A pension information planning method and system based on the Internet of Things and big data
Through the Internet of Things and big data technology, wearable devices are used to collect physiological signals and predict emotional states, which solves the problem of low accuracy in pension information planning caused by manual observation, and achieves higher accuracy and effectiveness.
Patent Information
- Application Number
- CN202410194107.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-21
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-02-21
AI Technical Summary
The existing nursing home planning methods rely on manual observation of facial expressions, and there are human subjective factors, resulting in low emotional classification accuracy and nursing information planning accuracy.
Using the Internet of Things and big data method, multi-channel human physiological signals are obtained through wearable acquisition devices, feature classification, cluster analysis and micro-state division are carried out, and micro-state segmentation is mapped into micro-state sequence features, input emotional state prediction model, and predict emotional state , and plan pension information based on the emotional calculation results.
It improves the accuracy of pension information planning, reduces the influence of human subjective factors, and enhances the accuracy and effectiveness of emotional recognition.
Smart Images

Figure CN117894465B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of physiological signal analysis, and in particular to a pension information planning method and system based on the Internet of Things and big data. Background Art
[0002] In the planning of nursing homes, in order to classify elderly users with the same personality into one nursing home, the existing method is to determine the emotions of different elderly users through manual experience observation of their facial expressions, and then classify the elderly users according to the manually classified emotional results. However, there are human subjective factors in the manual classification process, which leads to low accuracy of the manually determined results, resulting in low accuracy of the final elderly information planning. Summary of the invention
[0003] In view of the technical problems existing in the prior art, the present invention provides a pension information planning method and system based on the Internet of Things and big data, so as to improve the accuracy of pension information planning.
[0004] The technical solution of the present invention to solve the above technical problems is as follows: a pension information planning method based on the Internet of Things and big data, comprising:
[0005] Perform feature classification on the multi-channel human physiological signals of the target elderly users obtained through wearable collection devices to obtain the target physiological signal features;
[0006] Performing cluster analysis on the target physiological signal characteristics to obtain physiological characteristic categories, and performing micro-state division of physiological signals on the physiological characteristic categories to obtain micro-state division results;
[0007] Based on the microstate division result, the original physiological signal is mapped to the microstate category, the continuous physiological signal is converted into a microstate sequence, and the microstate sequence feature is extracted;
[0008] Inputting the microstate sequence features into an emotional state prediction model to obtain an emotional calculation result output by the emotional state prediction model; the emotional state prediction model is trained based on sample microstate sequence features of the Internet of Things and big data, and emotional labels corresponding to the sample microstate sequence features;
[0009] The pension information planning is performed according to the emotion calculation result.
[0010] According to a pension information planning method based on the Internet of Things and big data provided by the present invention, pension information planning is performed according to the emotion calculation result, including:
[0011] Target elderly care users with the same emotion calculation results are classified to obtain multiple elderly care user groups;
[0012] Plan each group of elderly care users to a different nursing home.
[0013] According to a pension information planning method based on the Internet of Things and big data provided by the present invention, feature classification is performed on multi-channel human physiological signals of target pension users acquired through wearable acquisition devices to obtain target physiological signal features, including:
[0014] Inputting the multi-channel human physiological signal into a first convolution module in a physiological signal classification model, performing dimensionality reduction processing on the multi-channel human physiological signal through the first convolution module to obtain shallow signal features;
[0015] The shallow signal features are respectively input into the physiological feature mixing module and the signal feature mixing module in the physiological signal classification model, the multi-scale physiological features of the shallow signal features are extracted by the physiological feature mixing module, and the multi-scale signal features of the shallow signal features are extracted by the signal feature mixing module;
[0016] The superposition result of the multi-scale physiological feature and the multi-scale signal feature is input into the second convolution module in the physiological signal classification model, and the superposition result is classified by the second convolution module to obtain the target physiological signal feature.
[0017] According to a pension information planning method based on the Internet of Things and big data provided by the present invention, a physiological feature mixing module includes a first normalization layer, a physiological adaptive feature modulation unit, a second normalization layer and a physiological mixer connected in series in sequence, and the multi-scale physiological features of the shallow signal features are extracted by the physiological feature mixing module, including:
[0018] Normalizing the shallow signal feature through the first normalization layer to obtain a first signal feature;
[0019] Performing physiological adaptive feature modulation on the first signal feature by the physiological adaptive feature modulation unit to obtain a first physiological feature;
[0020] Normalizing the first product of the shallow signal feature and the first physiological feature through the second normalization layer to obtain a second physiological feature;
[0021] The second physiological characteristic is subjected to physiological mixing processing by the physiological mixer to obtain a third physiological characteristic, and a superposition result of the third physiological characteristic and the first product is the multi-scale physiological characteristic.
[0022] According to a pension information planning method based on the Internet of Things and big data provided by the present invention, a physiological adaptive feature modulation unit includes a multi-scale physiological feature generation unit, a first convolution layer and a first activation function connected in series in sequence, and the physiological adaptive feature modulation unit performs physiological adaptive feature modulation on the first signal feature to obtain a first physiological feature, including:
[0023] Extracting a plurality of physiological characteristic signals corresponding to the first signal characteristic by the multi-scale physiological characteristic generating unit;
[0024] Convolving the concatenated result of the plurality of physiological characteristic signals through the first convolution layer to obtain a convolved physiological characteristic signal;
[0025] The convolved physiological characteristic signal is nonlinearly transformed by the first activation function to obtain a transformed physiological characteristic signal, and the product of the transformed physiological characteristic signal and the first signal characteristic is the first physiological characteristic.
[0026] According to a pension information planning method based on the Internet of Things and big data provided by the present invention, a multi-scale physiological feature generation unit includes a first deep convolution layer and a plurality of physiological feature generation branches connected in parallel, each physiological feature generation branch includes an adaptive average pooling layer, a second deep convolution layer and a nearest neighbor upsampling layer connected in series in sequence, and the pooling resolution of the adaptive average pooling layer included in each physiological feature generation branch is different. The method of extracting a plurality of physiological feature signals corresponding to the first signal feature through the multi-scale physiological feature generation unit includes:
[0027] Performing deep convolution on the first signal feature through the first deep convolution layer to obtain a first physiological feature signal;
[0028] Generate branches for each of the physiological characteristics, perform pooling processing on the first signal characteristics through the adaptive average pooling layer to obtain the pooled physiological characteristics; perform deep convolution on the pooled physiological characteristics through the second deep convolution layer to obtain the deep convolution physiological characteristics; upsample the deep convolution physiological characteristics through the nearest neighbor upsampling layer to obtain a second physiological characteristic signal;
[0029] The multiple physiological characteristic signals include the first physiological characteristic signal and the second physiological characteristic signal obtained by each physiological characteristic generation branch.
[0030] According to a pension information planning method based on the Internet of Things and big data provided by the present invention, a signal feature mixing module includes a third normalization layer, a signal adaptive feature modulation unit, a fourth normalization layer and a signal mixer connected in series in sequence, and the multi-scale signal feature of the shallow signal feature is extracted by the signal feature mixing module, including:
[0031] Normalizing the shallow signal feature through the third normalization layer to obtain a second signal feature;
[0032] Performing signal adaptive feature modulation on the second signal feature by the signal adaptive feature modulation unit to obtain a first signal feature;
[0033] Normalizing the second product of the shallow signal feature and the first signal feature through the fourth normalization layer to obtain a second signal feature;
[0034] The second signal feature is subjected to physiological mixing processing by the signal mixer to obtain a third signal feature, and a superposition result of the third signal feature and the second product is the multi-scale signal feature.
[0035] The present invention also provides a pension information planning system based on the Internet of Things and big data, comprising:
[0036] A feature classification module is used to perform feature classification on multi-channel human physiological signals of target elderly users acquired through wearable acquisition devices to obtain target physiological signal features;
[0037] A clustering division module is used to perform cluster analysis on the target physiological signal characteristics to obtain physiological characteristic categories, and to divide the physiological characteristic categories into microstates of physiological signals to obtain microstate division results;
[0038] A mapping module, used to map the original physiological signal to the microstate category based on the microstate division result, convert the continuous physiological signal into a microstate sequence, and extract the microstate sequence features;
[0039] A prediction module, used to input the microstate sequence features into an emotional state prediction model to obtain an emotional calculation result output by the emotional state prediction model; the emotional state prediction model is trained based on sample microstate sequence features of the Internet of Things and big data, and emotional labels corresponding to the sample microstate sequence features;
[0040] The pension planning module is used to plan pension information according to the emotion calculation results.
[0041] The present invention also provides an electronic device, comprising: a memory for storing a computer software program; a processor for reading and executing the computer software program, thereby implementing any of the above-mentioned methods for elderly care information planning based on the Internet of Things and big data.
[0042] The present invention also provides a non-transitory computer-readable storage medium, characterized in that a computer software program is stored in the storage medium, and when the computer software program is executed by a processor, it implements any of the above-mentioned pension information planning methods based on the Internet of Things and big data.
[0043] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for elderly care information planning based on the Internet of Things and big data.
[0044] The beneficial effect of the present invention is that the emotion calculation results of the elderly care users are accurately predicted through wearable collection devices, as well as the emotion state prediction model of the Internet of Things and big data, so that the elderly care information planning is carried out through the accurate emotion calculation results, thereby improving the accuracy of the elderly care information planning. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] Figure 1 It is a flow chart of the pension information planning method based on the Internet of Things and big data provided by the present invention;
[0046] Figure 2 is a schematic diagram of the structure of a physiological signal classification model provided by an embodiment of the present invention;
[0047] Figure 3 is a schematic diagram of the structure of a physiological characteristic mixing module provided in an embodiment of the present invention;
[0048] Figure 4 is a schematic diagram of the structure of a physiological adaptive feature modulation unit provided in an embodiment of the present invention;
[0049] Figure 5 is a schematic structural diagram of a physiological mixer provided by an embodiment of the present invention;
[0050] Figure 6 is a schematic diagram of the structure of a signal feature mixing module provided in an embodiment of the present invention;
[0051] Figure 7 is a schematic diagram of the structure of a signal adaptive feature modulation unit provided in an embodiment of the present invention;
[0052] Figure 8 is a structural schematic diagram of a signal mixer provided by an embodiment of the present invention;
[0053] Fig. 9It is a structural diagram of the pension information planning system based on the Internet of Things and big data provided by the present invention;
[0054] Fig.10 is a schematic diagram of an embodiment of an electronic device provided by an embodiment of the present invention;
[0055] Fig.11 It is a schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0057] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
[0058] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.
[0059] Reference Figure 1 The present invention describes a pension information planning method based on the Internet of Things and big data, comprising steps 100 to 500:
[0060] Step 100: feature classify the multi-channel human physiological signals of the target elderly user acquired through the wearable acquisition device to obtain target physiological signal features.
[0061] In the present invention, the use of wearable measurement devices to collect effective physiological signals is the basis for subsequent microstate analysis. Since the latency and duration of common peripheral physiological responses are usually between a few seconds and tens of seconds, the physiological measurement equipment required by this method should have a sampling rate of not less than 1Hz, and the measurement signal categories include: pulse wave, skin electricity, electrocardiogram, respiration, and skin temperature. At the same time, the signals of each channel should ensure time alignment when they are aggregated. In terms of data acquisition, wearable devices that meet the measurement requirements are selected. Considering that wearable devices are more often used in daily measurements, they need to have effective measurement capabilities in daily life scenarios. Specifically, the most important aspects are: signal quality and signal-to-noise ratio, anti-interference effect (such as anti-motion interference), data transmission or storage method, sampling rate, and continuous measurement time.
[0062] Step 200: cluster analysis is performed on target physiological signal features to obtain physiological feature categories, and the physiological feature categories are divided into microstates of physiological signals to obtain microstate division results.
[0063] Specifically, the features of multi-channel human physiological signals are extracted, and then the dimension is reduced and compressed. The features after the dimension compression are clustered to obtain physiological feature categories, and the micro-state division of physiological signals is performed, which specifically includes:
[0064] Acquire multi-channel human physiological signals, segment them according to time windows and extract standardized physiological features;
[0065] All physiological characteristics were summarized and reduced in dimension based on standardized physiological characteristics. Feature dimensions with higher weights were extracted through principal component analysis, and cluster analysis was performed on the physiological characteristics after dimension reduction.
[0066] Check the distribution of cluster analysis results, exclude cluster results with too small a sample ratio or only concentrated on a few individuals, and perform stability checks on cluster results;
[0067] The valid feature categories are named as the corresponding microstates, the microstate division is completed, and the cluster center points are marked to map to the physiological features of the original feature physiology.
[0068] In the present invention, preliminary analysis of physiological signals to determine microstate categories is the basis for subsequent feature extraction and emotion calculation. Since the response speed of peripheral physiological reactions is usually slower than that of EEG signals, and changes on a time scale of seconds, the time window used to calculate the characteristics of peripheral physiological reactions should not be shorter than 0.1s to capture the dynamic patterns of physiological reactions related to emotions. After extracting the features of multi-channel physiological signals, the data is reduced in dimension (such as principal component analysis) to compress the feature dimensions, and the obtained samples are clustered to obtain several major physiological feature categories, which are called microstates. Microstates should meet basic constraints, such as the overall proportion of each microstate cannot be too small (such as <0.1%), and should have a relatively uniform distribution at the cross-individual level. Microstates that do not meet the conditions are considered to be merged or recalculated with other microstates.
[0069] In the microstate division stage, the preprocessed physiological signals are first divided according to the time window and their main features are extracted. The time window is not shorter than 0.1s. Common physiological features include but are not limited to time domain features and frequency domain features, such as statistical indicators such as mean, standard deviation, skewness, kurtosis, and frequency domain attributes such as peak frequency and harmonics. In addition, the waveform characteristics of short-term physiological changes, as well as methods such as autoregressive coefficients and wavelet decomposition coefficients can also extract high-order continuous signal features. At the same time, there may be inherent individual differences in the physiological signals themselves, which are eliminated by intra-individual normalization. After extracting standardized physiological features, all samples are summarized and dimensionally reduced. The feature dimensions with higher weights are extracted by principal component analysis (PCA). The explained variance of the dimension reduction results is not less than 95%. The purpose of dimension reduction is to reduce the complexity of subsequent cluster analysis. Finally, the k-means, cluster density peak and other methods are used to cluster the physiological features after dimension reduction. The k-means can specify the number of category center points, so the effectiveness of the clustering results is evaluated by traversing different numbers of categories to determine the final clustering results. Clustering effectiveness evaluation indicators include but are not limited to silhouette coefficient and clustering error sum of squares.
[0070] Check the distribution of clustering results, exclude the clustering results with too small a proportion of samples and those concentrated on only a few individuals, and because k-means clustering has a certain degree of randomness, the clustering results should be checked for stability. Then, the valid feature categories are named as the corresponding microstates, and the physiological features of the original feature physiology are marked with their cluster center points mapped to them.
[0071] Specifically, after completing the classification of microstate categories, the mapping method from the original physiological signal to the microstate category is obtained;
[0072] Based on the mapping method of original physiological signals to microstate categories, longer continuous physiological signals are converted into microstate sequences, and the features of microstate sequences are extracted.
[0073] In the present invention, after completing the microstate classification, the mapping method from the original physiological signal to the microstate category can be obtained. On this basis, the longer continuous physiological signal can be converted into a microstate sequence, and the microstate sequence characteristics can be extracted. Among them, the microstate sequence characteristics include: the total proportion of each microstate, the average duration of each microstate, and the transition probability of switching between two microstates.
[0074] Step 400: Input the microstate sequence features into the emotional state prediction model to obtain the emotion calculation result output by the emotional state prediction model.
[0075] The emotional state prediction model is based on the sample micro-state sequence features of the Internet of Things and big data, and the emotion labels corresponding to the sample micro-state sequence features are trained. Among them, the emotion labels come from the emotion-induced material labels, the subjects' self-reports, and the comprehensive behavioral expressions.
[0076] In the present invention, when actual emotion computing is applied, wearable devices can be used to collect the physiological reactions of users in different emotional states, and then the micro-state characteristics of wearable physiological signals can be matched with the emotion labels obtained by subjective reports, etc. Based on such a correspondence, an emotion state prediction model for predicting the emotion state can be established. By optimizing the model parameters on the training set and using the external test set to test the stability of the micro-state division and the accuracy of the emotion computing model, the robustness and effectiveness of the model can be ensured.
[0077] This method will calculate the proportion of each microstate, the duration, and the transition probability of different microstates, and put the microstate features into the emotional state prediction model to complete the training of the emotion recognition model. The emotional labels of physiological data can come from emotion-induced material labels, self-reports of subjects, and comprehensive behavioral performance. Considering the category attributes that the labels may have, or continuous numerical scores, the prediction model will selectively construct classification models, regression models, etc. The accuracy of the prediction model requires additional verification / testing. The test samples should be strictly independent of the existing samples to ensure that they are not involved in cluster analysis, model training and other steps. The model is evaluated based on the accuracy of the test results and optimized.
[0078] Through the method of elderly care information planning based on the Internet of Things and big data provided by the present invention, an automated emotional state prediction model is established based on the temporal dynamic characteristics of emotional physiological responses and the micro-state clustering analysis method that has been preliminarily verified in EEG signals. Compared with the traditional static perspective of wearable emotional computing, the present invention takes into account the temporal process of emotional responses, which is expected to bring new increments to clarify the physiological response patterns specific to emotions, thereby improving the accuracy and effectiveness of emotion recognition. In addition, the method based on wearable physiological measurement has a lower user compliance, which helps to promote the implementation of wearable emotional computing technology and bring new growth points to promote the progress in the field of emotional intelligence.
[0079] Step 500: Perform elderly care information planning according to the emotion calculation results, specifically: classify target elderly care users with the same emotion calculation results to obtain multiple elderly care user groups, and plan each elderly care user group to a different nursing home.
[0080] The embodiment of the present invention accurately predicts the emotion calculation results of the elderly care users through wearable data collection devices, as well as the emotion state prediction model of the Internet of Things and big data, so as to carry out elderly care information planning through accurate emotion calculation results, thereby improving the accuracy of elderly care information planning.
[0081] In the embodiment of the present invention, the structure of the physiological signal classification model can be seen in Figure 2 As shown, Figure 2 is a structural schematic diagram of a physiological signal classification model provided by an embodiment of the present invention. The physiological signal classification model may include a first convolution module, a physiological feature mixing module, a signal feature mixing module and a second convolution module.
[0082] The multi-channel human physiological signals of the target elderly users obtained through wearable collection devices are classified to obtain the target physiological signal features, including:
[0083] Inputting the multi-channel human physiological signal into the first convolution module in the physiological signal classification model, performing dimensionality reduction processing on the multi-channel human physiological signal through the first convolution module to obtain shallow signal features;
[0084] The shallow signal features are respectively input into the physiological feature mixing module and the signal feature mixing module in the physiological signal classification model, the multi-scale physiological features of the shallow signal features are extracted through the physiological feature mixing module, and the multi-scale signal features of the shallow signal features are extracted through the signal feature mixing module;
[0085] The superposition results of the multi-scale physiological features and the multi-scale signal features are input into the second convolution module in the physiological signal classification model, and the superposition results are classified by the second convolution module to obtain the target physiological signal features.
[0086] For example, in an embodiment of the present invention, the first convolution module may include a 1×1 convolution layer and a 3×3 convolution layer connected in series, which may be specifically configured according to actual needs. In an embodiment of the present invention, the entire multi-channel human physiological signal is used as the input of the physiological signal classification model, so that in the process of classifying the multi-channel human physiological signal, the long-distance dependency in the entire multi-channel human physiological signal can be effectively captured, thereby improving the accuracy of the target physiological signal feature.
[0087] In the embodiment of the present invention, a physiological feature mixing module and a signal feature mixing module are combined to jointly realize multi-channel human physiological signal classification. Among them, the physiological feature mixing module is used to extract multi-scale physiological features of shallow signal features, and the signal feature mixing module is used to extract multi-scale signal features of shallow signal features. In this way, long-distance interactions are considered by fusing multi-scale physiological features and multi-scale signal features, thereby further improving the accuracy of target physiological signal features.
[0088] For example, in an embodiment of the present invention, the second convolution module may include a 1×1 convolution layer, which may be specifically set according to actual needs. It can be seen that in an embodiment of the present invention, when classifying a multi-channel human physiological signal, the multi-channel human physiological signal to be classified may be input into the first convolution module in the physiological signal classification model, and the multi-channel human physiological signal is subjected to dimensionality reduction processing by the first convolution module, and the shallow signal features obtained by dimensionality reduction are respectively input into the physiological feature mixing module and the signal feature mixing module in the physiological signal classification model, and the multi-scale physiological features of the shallow signal features are extracted by the physiological feature mixing module, and the multi-scale signal features of the shallow signal features are extracted by the signal feature mixing module; the superposition results of the multi-scale physiological features and the multi-scale signal features are input into the second convolution module in the physiological signal classification model, and the superposition results are classified by the second convolution module to obtain the target physiological signal features. In this way, the entire multi-channel human physiological signal is used as the input of the physiological signal classification model, which can effectively capture the long-distance dependency in the entire multi-channel human physiological signal; and the long-distance interaction is considered by the fusion of the multi-scale physiological features and the multi-scale signal features, which can effectively improve the accuracy of the classification results.
[0089] For example, in the embodiment of the present invention, the network structure of the physiological feature mixing module for extracting multi-scale physiological features can be seen in Figure 3 As shown, Figure 3 It is a structural schematic diagram of the physiological feature mixing module provided in an embodiment of the present invention. The physiological feature mixing module includes a first normalization layer, a physiological adaptive feature modulation unit, a second normalization layer and a physiological mixer connected in series in sequence. The multi-scale physiological features of shallow signal features are extracted through the physiological feature mixing module.
[0090] In an optional embodiment, the multi-scale physiological features of the shallow signal features are extracted by the physiological feature mixing module, including:
[0091] Normalizing the shallow signal features through the first normalization layer to obtain the first signal features;
[0092] Performing physiological adaptive feature modulation on the first signal feature by a physiological adaptive feature modulation unit to obtain a first physiological feature;
[0093] Normalizing the first product of the shallow signal feature and the first physiological feature through a second normalization layer to obtain a second physiological feature;
[0094] The second physiological characteristic is subjected to physiological mixing processing by a physiological mixer to obtain a third physiological characteristic, and the superposition result of the third physiological characteristic and the first product is a multi-scale physiological characteristic.
[0095] For example, in the embodiments of the present invention, see Figure 4 As shown, Figure 4 It is a structural schematic diagram of a physiological adaptive feature modulation unit provided in an embodiment of the present invention. The physiological adaptive feature modulation unit includes a multi-scale physiological feature generation unit, a first convolutional layer and a first activation function connected in series.
[0096] Combination Figure 4 The physiological adaptive feature modulation unit shown in the figure, when performing physiological adaptive feature modulation on the first signal feature through the physiological adaptive feature modulation unit, can first extract multiple physiological feature signals corresponding to the first signal feature through the multi-scale physiological feature generation unit; convolve the splicing result of the multiple physiological feature signals through the first convolution layer to obtain the convolved physiological feature signal; perform nonlinear transformation on the convolved physiological feature signal through the first activation function to obtain the transformed physiological feature signal, and the product of the transformed physiological feature signal and the first signal feature is the first physiological feature.
[0097] By way of example, in an embodiment of the present invention, a multi-scale physiological feature generation unit may include a first deep convolutional layer and multiple physiological feature generation branches connected in parallel, wherein each physiological feature generation branch includes an adaptive average pooling layer, a second deep convolutional layer, and a nearest neighbor upsampling layer connected in series in sequence, and the pooling resolution of the adaptive average pooling layer included in each physiological feature generation branch is different. In this way, multiple physiological feature generation branches with different pooling resolutions can be used to extract multiple physiological feature signals of different scales, thereby providing a basis for the subsequent acquisition of multi-scale physiological features.
[0098] When extracting multiple physiological characteristic signals corresponding to the first signal feature through the above-mentioned multi-scale physiological characteristic generation unit, the first signal feature can be first deeply convolved through a first deep convolution layer to obtain a first physiological characteristic signal; for each physiological characteristic generation branch, the first signal feature is pooled through an adaptive average pooling layer to obtain a pooled physiological feature; the pooled physiological feature is deeply convolved through a second deep convolution layer to obtain a deeply convolved physiological feature; the deeply convolved physiological feature is upsampled through a nearest neighbor upsampling layer to obtain a second physiological characteristic signal; wherein the multiple physiological characteristic signals include the first physiological characteristic signal and the second physiological characteristic signal obtained by each physiological characteristic generation branch.
[0099] In an embodiment of the present invention, the first depthwise convolutional layer may include a 3×3 depthwise convolution.
[0100] Taking the case where multiple physiological characteristics generate three branches as an example, for example, see the above Figure 4 As shown, the first physiological feature generation branch may include an adaptive average pooling layer with a pooling resolution of 2, a second deep convolution layer of a 3×3 deep convolution, and a nearest neighbor upsampling layer; the second physiological feature generation branch may include an adaptive average pooling layer with a pooling resolution of 4, a second deep convolution layer of a 3×3 deep convolution, and a nearest neighbor upsampling layer; the third physiological feature generation branch may include an adaptive average pooling layer with a pooling resolution of 8, a second deep convolution layer of a 3×3 deep convolution, and a nearest neighbor upsampling layer, which can be set according to actual needs.
[0101] After extracting multiple physiological characteristic signals corresponding to the first signal feature through the multi-scale physiological characteristic generation unit, the splicing result of the multiple physiological characteristic signals can be convolved through the first convolution layer. For example, the first convolution layer can include a 3×3 depth convolution to obtain the convolved physiological characteristic signal; and the convolved physiological characteristic signal is nonlinearly transformed through the first activation function. For example, the first activation function can be a Gaussian Error Linear Unit (GELU) activation function to obtain the transformed physiological characteristic signal, and the transformed physiological characteristic signal is multiplied by the first signal feature to obtain the product of the first physiological feature.
[0102] For example, in the embodiment of the present invention, the structure of the physiological mixer can be seen in Figure 5 As shown, Figure 5 : is a schematic diagram of the structure of the physiological mixer provided in an embodiment of the present invention. The physiological mixer may include a 3×3 depth convolution, a GELU activation function and another 3×3 depth convolution connected in series. The specific configuration may be based on actual needs. Here, the embodiment of the present invention is only based on Figure 4 The physiological mixer shown is used as an example for explanation, but it does not mean that the embodiment of the present invention is limited to this.
[0103] In combination with the above description, the multi-scale physiological features of the multi-channel human physiological signals can be extracted through the physiological adaptive feature modulation unit in the physiological feature mixing module; the local information of the multi-channel human physiological signals can be captured through the physiological mixer in the physiological feature mixing module, thereby effectively improving the accuracy of the acquired multi-scale physiological features, so that when the multi-channel human physiological signals are subsequently classified in combination with the multi-scale physiological features, the accuracy of the classification results can be effectively improved.
[0104] Based on any of the above embodiments, for example, in an embodiment of the present invention, the signal feature mixing module for extracting multi-scale signal features can be referred to in Figure 6 As shown, Figure 6 It is a structural schematic diagram of a signal feature mixing module provided in an embodiment of the present invention. The signal feature mixing module includes a third normalization layer, a signal adaptive feature modulation unit, a fourth normalization layer and a signal mixer connected in series in sequence. The signal feature mixing module is used to extract multi-scale signal features of shallow signal features.
[0105] In an optional embodiment, the shallow signal feature is normalized by a third normalization layer to obtain a second signal feature;
[0106] Performing signal adaptive feature modulation on the second signal feature by a signal adaptive feature modulation unit to obtain a first signal feature;
[0107] Normalizing the second product of the shallow signal feature and the first signal feature through a fourth normalization layer to obtain a second signal feature;
[0108] The second signal feature is subjected to physiological mixing processing by a signal mixer to obtain a third signal feature, and a superposition result of the third signal feature and the second product is the multi-scale signal feature.
[0109] For example, in the embodiments of the present invention, see Figure 7 As shown, Figure 7 It is a structural schematic diagram of a signal adaptive feature modulation unit provided in an embodiment of the present invention, wherein the signal adaptive feature modulation unit comprises a convolution unit, a second convolution layer and a second activation function connected in series in sequence.
[0110] Combination Figure 7The signal adaptive feature modulation unit shown in the figure, when performing signal adaptive feature modulation on the second signal feature through the signal adaptive feature modulation unit, can first convolve the second signal feature through a convolution unit to obtain multiple signal features; convolve the addition result of the multiple signal features through a second convolution layer to obtain a first signal feature after convolution; perform a nonlinear transformation on the first signal feature through a second activation function to obtain a transformed signal feature, and the product of the transformed signal feature and the second signal feature is the first signal feature.
[0111] By way of example, in an embodiment of the present invention, the convolution unit includes multiple convolution branches in parallel, and the convolution kernels of the multiple convolution branches are different. When the second signal feature is convolved by the convolution unit, the second signal feature is convolved by the convolution branch for each convolution branch to obtain the convolved second signal feature; wherein the multiple signal features include the second signal features obtained by each convolution branch.
[0112] Taking the convolution unit including three convolution branches as an example, for example, see the above Fig. 9 As shown, a convolution branch may include a 1×1×3 convolution, a convolution branch may include a 1×1×5 convolution, and a convolution branch may include a 1×1×7 convolution, which can be set according to actual needs.
[0113] After convolving the second signal feature through the convolution unit to obtain multiple signal features, the addition result of the multiple signal features can be convolved through the second convolution layer. For example, the second convolution layer can include a 1×1 depth convolution to obtain the convolved first signal feature; and the first signal feature is nonlinearly transformed through the second activation function. For example, the second activation function can be a GELU activation function to obtain the transformed signal feature, and the transformed signal feature is multiplied by the second signal feature to obtain the product of the first signal feature.
[0114] For example, in the embodiment of the present invention, the structure of the signal mixer can be seen in Figure 8 As shown, Figure 8 is a schematic diagram of the structure of a signal mixer provided in an embodiment of the present invention. The signal mixer may include a 1×1 depth convolution, a GELU activation function and another 1×1 depth convolution connected in series. The specific configuration may be based on actual needs. Here, the embodiment of the present invention is only based on Figure 8 The physiological mixer shown is used as an example for explanation, but it does not mean that the embodiment of the present invention is limited to this.
[0115] In combination with the above description, the multi-scale signal features of the multi-channel human physiological signals can be extracted through the signal adaptive feature modulation unit in the signal feature mixing module; the local information of the multi-channel human physiological signals can be captured through the signal mixer in the signal feature mixing module, thereby effectively improving the accuracy of the acquired multi-scale signal features, so that when the multi-channel human physiological signals are subsequently classified in combination with the multi-scale signal features, the accuracy of the classification results can be effectively improved.
[0116] The following is a description of the elderly care information planning system based on the Internet of Things and big data provided by the present invention. The elderly care information planning system based on the Internet of Things and big data described below and the elderly care information planning method based on the Internet of Things and big data described above can be referred to each other. Fig. 9 As shown, the system includes a feature classification module 901, a clustering module 902, a mapping module 903, a prediction module 904 and a pension planning module 905, wherein:
[0117] The feature classification module 901 is used to perform feature classification on the multi-channel human physiological signals of the target elderly user acquired through the wearable acquisition device to obtain the target physiological signal features;
[0118] The clustering module 902 is used to perform cluster analysis on the target physiological signal characteristics to obtain physiological characteristic categories, and to perform micro-state division of the physiological signal on the physiological characteristic categories to obtain micro-state division results;
[0119] A mapping module 903 is used to map the original physiological signal to the microstate category based on the microstate division result, convert the continuous physiological signal into a microstate sequence, and extract the microstate sequence features;
[0120] Prediction module 904, used to input the microstate sequence features into an emotional state prediction model to obtain an emotional calculation result output by the emotional state prediction model; the emotional state prediction model is trained based on sample microstate sequence features of the Internet of Things and big data, and emotional labels corresponding to the sample microstate sequence features;
[0121] The pension planning module 905 is used to plan pension information according to the emotion calculation result.
[0122] This embodiment uses wearable data collection devices, as well as the Internet of Things and big data emotion state prediction models to accurately predict the emotion calculation results of elderly care users, thereby performing elderly care information planning based on accurate emotion calculation results, thereby improving the accuracy of elderly care information planning.
[0123] See also Fig.10 , Fig.10 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 7 As shown, an embodiment of the present invention provides an electronic device 1000, including a memory 1010, a processor 1020, and a computer program 1011 stored in the memory 1010 and executable on the processor 1020. When the processor 1020 executes the computer program 1011, the following steps are implemented:
[0124] Perform feature classification on the multi-channel human physiological signals of the target elderly users obtained through wearable collection devices to obtain the target physiological signal features;
[0125] Performing cluster analysis on the target physiological signal characteristics to obtain physiological characteristic categories, and performing micro-state division of physiological signals on the physiological characteristic categories to obtain micro-state division results;
[0126] Based on the microstate division result, the original physiological signal is mapped to the microstate category, the continuous physiological signal is converted into a microstate sequence, and the microstate sequence feature is extracted;
[0127] Inputting the microstate sequence features into an emotional state prediction model to obtain an emotional calculation result output by the emotional state prediction model; the emotional state prediction model is trained based on sample microstate sequence features of the Internet of Things and big data, and emotional labels corresponding to the sample microstate sequence features;
[0128] The pension information planning is performed according to the emotion calculation result.
[0129] See also Fig.11 , Fig.11 Schematic diagram of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. Fig.11 As shown, this embodiment provides a computer-readable storage medium 1100, on which a computer program 1111 is stored. When the computer program 1111 is executed by a processor, the following steps are implemented:
[0130] Perform feature classification on the multi-channel human physiological signals of the target elderly users obtained through wearable collection devices to obtain the target physiological signal features;
[0131] Performing cluster analysis on the target physiological signal characteristics to obtain physiological characteristic categories, and performing micro-state division of physiological signals on the physiological characteristic categories to obtain micro-state division results;
[0132] Based on the microstate division result, the original physiological signal is mapped to the microstate category, the continuous physiological signal is converted into a microstate sequence, and the microstate sequence feature is extracted;
[0133] Inputting the microstate sequence features into an emotional state prediction model to obtain an emotional calculation result output by the emotional state prediction model; the emotional state prediction model is trained based on sample microstate sequence features of the Internet of Things and big data, and emotional labels corresponding to the sample microstate sequence features;
[0134] The pension information planning is performed according to the emotion calculation result.
[0135] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0136] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0137] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0138] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0139] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0140] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0141] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A pension information planning method based on the Internet of Things and big data, characterized in that: include Perform feature classification on the multi-channel human physiological signals of the target elderly users obtained through wearable collection devices to obtain the target physiological signal features; Performing cluster analysis on the target physiological signal characteristics to obtain physiological characteristic categories, and performing micro-state division of physiological signals on the physiological characteristic categories to obtain micro-state division results; Based on the microstate division result, the original physiological signal is mapped to the microstate category, the continuous physiological signal is converted into a microstate sequence, and the microstate sequence feature is extracted; Inputting the microstate sequence features into an emotional state prediction model to obtain an emotional calculation result output by the emotional state prediction model; the emotional state prediction model is trained based on sample microstate sequence features of the Internet of Things and big data, and emotional labels corresponding to the sample microstate sequence features; Performing pension information planning based on the emotion calculation results; The method of performing feature classification on the multi-channel human physiological signals of the target elderly user acquired through the wearable acquisition device to obtain the target physiological signal features includes: Inputting the multi-channel human physiological signal into a first convolution module in a physiological signal classification model, performing dimensionality reduction processing on the multi-channel human physiological signal through the first convolution module to obtain shallow signal features; The shallow signal features are respectively input into the physiological feature mixing module and the signal feature mixing module in the physiological signal classification model, the multi-scale physiological features of the shallow signal features are extracted by the physiological feature mixing module, and the multi-scale signal features of the shallow signal features are extracted by the signal feature mixing module; Inputting the superposition result of the multi-scale physiological feature and the multi-scale signal feature into the second convolution module in the physiological signal classification model, and classifying the superposition result through the second convolution module to obtain the target physiological signal feature; The physiological feature mixing module includes a first normalization layer, a physiological adaptive feature modulation unit, a second normalization layer and a physiological mixer connected in series in sequence. The multi-scale physiological feature of the shallow signal feature is extracted by the physiological feature mixing module, including: Normalizing the shallow signal feature through the first normalization layer to obtain a first signal feature; Performing physiological adaptive feature modulation on the first signal feature by the physiological adaptive feature modulation unit to obtain a first physiological feature; Normalizing the first product of the shallow signal feature and the first physiological feature through the second normalization layer to obtain a second physiological feature; Performing physiological mixing processing on the second physiological characteristic by the physiological mixer to obtain a third physiological characteristic, and a superposition result of the third physiological characteristic and the first product is the multi-scale physiological characteristic; The physiological adaptive feature modulation unit includes a multi-scale physiological feature generation unit, a first convolution layer, and a first activation function connected in series in sequence, and the physiological adaptive feature modulation unit performs physiological adaptive feature modulation on the first signal feature to obtain a first physiological feature, including: Extracting a plurality of physiological characteristic signals corresponding to the first signal characteristic by the multi-scale physiological characteristic generating unit; Convolving the concatenated result of the plurality of physiological characteristic signals through the first convolution layer to obtain a convolved physiological characteristic signal; Performing a nonlinear transformation on the convolved physiological characteristic signal through the first activation function to obtain a transformed physiological characteristic signal, wherein the product of the transformed physiological characteristic signal and the first signal characteristic is the first physiological characteristic; The multi-scale physiological feature generation unit includes a first deep convolution layer and a plurality of physiological feature generation branches connected in parallel, each physiological feature generation branch includes an adaptive average pooling layer, a second deep convolution layer and a nearest neighbor upsampling layer connected in series in sequence, the adaptive average pooling layer included in each physiological feature generation branch has a different pooling resolution, and the extraction of the plurality of physiological feature signals corresponding to the first signal feature by the multi-scale physiological feature generation unit includes: Performing deep convolution on the first signal feature through the first deep convolution layer to obtain a first physiological feature signal; Generate branches for each of the physiological characteristics, perform pooling processing on the first signal characteristics through the adaptive average pooling layer to obtain the pooled physiological characteristics; perform deep convolution on the pooled physiological characteristics through the second deep convolution layer to obtain the deep convolution physiological characteristics; upsample the deep convolution physiological characteristics through the nearest neighbor upsampling layer to obtain a second physiological characteristic signal; Wherein, the multiple physiological characteristic signals include the first physiological characteristic signal and the second physiological characteristic signal obtained by each physiological characteristic generation branch; The signal feature mixing module includes a third normalization layer, a signal adaptive feature modulation unit, a fourth normalization layer and a signal mixer connected in series in sequence, and the multi-scale signal feature of the shallow signal feature is extracted by the signal feature mixing module, including: Normalizing the shallow signal feature through the third normalization layer to obtain a second signal feature; Performing signal adaptive feature modulation on the second signal feature by the signal adaptive feature modulation unit to obtain a first signal feature; Normalizing the second product of the shallow signal feature and the first signal feature through the fourth normalization layer to obtain a second signal feature; The second signal feature is subjected to physiological mixing processing by the signal mixer to obtain a third signal feature, and a superposition result of the third signal feature and the second product is the multi-scale signal feature.
2. The method for planning pension information based on the Internet of Things and big data according to claim 1 is characterized in that: The performing pension information planning according to the emotion calculation result includes: Target elderly care users with the same emotion calculation results are classified to obtain multiple elderly care user groups; Plan each group of elderly care users to a different nursing home.
3. A pension information planning system based on the Internet of Things and big data, characterized in that: include: A feature classification module is used to perform feature classification on multi-channel human physiological signals of target elderly users acquired through wearable acquisition devices to obtain target physiological signal features; A clustering division module is used to perform cluster analysis on the target physiological signal characteristics to obtain physiological characteristic categories, and to divide the physiological characteristic categories into microstates of physiological signals to obtain microstate division results; A mapping module, used to map the original physiological signal to the microstate category based on the microstate division result, convert the continuous physiological signal into a microstate sequence, and extract the microstate sequence features; A prediction module, used to input the microstate sequence features into an emotional state prediction model to obtain an emotional calculation result output by the emotional state prediction model; the emotional state prediction model is trained based on sample microstate sequence features of the Internet of Things and big data, and emotional labels corresponding to the sample microstate sequence features; A pension planning module, used for planning pension information according to the emotion calculation results; The method of performing feature classification on the multi-channel human physiological signals of the target elderly user acquired through the wearable acquisition device to obtain the target physiological signal features includes: Inputting the multi-channel human physiological signal into a first convolution module in a physiological signal classification model, performing dimensionality reduction processing on the multi-channel human physiological signal through the first convolution module to obtain shallow signal features; The shallow signal features are respectively input into the physiological feature mixing module and the signal feature mixing module in the physiological signal classification model, the multi-scale physiological features of the shallow signal features are extracted by the physiological feature mixing module, and the multi-scale signal features of the shallow signal features are extracted by the signal feature mixing module; Inputting the superposition result of the multi-scale physiological feature and the multi-scale signal feature into the second convolution module in the physiological signal classification model, and classifying the superposition result through the second convolution module to obtain the target physiological signal feature; The physiological feature mixing module includes a first normalization layer, a physiological adaptive feature modulation unit, a second normalization layer and a physiological mixer connected in series in sequence. The multi-scale physiological feature of the shallow signal feature is extracted by the physiological feature mixing module, including: Normalizing the shallow signal feature through the first normalization layer to obtain a first signal feature; Performing physiological adaptive feature modulation on the first signal feature by the physiological adaptive feature modulation unit to obtain a first physiological feature; Normalizing the first product of the shallow signal feature and the first physiological feature through the second normalization layer to obtain a second physiological feature; Performing physiological mixing processing on the second physiological characteristic by the physiological mixer to obtain a third physiological characteristic, and a superposition result of the third physiological characteristic and the first product is the multi-scale physiological characteristic; The physiological adaptive feature modulation unit includes a multi-scale physiological feature generation unit, a first convolution layer, and a first activation function connected in series in sequence, and the physiological adaptive feature modulation unit performs physiological adaptive feature modulation on the first signal feature to obtain a first physiological feature, including: Extracting a plurality of physiological characteristic signals corresponding to the first signal characteristic by the multi-scale physiological characteristic generating unit; Convolving the concatenated result of the plurality of physiological characteristic signals through the first convolution layer to obtain a convolved physiological characteristic signal; Performing a nonlinear transformation on the convolved physiological characteristic signal through the first activation function to obtain a transformed physiological characteristic signal, wherein the product of the transformed physiological characteristic signal and the first signal characteristic is the first physiological characteristic; The multi-scale physiological feature generation unit includes a first deep convolution layer and a plurality of physiological feature generation branches connected in parallel, each physiological feature generation branch includes an adaptive average pooling layer, a second deep convolution layer and a nearest neighbor upsampling layer connected in series in sequence, the adaptive average pooling layer included in each physiological feature generation branch has a different pooling resolution, and the extraction of the plurality of physiological feature signals corresponding to the first signal feature by the multi-scale physiological feature generation unit includes: Performing deep convolution on the first signal feature through the first deep convolution layer to obtain a first physiological feature signal; Generate branches for each of the physiological characteristics, perform pooling processing on the first signal characteristics through the adaptive average pooling layer to obtain the pooled physiological characteristics; perform deep convolution on the pooled physiological characteristics through the second deep convolution layer to obtain the deep convolution physiological characteristics; upsample the deep convolution physiological characteristics through the nearest neighbor upsampling layer to obtain a second physiological characteristic signal; Wherein, the multiple physiological characteristic signals include the first physiological characteristic signal and the second physiological characteristic signal obtained by each physiological characteristic generation branch; The signal feature mixing module includes a third normalization layer, a signal adaptive feature modulation unit, a fourth normalization layer and a signal mixer connected in series in sequence, and the multi-scale signal feature of the shallow signal feature is extracted by the signal feature mixing module, including: Normalizing the shallow signal feature through the third normalization layer to obtain a second signal feature; Performing signal adaptive feature modulation on the second signal feature by the signal adaptive feature modulation unit to obtain a first signal feature; Normalizing the second product of the shallow signal feature and the first signal feature through the fourth normalization layer to obtain a second signal feature; The second signal feature is subjected to physiological mixing processing by the signal mixer to obtain a third signal feature, and a superposition result of the third signal feature and the second product is the multi-scale signal feature.
4. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing the pension information planning method based on the Internet of Things and big data as described in claim 1 or 2.
5. A non-transitory computer-readable storage medium, characterized in that: The storage medium stores a computer software program, and when the computer software program is executed by the processor, the pension information planning method based on the Internet of Things and big data as claimed in claim 1 or 2 is implemented.
