the elderly object; the analysis result generation module 150 is used to obtain the analysis result based on the sparse significant fusion representation vector of the cognitive-physiological state of the elderly object, and the analysis result is used to indicate whether a cognitive deterioration warning prompt is generated.

[0039] In the embodiment of the present application, the elderly object related state data acquisition module 110 is used to obtain the physiological measurement data and Montreal cognitive scale data of the analyzed elderly object. It should be understood that the physiological measurement data of the analyzed elderly object includes the physiological index data of the elderly object such as height, weight, body temperature, blood pressure, blood sugar, bone density, trace elements, blood oxygen, blood uric acid, total cholesterol, etc., and the Montreal cognitive scale data of the analyzed elderly object mainly includes the cognitive assessment of the visual-spatial execution ability, naming, memory, attention, language fluency, abstract thinking, delayed memory, orientation, etc. of the elderly object. Specifically, the physiological measurement data provides a multi-dimensional perspective on the overall health of the elderly. These indicators can reflect the health status of the elderly object's body in multiple aspects such as metabolism, circulation, and endocrine, and these factors are closely related to cognitive function. For example, cardiovascular health directly affects the blood supply to the brain, which in turn affects cognitive ability; poor management of chronic diseases such as diabetes and hypertension can lead to a decline in cognitive function. Therefore, the integration of these physiological data helps to comprehensively assess the health status of the elderly and identify possible cognitive degradation risks in advance. The indicators in the Montreal cognitive scale target different cognitive fields and can comprehensively reflect the cognitive ability of the elderly. By measuring these cognitive functions, potential cognitive deterioration can be discovered in a timely manner, especially in the early stages. Changes in these indicators are often early signs of cognitive decline. In particular, in a specific implementation of the present application, physiological measurement data can be obtained by measuring the elderly subject being analyzed using a health all-in-one machine; Montreal Cognitive Scale data can be obtained by filling out the elderly subject being analyzed.

[0040] In an embodiment of the present application, the physiological state semantic coding module 120 of the elderly object is used to embed the encoding of each data item in the physiological measurement data to obtain a sequence of physiological measurement data item embedding coding vectors, and then input the sequence into the physiological state semantic association encoder to obtain the physiological state semantic coding feature vector of the elderly object. Accordingly, considering that the physiological measurement data may include multiple types of data, such as heart rate, blood pressure, body temperature, etc., the dimensions and ranges of these data may be different, and each of these data items contains important semantic information. Based on this, in the technical solution of the present application, the physiological data embedding coding matrix is ​​used to embed the encoding of each data item in the physiological measurement data to obtain a sequence of physiological measurement data item embedding coding vectors. That is, the physiological data embedding coding matrix can reduce the dimension of each data item, while retaining important information as much as possible, providing data support for subsequent semantic feature extraction and encoding. Then, in order to further capture and mine the semantic associations about physiological states between different semantic data spans in the sequence of the physiological measurement data item embedded coding vectors, in the technical solution of the present application, the sequence of the physiological measurement data item embedded coding vectors is input into the transformer-based physiological state semantic association encoder to obtain the physiological state semantic coding feature vector of the elderly object. In particular, the transformer model is particularly suitable for processing this type of data, and it can capture long-term dependencies in the data, which is very important for understanding changes in physiological states.

[0041] Specifically, Figure 3 FIG. 1 is a block diagram of a semantic coding module for the physiological state of an elderly subject in an intelligent early warning system for cognitive function deterioration of the elderly according to an embodiment of the present application. Figure 3 As shown, the physiological state semantic coding module 120 for elderly subjects includes: an elderly subject physiological measurement data embedding coding unit 121, which is used to embed and code each data item in the physiological measurement data using a physiological data embedding coding matrix to obtain a sequence of embedded coding vectors of the physiological measurement data items; and an elderly subject physiological state semantic coding feature generating unit 122, which is used to input the sequence of embedded coding vectors of the physiological measurement data items into a converter-based physiological state semantic association encoder to obtain the elderly subject physiological state semantic coding feature vector.

[0042] In an embodiment of the present application, the semantic coding module 130 for the cognitive state of the elderly object is used to embed the encoding of each data item in the Montreal Cognitive Scale data to obtain a sequence of cognitive data item embedding encoding vectors, and then input the sequence into the cognitive state semantic association encoder to obtain the semantic encoding feature vector of the cognitive state of the elderly object. Accordingly, considering that the Montreal Cognitive Scale data may contain different types of scale scores, such as memory, attention, language, etc., and the key features in these data have an important impact on the assessment of cognitive deterioration. Based on this, in the technical solution of the present application, the cognitive data embedding encoding matrix is ​​used to embed the encoding of each data item in the Montreal Cognitive Scale data to mine and extract key feature information, and convert it into a unified numerical representation to obtain a sequence of cognitive data item embedding encoding vectors. Then, considering that cognitive scale data is usually presented in a sequence form, for example, different parts of the Montreal Cognitive Scale (MoCA) may be evaluated in a specific order, and there are interrelated influences between different data spans. The transformer model can effectively process such sequence data and capture the time dependency and order relationship in the data. Therefore, in the technical solution of the present application, the sequence of the cognitive data items embedded in the coding vector is input into the transformer-based cognitive state semantic association encoder to obtain the cognitive state semantic coding feature vector of the elderly object. In particular, the transformer-based cognitive state semantic association encoder is essentially a Transformer model.

[0043] Specifically, Figure 4 FIG. 1 is a block diagram of a semantic coding module for cognitive status of elderly subjects in an intelligent early warning system for cognitive function deterioration of the elderly according to an embodiment of the present application. Figure 4 As shown, the cognitive state semantic coding module 130 for elderly subjects comprises: an elderly subject cognitive data embedding coding unit 131, which is used to embed and code each data item in the Montreal Cognitive Scale data using a cognitive data embedding coding matrix to obtain a sequence of embedded coding vectors of the cognitive data items; and an elderly subject cognitive state semantic coding feature generating unit 132, which is used to input the sequence of embedded coding vectors of the cognitive data items into a converter-based cognitive state semantic association encoder to obtain the semantic coding feature vector of the cognitive state of the elderly subject.

[0044] In an embodiment of the present application, the elderly object cognitive-physiological state fusion module 140 is used to input the elderly object cognitive state semantic coding feature vector and the elderly object physiological state semantic coding feature vector into a significant fusion network based on feature principal component query matching to obtain a sparse significant fusion representation vector of the elderly object cognitive-physiological state. It should be understood that the physiological state and the cognitive state are interrelated and closely related to the health status of the elderly. In order to better mine the significant correlation features between the elderly object cognitive state semantic coding feature vector and the elderly object physiological state semantic coding feature vector, so as to better characterize the comprehensive health status of the elderly and improve the accuracy of cognitive degeneration judgment. In the technical solution of the present application, the elderly object cognitive state semantic coding feature vector and the elderly object physiological state semantic coding feature vector are significantly fused by introducing a significant fusion network based on feature principal component query matching to obtain a sparse significant fusion representation vector of the elderly object cognitive-physiological state. It is worth mentioning that the saliency fusion network based on feature principal component query matching is an advanced feature processing technology that combines feature sparsification, principal component analysis (PCA) and feature saliency fusion methods to construct key and prominent correlation representations between feature vectors. This technology aims to extract and integrate the most representative and discriminative features from the data to form a more accurate and efficient feature representation.

[0045] In detail, first, the semantic encoding feature vector of the cognitive state of the elderly subject and the semantic encoding feature vector of the physiological state of the elderly subject are standardized to eliminate the differences of different dimensions and numerical ranges, so that each feature has the same importance in the model, so as to obtain the global semantic encoding feature vector of the standardized test results and the local semantic encoding feature vector of the standardized test results. Then, the covariance matrix of the global semantic encoding feature vector of the standardized test results and the local semantic encoding feature vector of the standardized test results are calculated to reveal the correlation between the data, so as to obtain the sample covariance matrix of the cognitive state of the elderly subject and the sample covariance matrix of the physiological state of the elderly subject. Next, the two covariance matrices are subjected to matrix decomposition-based feature vector extraction to obtain a set of principal component feature vectors of the cognitive state of the elderly subject and a set of principal component feature vectors of the physiological state of the elderly subject. That is, the principal component analysis algorithm is used to decompose the matrix based on the matrix to convert the high-dimensional data set into a low-dimensional representation, which helps to reduce the complexity of the data while retaining the most important information. In particular, principal component analysis (PCA) is an effective feature extraction technique that converts the original data into a set of linearly unrelated variables through linear transformation, which are sorted according to the size of the variance to capture the main changes in the data. After that, the two sets are matched based on the maximum approximate query between feature vectors, and the most relevant semantic query matching is performed by calculating the similarity between each feature, so as to find the best matching pair. Then, the gating mechanism is used to perform a fine-grained semantic union on the best matching pair to gain a deeper understanding of the associated semantic information in the physiological state and cognitive state, and to help screen and select the most relevant information, thereby filtering out noise and irrelevant data, and obtaining a set of principal component fusion feature vectors of the cognitive state and physiological state of the elderly object. Finally, each set of fusion features after the union is cascaded to obtain the sparse significant fusion representation vector of the cognitive-physiological state of the elderly object. In this way, features at different levels or from different sources can be combined to enhance the final feature representation, so that it can more comprehensively capture the complexity and diversity of the semantic association between the cognitive and physiological states of the elderly object.

[0046] Specifically, Figure 5 FIG. 1 is a block diagram of a cognitive-physiological state fusion module for elderly subjects in an intelligent early warning system for cognitive function deterioration of the elderly according to an embodiment of the present application. Figure 5As shown, the elderly subject cognitive-physiological state fusion module 140 includes: an elderly subject cognitive-physiological state semantic feature standardization unit 141, which is used to standardize the elderly subject cognitive state semantic coding feature vector and the elderly subject physiological state semantic coding feature vector to obtain a standardized elderly subject cognitive state semantic coding feature vector and a standardized elderly subject physiological state semantic coding feature vector; an elderly subject cognitive-physiological state sample covariance matrix calculation unit 142, which is used to calculate the sample covariance matrix of the standardized elderly subject cognitive state semantic coding feature vector and the standardized elderly subject physiological state semantic coding feature vector to obtain an elderly subject cognitive state sample covariance matrix and an elderly subject physiological state sample covariance matrix; an elderly subject cognitive-physiological state principal component feature vector generation unit 143, which is used to perform matrix decomposition-based feature vector extraction on the elderly subject cognitive state sample covariance matrix and the elderly subject physiological state sample covariance matrix to obtain a set of elderly subject cognitive state principal component feature vectors and an elderly subject physiological state principal component feature vector set. A set of feature vectors; a language-behavior maximum approximate query unit 144, which is used to input the set of the main component feature vectors of the elderly object's cognitive state and the set of the main component feature vectors of the elderly object's physiological state into a maximum approximate query matching network to obtain a set of best matching pairs of the main component feature vectors of the elderly object's cognitive state and the main component feature vectors of the elderly object's physiological state; an elderly object cognitive state-physiological state main component fusion unit 145, which is used to input each best matching pair of the main component feature vectors of the elderly object's cognitive state and the main component feature vectors of the elderly object's physiological state in the set of best matching pairs of the main component feature vectors of the elderly object's cognitive state and the main component feature vectors of the elderly object's physiological state into a semantic fine-grained gated joint module to obtain a set of the main component fusion feature vectors of the elderly object's cognitive state-physiological state; a sparse significant fusion representation generation unit 146 for the elderly object's cognitive-physiological state, which is used to cascade the set of the main component fusion feature vectors of the elderly object's cognitive state-physiological state to obtain the sparse significant fusion representation vector of the elderly object's cognitive-physiological state.

[0047] More specifically, in the embodiment of the present application, the semantic feature standardization unit 141 of the cognitive physiological state of the elderly object is used to: respectively calculate the mean and standard deviation of the semantic coding feature vector of the cognitive state of the elderly object to obtain the mean and standard deviation of the semantic coding feature of the cognitive state of the elderly object; after subtracting the semantic coding feature vector of the cognitive state of the elderly object from the mean of the semantic coding feature of the cognitive state of the elderly object by position, the calculated semantic offset vector of the cognitive state of the elderly object and the standard deviation of the semantic coding feature of the cognitive state of the elderly object are divided by position to obtain the standardized semantic coding feature vector of the cognitive state of the elderly object; respectively calculate the mean and standard deviation of the semantic coding feature vector of the physiological state of the elderly object to obtain the mean and standard deviation of the semantic coding feature of the physiological state of the elderly object; after subtracting the semantic coding feature vector of the physiological state of the elderly object from the mean of the semantic coding feature of the physiological state of the elderly object by position, the calculated semantic offset vector of the physiological state of the elderly object and the standard deviation of the semantic coding feature of the physiological state of the elderly object are divided by position to obtain the standardized semantic coding feature vector of the physiological state of the elderly object.

[0048] More specifically, in an embodiment of the present application, the elderly subject cognitive physiological state sample covariance matrix calculation unit 142 is used to: after multiplying the transposed vector of the standardized elderly subject cognitive state semantic coding feature vector by the standardized elderly subject cognitive state semantic coding feature vector, divide the obtained standardized elderly subject cognitive state semantic association matrix by the value obtained by subtracting one from the length of the standardized elderly subject cognitive state semantic coding feature vector to obtain the elderly subject cognitive state sample covariance matrix; after multiplying the transposed vector of the standardized elderly subject physiological state semantic coding feature vector by the standardized elderly subject physiological state semantic coding feature vector, divide the obtained standardized elderly subject physiological state semantic association matrix by the value obtained by subtracting one from the length of the standardized elderly subject physiological state semantic coding feature vector to obtain the elderly subject physiological state sample covariance matrix.

[0049] More specifically, in the embodiment of the present application, the language-behavior maximum approximation query unit 144 is used to: extract a predetermined principal component feature vector of the cognitive state of the elderly object from the set of principal component feature vectors of the cognitive state of the elderly object; calculate the cosine similarity between the predetermined principal component feature vector of the cognitive state of the elderly object and each principal component feature vector of the physiological state of the elderly object in the set of principal component feature vectors of the physiological state of the elderly object to obtain a set of matching query similarities; and use the principal component feature vector of the physiological state of the elderly object corresponding to the largest matching query similarity in the set of matching query similarities and the predetermined principal component feature vector of the cognitive state of the elderly object as the best matching pair of the predetermined principal component feature vector of the cognitive state of the elderly object and the principal component feature vector of the physiological state of the elderly object.

[0050] More specifically, in the embodiment of the present application, the elderly subject cognitive state-physiological state principal component fusion unit 145 is used to: respectively calculate the position difference, position dot product and position addition between the best matching pair of the elderly subject cognitive state principal component feature vector and the elderly subject physiological state principal component feature vector to obtain the elderly subject cognitive state-physiological state principal component difference vector, the elderly subject cognitive state-physiological state principal component dot product vector and the elderly subject cognitive state-physiological state principal component sum vector; cascade the elderly subject cognitive state-physiological state principal component difference vector, the elderly subject cognitive state-physiological state principal component dot product vector and the elderly subject cognitive state-physiological state principal component sum vector, and then perform one-dimensional convolution encoding to obtain the elderly subject cognitive state-physiological state principal component multi-dimensional fusion vector; perform local window-based maximum pooling processing on the elderly subject cognitive state-physiological state principal component multi-dimensional fusion vector to obtain the elderly subject cognitive state-physiological state principal component fusion feature vector.

[0051] In the embodiment of the present application, specifically, the elderly object cognitive-physiological state fusion module 140 is used to: input the elderly object cognitive state semantic coding feature vector and the elderly object physiological state semantic coding feature vector into a significant fusion network based on feature principal component query matching, and process them with the following significant fusion formula to obtain the elderly object cognitive-physiological state sparse significant fusion representation vector; wherein the significant fusion formula is: ; ; ; ; ; ; ; ; ; ; ; ; ;in, and represent the semantic coding feature vector of the cognitive state of the elderly subject and the semantic coding feature vector of the physiological state of the elderly subject respectively, and are respectively the means of the semantic encoding feature vector of the cognitive state of the elderly subject and the semantic encoding feature vector of the physiological state of the elderly subject, and are the standard deviations of the semantic encoding feature vector of the cognitive state of the elderly subject and the semantic encoding feature vector of the physiological state of the elderly subject, respectively. and are the standardized semantic coding feature vector of cognitive state of elderly subjects and the standardized semantic coding feature vector of physiological state of elderly subjects, respectively. and They are and The transposed vector of and are the lengths of the standardized semantic encoding feature vector of cognitive state of elderly subjects and the lengths of the standardized semantic encoding feature vector of physiological state of elderly subjects, respectively. and are respectively the sample covariance matrix of the cognitive state of the elderly subject and the sample covariance matrix of the physiological state of the elderly subject, and They are the principal component orthogonal matrix of the cognitive state of the elderly and the principal component orthogonal matrix of the physiological state of the elderly, and are the diagonal matrices of cognitive state and physiological state of the elderly subjects, respectively. The diagonal elements of the matrix are The diagonal matrix of cognitive states of elderly subjects, are the weight values ​​of the principal component eigenvectors of the cognitive states of each elderly subject, The diagonal elements of the matrix are The diagonal matrix of the physiological state of the elderly subjects, are the weight values ​​of the principal component eigenvectors of the physiological states of each elderly subject, and They are and The transposed matrix of is the principal component feature vector of the cognitive state of each elderly subject in the set of principal component feature vectors of the cognitive state of the elderly subject, is the principal component feature vector of the physiological state of each elderly subject in the set of principal component feature vectors of the physiological state of the elderly subject, is the first in the set of principal component eigenvectors of the cognitive state of the elderly subject. The principal component eigenvector of the cognitive state of the elderly subjects, and are respectively the principal component eigenvectors of the physiological state of the elderly subject. and The principal component eigenvectors of the physiological states of the elderly subjects, For calculation and The vector inner product between To calculate the one-norm of a vector, To return the maximum value value, is the maximum approximate matching value, , and They are position difference, position dot multiplication and position addition respectively. For cascade processing, is a one-dimensional convolutional coding operation, is the maximum pooling operation, is the first in the set of principal component fusion feature vectors of the cognitive state and physiological state of the elderly subject. The principal component fusion feature vector of cognitive state and physiological state of elderly subjects, is the number of feature vectors in the set of principal component fusion feature vectors of the cognitive state and physiological state of the elderly subject, is a sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject.

[0052] In an embodiment of the present application, the analysis result generation module 150 is used to obtain an analysis result based on the sparse significant fusion representation vector of the cognitive-physiological state of the elderly object, and the analysis result is used to indicate whether a cognitive deterioration warning prompt is generated. Specifically, in an embodiment of the present application, the analysis result generation module 150 is used to: input the sparse significant fusion representation vector of the cognitive-physiological state of the elderly object into the cognitive deterioration warning analysis module based on the classifier to obtain the analysis result, and the analysis result is used to indicate whether a cognitive deterioration warning prompt is generated. That is, the sparse significant fusion representation vector of the cognitive-physiological state of the elderly object obtained by significantly fusion of the semantic encoding feature vector of the cognitive state of the elderly object and the semantic encoding feature vector of the physiological state of the elderly object is used for classification processing, so as to automatically determine whether a cognitive deterioration warning prompt is generated. In this way, the semantic relationship between physiology and cognition can be captured and represented more accurately, which helps to reveal deeper connections and complex interactions between data, thereby enhancing the model's ability to recognize cognitive deterioration features, thereby improving the accuracy of predicting cognitive impairment.

[0053] Preferably, considering that the semantic encoding feature vector of the physiological state of the elderly subject and the semantic encoding feature vector of the cognitive state of the elderly subject respectively represent the associated semantic features of the physiological measurement data and the Montreal Cognitive Scale data based on the embedded coding context, when performing significant fusion based on feature principal component query matching, there will be a semantic fusion offset caused by the imbalance of feature query matching significance under the difference in source semantic embedded coding context, so as to further improve the semantic fusion regression comprehensibility of the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject, thereby improving the accuracy of the analysis result obtained by inputting it into the classifier-based cognitive degradation early warning analysis module.

[0054] Therefore, when the sparsely significant fused representation vector of the cognitive-physiological state of the elderly subject is input into the classifier-based cognitive degradation early warning analysis module to obtain the analysis result, the present application optimizes the sparsely significant fused representation vector of the cognitive-physiological state of the elderly subject, including the steps of: subtracting the 0 norm of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly subject from the length of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly subject to obtain the sparsely significant fused isolated representation value of the cognitive-physiological state of the elderly subject; calculating the logarithm of the sum of the square of the sparsely significant fused isolated representation value of the cognitive-physiological state of the elderly subject and the sparsely significant fused isolated representation value of the cognitive-physiological state of the elderly subject with base 2 to obtain the sparsely significant fused information order value of the cognitive-physiological state of the elderly subject; calculating the exponential value of the difference between the sparsely significant fused isolated representation value of the cognitive-physiological state of the elderly subject minus one and taking each eigenvalue of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly subject as the base. A power function is used, and the point multiplication is performed with the sparse significant fusion information order value of the cognitive-physiological state of the elderly object to obtain the sparse significant fusion leading vector of the cognitive-physiological state of the elderly object; the sparse significant fusion representation vector of the cognitive-physiological state of the elderly object is multiplied by the difference between the sparse significant fusion isolated representation value of the cognitive-physiological state of the elderly object minus one, and then the point multiplication is performed with the reciprocal of the sparse significant fusion isolated representation value of the cognitive-physiological state of the elderly object to obtain the sparse significant fusion field constraint vector of the cognitive-physiological state of the elderly object; an exponential function is calculated with the natural constant as the base, and each eigenvalue of the sparse significant fusion field constraint vector of the cognitive-physiological state of the elderly object is an exponential function to obtain the sparse significant fusion field bias vector of the cognitive-physiological state of the elderly object; the sparse significant fusion leading vector of the cognitive-physiological state of the elderly object is added with the sparse significant fusion field bias vector of the cognitive-physiological state of the elderly object to obtain the optimized sparse significant fusion representation vector of the cognitive-physiological state of the elderly object.

[0055] The optimization process of the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject is expressed as: ; ;in, represents the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject, represents the length of the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject, represents the zero norm of the eigenvector, represents the sparse significant fusion isolated representation value of the cognitive-physiological state of the elderly subject, It means point multiplication by position. To calculate a power function with each eigenvalue of the sparse significant fusion representation vector of the elderly subject's cognitive-physiological state as the base and the difference between the sparse significant fusion isolated representation value of the elderly subject's cognitive-physiological state minus one as the exponent, Indicates adding by position point. represents the natural exponential function, Represents the optimized sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject.

[0056] Therefore, for the high-dimensional feature manifold of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly object, the eigenvalues ​​of the feature set are used as the vector field representation of the aggregation dimension, and the superposition value of the vector field of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly object at the isolated zero position is used as the order information to fix the local position of the eigenvalues ​​of its feature set, and a bias for the reversibility of the feature regression distribution field of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly object is added as a reward to achieve the mapping target tracking of the regression distribution of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly object for the eigenvalue position, so that the feature set of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly object perceives the mapping migration to the aggregate distribution, thereby improving the accuracy of the analysis result obtained by inputting the sparsely significant fused representation vector of the cognitive-physiological state of the elderly object into the classifier-based cognitive degradation early warning analysis module by improving the semantic fusion regression comprehensibility of the sparsely significant fused representation vector of the cognitive-physiological state of the elderly object. In this way, the semantic relationship between physiology and cognition can be captured and represented more accurately, which helps to reveal deeper connections and complex interactions between data, thereby enhancing the model's ability to identify the characteristics of cognitive degeneration and improving the accuracy of predicting cognitive impairment, thereby providing a more intelligent early warning system for cognitive function deterioration in the elderly.

[0057] In summary, the intelligent early warning system 100 for cognitive function deterioration in the elderly based on the embodiment of the present application is explained, which uses a data analysis technology based on deep learning to embed and semantically associate the physiological measurement data of the analyzed elderly object and each data item in the Montreal Cognitive Scale data, so as to automatically determine whether to generate a cognitive deterioration early warning prompt based on the significant fusion representation features between the physiological state semantic features of the physiological measurement data and the cognitive state semantic features of the cognitive scale data based on principal component query matching. In this way, the semantic relationship between physiology and cognition can be captured and represented more accurately, which helps to enhance the model's ability to recognize cognitive deterioration features, thereby improving the accuracy of predicting cognitive impairment.

[0058] As described above, the intelligent early warning system 100 for cognitive function degradation of the elderly according to the embodiment of the present application can be implemented in various terminal devices, such as a server for intelligent early warning of cognitive function degradation of the elderly. In one example, the intelligent early warning system 100 for cognitive function degradation of the elderly according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the intelligent early warning system 100 for cognitive function degradation of the elderly can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the intelligent early warning system 100 for cognitive function degradation of the elderly can also be one of the many hardware modules of the terminal device.

[0059] Alternatively, in another example, the intelligent early warning system for cognitive function degradation in the elderly 100 and the terminal device may also be separate devices, and the intelligent early warning system for cognitive function degradation in the elderly 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0060] Figure 6 Flow chart of the intelligent early warning method for cognitive function deterioration in the elderly according to an embodiment of the present application. Figure 6As shown, in the intelligent early warning method for cognitive function deterioration of the elderly, the method includes: S110, obtaining physiological measurement data and Montreal cognitive scale data of the analyzed elderly object; S120, embedding and coding each data item in the physiological measurement data to obtain a sequence of physiological measurement data item embedding coding vectors, and then inputting the sequence into a physiological state semantic association encoder to obtain a physiological state semantic coding feature vector of the elderly object; S130, embedding and coding each data item in the Montreal cognitive scale data to obtain a sequence of cognitive data item embedding coding vectors, and then inputting the sequence into a cognitive state semantic association encoder to obtain a cognitive state semantic coding feature vector of the elderly object; S140, inputting the cognitive state semantic coding feature vector of the elderly object and the physiological state semantic coding feature vector of the elderly object into a significant fusion network based on feature principal component query matching to obtain a sparse significant fusion representation vector of the cognitive-physiological state of the elderly object; S150, based on the sparse significant fusion representation vector of the cognitive-physiological state of the elderly object, obtaining an analysis result, and the analysis result is used to indicate whether a cognitive deterioration warning prompt is generated.

[0061] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned intelligent early warning method for cognitive function deterioration in the elderly have been described in the above reference. Figures 1 to 5 The invention has been introduced in detail in the description of the intelligent early warning system for cognitive function deterioration in the elderly, and therefore, its repeated description will be omitted.

[0062] In summary, the intelligent early warning method for cognitive function deterioration of the elderly based on the embodiment of the present application is explained, which uses a data analysis technology based on deep learning to embed the physiological measurement data of the analyzed elderly object and each data item in the Montreal Cognitive Scale data to encode and semantically associate, so as to automatically judge whether to generate a cognitive deterioration early warning prompt based on the significant fusion representation features between the physiological state semantic features of the physiological measurement data and the cognitive state semantic features of the cognitive scale data based on principal component query matching. In this way, the semantic relationship between physiology and cognition can be captured and represented more accurately, which helps to enhance the model's ability to recognize cognitive deterioration features, thereby improving the accuracy of predicting cognitive impairment.

[0063] The above is only a preferred embodiment of the present application and does not constitute any form of limitation to the present application. Although the present application has been disclosed as a preferred embodiment as above, it is not intended to limit the present application. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of the technical contents disclosed above without departing from the scope of the technical solution of the present application. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present application without departing from the content of the technical solution of the present application still falls within the scope of the technical solution of the present application.

Claims

1. An intelligent early warning system for cognitive function deterioration in the elderly, characterized in that: include: The module for collecting data on the relevant status of the elderly subjects is used to obtain the physiological measurement data and Montreal Cognitive Scale data of the elderly subjects being analyzed; A physiological state semantic coding module for elderly subjects, used for embedding and coding each data item in the physiological measurement data to obtain a sequence of physiological measurement data item embedding coding vectors, and then inputting the sequence into a physiological state semantic association encoder to obtain a physiological state semantic coding feature vector for elderly subjects; A semantic coding module for cognitive state of elderly subjects, used for embedding and coding each data item in the Montreal Cognitive Scale data to obtain a sequence of cognitive data item embedding coding vectors, and then inputting the sequence into a cognitive state semantic association encoder to obtain a semantic coding feature vector of cognitive state of elderly subjects; The elderly object cognitive-physiological state fusion module is used to input the elderly object cognitive state semantic encoding feature vector and the elderly object physiological state semantic encoding feature vector into a significant fusion network based on feature principal component query matching to obtain a sparse significant fusion representation vector of the elderly object cognitive-physiological state; An analysis result generating module, used to obtain an analysis result based on the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject, wherein the analysis result is used to indicate whether a cognitive degradation warning prompt is generated; The physiological state semantic encoding module of the elderly subject includes: An elderly subject physiological measurement data embedding coding unit, used to embed the data items in the physiological measurement data using a physiological data embedding coding matrix to obtain a sequence of embedding coding vectors of the physiological measurement data items; A physiological state semantic coding feature generating unit for an elderly subject, configured to input the sequence of the physiological measurement data items embedded in the coding vector into a converter-based physiological state semantic association encoder to obtain the physiological state semantic coding feature vector for the elderly subject; The semantic coding module for cognitive status of elderly subjects includes: An elderly subject cognitive data embedding coding unit, used for embedding coding each data item in the Montreal Cognitive Scale data using a cognitive data embedding coding matrix to obtain a sequence of cognitive data item embedding coding vectors; A unit for generating semantic coding features of cognitive state of elderly subjects, used for inputting the sequence of cognitive data items embedded in coding vectors into a converter-based cognitive state semantic association encoder to obtain the semantic coding feature vector of cognitive state of elderly subjects; The elderly subject cognitive-physiological state fusion module includes: A semantic feature standardization unit for cognitive physiological state of an elderly subject, used for standardizing the semantic coding feature vector of cognitive state of the elderly subject and the semantic coding feature vector of physiological state of the elderly subject to obtain a standardized semantic coding feature vector of cognitive state of the elderly subject and a standardized semantic coding feature vector of physiological state of the elderly subject; The elderly subject cognitive physiological state sample covariance matrix calculation unit is used to calculate the sample covariance matrix of the standardized elderly subject cognitive state semantic coding feature vector and the standardized elderly subject physiological state semantic coding feature vector to obtain the elderly subject cognitive state sample covariance matrix and the elderly subject physiological state sample covariance matrix; A principal component eigenvector generating unit for the cognitive physiological state of an elderly subject, used for extracting eigenvectors based on matrix decomposition from the sample covariance matrix of the cognitive state of the elderly subject and the sample covariance matrix of the physiological state of the elderly subject to obtain a set of principal component eigenvectors of the cognitive state of the elderly subject and a set of principal component eigenvectors of the physiological state of the elderly subject; A language-behavior maximum approximation query unit, used for inputting the set of principal component feature vectors of the cognitive state of the elderly subject and the set of principal component feature vectors of the physiological state of the elderly subject into a maximum approximation query matching network to obtain a set of best matching pairs of principal component feature vectors of the cognitive state of the elderly subject and principal component feature vectors of the physiological state of the elderly subject; The elderly subject cognitive state-physiological state principal component fusion unit is used to input each best matching pair of the elderly subject cognitive state principal component feature vector and the elderly subject physiological state principal component feature vector in the set of best matching pairs of the elderly subject cognitive state principal component feature vector and the elderly subject physiological state principal component feature vector into the semantic fine-grained gated joint module to obtain a set of elderly subject cognitive state-physiological state principal component fusion feature vectors; The sparse significant fusion representation generation unit of the cognitive-physiological state of the elderly subject is used to cascade the set of principal component fusion feature vectors of the cognitive state-physiological state of the elderly subject to obtain the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject.

2. The intelligent early warning system for cognitive function deterioration in the elderly according to claim 1 is characterized in that: The semantic feature standardization unit of the cognitive physiological state of the elderly subject is used for: Calculating the mean and standard deviation of the semantic coding feature vector of the cognitive state of the elderly subject respectively to obtain the mean of the semantic coding feature of the cognitive state of the elderly subject and the standard deviation of the semantic coding feature of the cognitive state of the elderly subject; After subtracting the semantic coding feature vector of the cognitive state of the elderly subject from the mean value of the semantic coding feature of the cognitive state of the elderly subject by position, the positional division of the calculated semantic offset vector of the cognitive state of the elderly subject and the standard deviation of the semantic coding feature of the cognitive state of the elderly subject is performed to obtain the standardized semantic coding feature vector of the cognitive state of the elderly subject; Calculating the mean and standard deviation of the semantic coding feature vector of the physiological state of the elderly subject respectively to obtain the mean of the semantic coding feature of the physiological state of the elderly subject and the standard deviation of the semantic coding feature of the physiological state of the elderly subject; After positionally subtracting the semantic coding feature vector of the physiological state of the elderly object from the mean of the semantic coding features of the physiological state of the elderly object, the calculated semantic offset vector of the physiological state of the elderly object and the standard deviation of the semantic coding features of the physiological state of the elderly object are divided by position to obtain the standardized semantic coding feature vector of the physiological state of the elderly object.

3. The intelligent early warning system for cognitive function deterioration in the elderly according to claim 2 is characterized in that: The elderly subject cognitive physiological state sample covariance matrix calculation unit is used to: After multiplying the transposed vector of the standardized elderly subject cognitive state semantic coding feature vector by the standardized elderly subject cognitive state semantic coding feature vector, the obtained standardized elderly subject cognitive state semantic association matrix is ​​divided by the value obtained by subtracting one from the length of the standardized elderly subject cognitive state semantic coding feature vector to obtain the elderly subject cognitive state sample covariance matrix; After multiplying the transposed vector of the standardized semantic coding feature vector of the physiological state of elderly subjects by the standardized semantic coding feature vector of the physiological state of elderly subjects, the obtained semantic association matrix of the physiological state of elderly subjects is divided by the value obtained by subtracting one from the length of the standardized semantic coding feature vector of the physiological state of elderly subjects by position to obtain the sample covariance matrix of the physiological state of elderly subjects.

4. The intelligent early warning system for cognitive function deterioration in the elderly according to claim 3 is characterized in that: The language-behavior maximum approximation query unit is used to: Extracting a predetermined principal component feature vector of the cognitive state of the elderly subject from the set of principal component feature vectors of the cognitive state of the elderly subject; Calculating the cosine similarity between the predetermined principal component feature vector of the cognitive state of the elderly subject and each principal component feature vector of the physiological state of the elderly subject in the set of principal component feature vectors of the physiological state of the elderly subject to obtain a set of matching query similarities; The principal component feature vector of the physiological state of the elderly subject corresponding to the largest matching query similarity in the set of matching query similarities and the predetermined principal component feature vector of the cognitive state of the elderly subject are taken as the best matching pair of the predetermined principal component feature vector of the cognitive state of the elderly subject and the principal component feature vector of the physiological state of the elderly subject.

5. The intelligent early warning system for cognitive function deterioration in the elderly according to claim 4 is characterized in that: The elderly subject cognitive state-physiological state principal component fusion unit is used to: Respectively calculating the position difference, position dot product and position addition between the best matching pairs of the principal component feature vectors of the elderly subject's cognitive state and the principal component feature vectors of the elderly subject's physiological state to obtain the principal component difference vector of the elderly subject's cognitive state-physiological state, the principal component dot product vector of the elderly subject's cognitive state-physiological state and the principal component sum vector of the elderly subject's cognitive state-physiological state; The difference vector of the principal component of the elderly subject's cognitive state-physiological state, the dot product vector of the principal component of the elderly subject's cognitive state-physiological state and the sum vector of the principal component of the elderly subject's cognitive state-physiological state are cascaded and then one-dimensional convolutional encoded to obtain a multi-dimensional fusion vector of the principal component of the elderly subject's cognitive state-physiological state; The multi-dimensional fusion vector of the principal component of the cognitive state and physiological state of the elderly subject is subjected to a local window-based maximum pooling process to obtain a principal component fusion feature vector of the cognitive state and physiological state of the elderly subject.

6. The intelligent early warning system for cognitive function deterioration in the elderly according to claim 5 is characterized in that: The analysis result generating module is used to input the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject into the classifier-based cognitive degradation early warning analysis module to obtain the analysis result, and the analysis result is used to indicate whether a cognitive degradation early warning prompt is generated.

7. An intelligent early warning method for cognitive function deterioration in the elderly according to the intelligent early warning system for cognitive function deterioration in the elderly according to claim 1, characterized in that: include: Obtain physiological measurement data and Montreal Cognitive Scale data from the analyzed elderly subjects; Embedding and encoding each data item in the physiological measurement data to obtain a sequence of physiological measurement data item embedding encoding vectors, and then inputting the sequence into a physiological state semantic association encoder to obtain a physiological state semantic encoding feature vector of the elderly subject; Embedding and encoding each data item in the Montreal Cognitive Scale data to obtain a sequence of cognitive data item embedding encoding vectors, and then inputting the sequence into a cognitive state semantic association encoder to obtain a cognitive state semantic encoding feature vector of the elderly subject; Inputting the semantic encoding feature vector of the cognitive state of the elderly subject and the semantic encoding feature vector of the physiological state of the elderly subject into a significant fusion network based on feature principal component query matching to obtain a sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject; Based on the sparse significant fusion representation vector of the cognitive-physiological state of the elderly subject, an analysis result is obtained, and the analysis result is used to indicate whether a cognitive degradation early warning prompt is generated.

Images