Physiological signal processing, model training and prediction method and device based on human factor intelligence, edge device and medium

Abstract

This invention discloses a method, device, edge device, and medium for physiological signal processing, model training, and prediction based on human-centric intelligence. The physiological signal processing method based on human-centric intelligence includes: acquiring reference physiological data; extracting multiple data segments from the reference physiological data; extracting features from the multiple data segments to obtain multiple initial feature data; and fusing the multiple initial feature data to obtain baseline features, wherein the baseline features are used as reference features for physiological state prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical fields of computer vision, human intelligence, etc., and in particular to a physiological signal processing method, a model training and prediction method and device based on human intelligence, an edge device and a medium. BACKGROUND

[0002] Physiological signals have been widely concerned by researchers due to their ability to objectively and truly reflect the physiological, mental and emotional states of individuals. However, how to process physiological signal data to obtain the required key feature information has become a research focus.

[0003] In related technologies, a data segment is directly extracted from the collected physiological signal for related processing to obtain baseline features. The baseline features depend on the experience of the individual in selecting the data segment, resulting in low accuracy and reference of the baseline features. Subsequent related data processing based on the baseline features will affect the accuracy of the processing results. SUMMARY

[0004] The embodiments of the present application aim to at least solve one of the technical problems in the related art. To this end, one object of the present application is to provide a physiological signal processing method based on human intelligence, a physiological state model training method, a physiological state prediction method, a device, an edge device and a storage medium, to improve the accuracy and reference of baseline features.

[0005] The embodiments of the present application provide a physiological signal processing method based on human intelligence, which comprises: obtaining reference physiological data; extracting a plurality of data segments from the reference physiological data; performing feature extraction on the plurality of data segments respectively to obtain a plurality of initial feature data; and fusing the plurality of initial feature data to obtain baseline features, wherein the baseline features are used as reference features for physiological state prediction.

[0006] Exemplarily, each initial feature data includes I sub-feature data, I being an integer greater than or equal to 1; fusing the plurality of initial feature data to obtain baseline features comprises: for each i-th sub-feature data corresponding to each initial feature data, processing a plurality of i-th sub-feature data corresponding to the plurality of initial feature data to obtain a feature value, wherein 1≤i≤I; and obtaining baseline features based on the I feature values.

[0007] Exemplarily, the reference physiological data includes data corresponding to K acquisition time points, K being an integer greater than 1; and the plurality of data segments are extracted from the reference physiological data, including: extracting each data segment from the reference physiological data according to a preset data length k, wherein 1≤k≤K, the data length of each data segment is the preset data length k, and any two data segments in the plurality of data segments have repetition or do not have repetition.

[0008] Exemplarily, the reference physiological data includes M sequence data, M being an integer greater than or equal to 1, and each sequence data being acquired by a corresponding data acquisition channel; and the plurality of data segments are extracted from the reference physiological data, including: extracting part of the sequence data from each sequence data to obtain M sub-sequence data, and the M sub-sequence data forming a data segment.

[0009] Exemplarily, the reference physiological data includes N groups of data, N being an integer greater than or equal to 1, and each group of data being acquired by data acquisition on a corresponding measured person; each group of data includes M sequence data, M being an integer greater than or equal to 1, and each sequence data being acquired by a corresponding data acquisition channel; and the plurality of data segments are extracted from the reference physiological data, including: for the M×N sequence data, extracting part of the sequence data from each sequence data to obtain M×N sub-sequence data, and the M×N sub-sequence data forming a data segment.

[0010] Another embodiment of the present application provides a physiological state model training method, which includes: acquiring physiological sample data, wherein the physiological sample data includes a state label; performing feature extraction on the physiological sample data to obtain sample feature data; performing standardization processing on the sample feature data based on baseline features to obtain standardized sample feature data, wherein the baseline features are obtained based on the above-mentioned physiological signal processing method based on human intelligence; inputting the standardized sample feature data into a to-be-trained physiological state model to perform state prediction and obtain a physiological state prediction result; and adjusting model parameters of the to-be-trained physiological state model based on the physiological state prediction result and the state label.

[0011] Another embodiment of the present application provides a physiological state prediction method, which includes: acquiring to-be-predicted physiological data; performing feature extraction on the to-be-predicted physiological data to obtain target feature data; performing standardization processing on the target feature data based on baseline features to obtain standardized target feature data, wherein the baseline features are obtained based on the above-mentioned physiological signal processing method based on human intelligence; inputting the standardized target feature data into a trained physiological state model to perform state prediction and obtain a physiological state prediction result.

[0012] Another embodiment of the present application provides a physiological signal processing device based on human factor intelligence, the physiological signal processing device based on human factor intelligence comprises: a first acquisition module configured to acquire reference physiological data; a first extraction module configured to extract a plurality of data segments from the reference physiological data, and perform feature extraction on the plurality of data segments to obtain a plurality of initial feature data; and a fusion module configured to fuse the plurality of initial feature data to obtain baseline features, wherein the baseline features are used as reference features for physiological state prediction.

[0013] Another embodiment of the present application provides a physiological state model training device, the physiological state model training device comprises: a second acquisition module configured to acquire physiological sample data, wherein the physiological sample data comprises a state label; a second extraction module configured to perform feature extraction on the physiological sample data to obtain sample feature data; a first processing module configured to perform standardization processing on the sample feature data based on the baseline features to obtain standardized sample feature data, wherein the baseline features are obtained based on the physiological signal processing device based on human factor intelligence; a first prediction module configured to input the standardized sample feature data into a physiological state model to be trained to perform state prediction and obtain a physiological state prediction result; and an adjustment module configured to adjust model parameters of the physiological state model to be trained based on the physiological state prediction result and the state label.

[0014] Another embodiment of the present application provides a physiological state prediction device, the physiological state prediction device comprises: a third acquisition module configured to acquire physiological data to be predicted; a third extraction module configured to perform feature extraction on the physiological data to be predicted to obtain target feature data; a second processing module configured to perform standardization processing on the target feature data based on the baseline features to obtain standardized target feature data, wherein the baseline features are obtained based on the physiological signal processing device based on human factor intelligence; and a second prediction module configured to input the standardized target feature data into a trained physiological state model to perform state prediction and obtain a physiological state prediction result.

[0015] Another embodiment of the present application provides an edge device comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method of any of the above embodiments when executing the computer program.

[0016] An embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the method of any of the above embodiments when executed by a processor.

[0017] In the above embodiments, the physiological signal processing method based on human factors intelligence includes: acquiring reference physiological data; extracting multiple data segments from the reference physiological data; performing feature extraction on each of the multiple data segments to obtain multiple initial feature data; and fusing the multiple initial feature data to obtain baseline features, wherein the baseline features are used as reference features for physiological state prediction. This method improves the accuracy and reliability of baseline features by extracting features from multiple data segments to obtain multiple initial feature data and then fusing the multiple initial feature data to obtain baseline features.

[0018] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0019] Figure 1 A flowchart of a physiological signal processing method based on human factors intelligence provided for an embodiment of the present invention;

[0020] Figure 2 A schematic diagram of resting-state physiological signal feature extraction of the brain provided for an embodiment of the present invention;

[0021] Figure 3 A schematic diagram illustrating the acquisition of multiple bodily physiological signal data provided in an embodiment of the present invention;

[0022] Figure 4 Flowchart of a physiological state model training method provided for another embodiment of the present invention;

[0023] Figure 5 A schematic diagram of brain task-state physiological signal feature extraction provided for another embodiment of the present invention;

[0024] Figure 6 Flowchart of a physiological state prediction method provided in another embodiment of the present invention;

[0025] Figure 7 A block diagram of a physiological signal processing device based on human factors intelligence, provided for another embodiment of the present invention;

[0026] Figure 8 Block diagram of a training device for a physiological state model provided in another embodiment of the present invention;

[0027] Figure 9 A block diagram of a physiological state prediction device provided for another embodiment of the present invention. Detailed Implementation

[0028] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0029] Physiological signals have garnered widespread attention from researchers because they objectively and accurately reflect an individual's physiological, mental, and emotional states. However, how to process physiological signal data to obtain the necessary key feature information has become a research focus.

[0030] When using machine learning methods to obtain baseline features, one can acquire physiological signal data of the brain over a period of time and extract features. The baseline features are then obtained by subtracting or dividing the brain's task-state data from the baseline features, or by subtracting and then dividing. Alternatively, purely offline methods can be used to obtain baseline features by extracting data fragments from a period prior to the event and extracting their features. However, the selection of these data fragments is susceptible to instability and other factors, and the selection of these fragments may be biased towards personal experience, thus affecting the results of subsequent data processing based on the baseline features.

[0031] Therefore, this invention provides a physiological signal processing method based on human factors intelligence. By extracting features from multiple data segments, multiple initial feature data are obtained. The multiple initial feature data are then fused to obtain baseline features. This method solves the bias problems caused by human experience and insufficient baseline feature extraction samples. It is more suitable for use in real-time operations for different time periods of the same subject and for cross-subject situations. It is also more stable and improves the accuracy and reference value of baseline features.

[0032] This article uses electroencephalogram (EEG) data as an example to analyze physiological signal processing methods based on human-centric intelligence. EEG data is divided into resting-state data and task-state data. Task-state data mainly refers to the brain's state when performing specific tasks such as memory, recognition, and movement, while resting-state data refers to the brain's state when it is not performing specific cognitive tasks, remaining quiet, relaxed, and awake. Resting-state data represents the fundamental and essential state among the various complex states of the brain.

[0033] Figure 1 A flowchart of a physiological signal processing method based on human factors intelligence provided for an embodiment of the present invention;

[0034] like Figure 1 As shown, the physiological signal processing method 100 based on human intelligence includes steps S110 to S140.

[0035] Step S110: Obtain reference physiological data.

[0036] For example, physiological signals of an individual are collected to obtain reference physiological data, which may include, for example, physiological signal data such as electroencephalogram (EEG), electrocardiogram (ECG), skin conductance, electromyography (EMG), blood oxygenation, respiration, and pupillary light.

[0037] Step S120: Extract multiple data fragments from the reference physiological data.

[0038] For example, multiple data points are extracted from reference physiological data according to the corresponding rules to form multiple data segments.

[0039] Step S130: Extract features from multiple data segments to obtain multiple initial feature data.

[0040] For example, for multiple data segments, feature extraction is performed on each data segment to obtain multiple feature data, which are used as initial feature data.

[0041] Step S140: Multiple initial feature data are fused to obtain baseline features, which are used as reference features for physiological state prediction.

[0042] For example, for the brain, baseline features usually refer to the feature data obtained by feature extraction and fusion processing of the brain's resting state data. When predicting physiological state (usually task state), it is necessary to extract features from the task state data, perform correlation operations between the obtained feature data and the baseline features, and input the results into the relevant model to obtain the predicted physiological state.

[0043] In the above embodiments, multiple initial feature data are obtained by extracting features from multiple data segments, and the multiple initial feature data are fused to obtain baseline features. This method solves the bias problems caused by human experience and too few baseline feature extraction samples, and improves the accuracy and reference value of baseline features.

[0044] Figure 2 A schematic diagram illustrating the extraction of resting-state physiological signal features of the brain, provided for an embodiment of the present invention.

[0045] The reference physiological data includes M sequence data, where M is an integer greater than or equal to 1. Each sequence data is acquired through a corresponding data acquisition channel. Multiple data segments are extracted from the reference physiological data, including: extracting partial sequence data from each sequence data to obtain M sub-sequence data, and the M sub-sequence data forming a data segment.

[0046] For example, such as Figure 2 As shown in the figure, the horizontal axis represents the time of physiological data collection, and the vertical axis represents the channel. Reference physiological data is obtained through different acquisition channels. For example, M sequence data are obtained through M channels.Figure 2 Taking the reference physiological data as an example, which includes 14 sequence data (i.e., M=14), each sequence data corresponds to a data acquisition channel.

[0047] by Figure 2 For example, six data segments are extracted from the reference physiological data, namely data segment 01, data segment 02, data segment 03, data segment 04, data segment 05, and data segment 06. Each data segment extracts a portion of the sequence data from each sequence data. Taking data segment 01 as an example, data segment 01 obtains a portion of the sequence data from each of the 14 sequence data, thus obtaining 14 sub-sequence data. Data segment 01 is composed of 14 sub-sequence data.

[0048] In the above embodiments, when using a single channel for data acquisition, the cost is lower and it is easier to implement and operate because there is no need for complex multi-channel synchronization and coordination circuits. When using multiple channels for data acquisition, different types of data from multiple channels can be acquired simultaneously, greatly improving the efficiency of data acquisition and obtaining richer information for accurate analysis of the state of the person being tested.

[0049] In one example, the reference physiological data includes data corresponding to K acquisition times, where K is an integer greater than 1; multiple data segments are extracted from the reference physiological data, including: extracting each data segment from the reference physiological data according to a preset data length k, where 1≤k≤K, the data length of each data segment is the preset data length k, and there may be or may not be duplicates between any two data segments among the multiple data segments.

[0050] For example, such as Figure 2 As shown, the horizontal axis represents time, encompassing multiple acquisition moments (i.e., K acquisition moments). Six data segments are extracted from the reference physiological data: data segment 01, data segment 02, data segment 03, data segment 04, data segment 05, and data segment 06. Taking data segment 01 as an example, when extracting from M = 14 sequence data, extraction is performed according to a preset length k, resulting in 14 subsequence data of length k, thus forming data segment 01. Figure 2 As shown, there are overlapping parts among data fragments 01, 02, 03, 04, 05, and 06, but there may also be non-overlapping parts. The analysis takes the overlapping part as an example, but it is not a specific limitation.

[0051] In the above embodiments, when extracting data segments, multiple data segments are extracted according to a preset data length. This avoids the problem of insufficient sampling samples at a single moment and the problem of mismatched data lengths, making it more suitable for real-time prediction scenarios. Furthermore, if there is no overlap between data segments, more and richer information can be extracted from the reference physiological data to obtain baseline features. If there is overlap between data segments, it is suitable for scenarios where the amount of reference physiological data is small.

[0052] In one embodiment, the feature data can be extracted and used as the initial feature data.

[0053] For example, feature extraction processing can be performed on multiple data segments according to preset feature indicators to obtain feature data corresponding to the preset feature indicators, and the multiple feature data corresponding to the preset feature indicators can be used as initial feature data.

[0054] Preset characteristic indicators may include, for example, total energy (δ+θ+α+β), absolute energy of delta wave, absolute energy of θ wave, absolute energy of α wave, absolute energy of β wave, Renyi entropy, wavelet transform, wavelet absolute mean, Fourier transform mean coefficient, minimum value indicator, maximum value indicator, average value indicator, variance indicator, standard deviation indicator, median difference indicator, maximum peak value of spectrum indicator, etc. Preset characteristic indicators can be set according to actual conditions, and will not be listed one by one here.

[0055] In another embodiment, a neural network can be used to extract features from multiple data segments to obtain initial feature data.

[0056] For example, the subsequence data corresponding to each data segment can be input into a neural network for feature extraction to obtain initial feature data.

[0057] Neural networks include, for example, Long Short-Term Memory (LSTM) networks, Bidirectional Long Short-Term Memory (BiLSTM) networks, and other networks capable of extracting features from time-series data.

[0058] Neural networks, as a deep learning method, possess powerful signal processing and recognition capabilities, and can be applied to the analysis of physiological signals. When using neural networks to detect physiological signals, it is first necessary to select a suitable network structure and set appropriate parameters for learning and training on the physiological data. Compared to traditional machine learning algorithms that require manual feature extraction and selection, neural networks can automatically select and extract features from the input physiological data, reducing manual costs and improving feature extraction efficiency.

[0059] In another example, each initial feature data includes I sub-feature data, where I is an integer greater than or equal to 1; multiple initial feature data are fused to obtain baseline features, including: for the i-th sub-feature data corresponding to each initial feature data, multiple i-th sub-feature data corresponding to multiple initial feature data are processed to obtain feature values, where 1≤i≤I; based on the I feature values, the baseline features are obtained.

[0060] For example, such as Figure 2 As shown, feature extraction is performed on data fragments 01, 02, 03, 04, 05, and 06 respectively, resulting in six initial feature data. Taking the initial feature data corresponding to data fragment 01 as an example, it includes I sub-feature data, namely X 1,1 X 1,2 …X 1,I-1 X 1,I By fusing the six initial feature data corresponding to the six data segments, an initial feature matrix of size 6xI can be obtained. By taking the median of each column of the initial feature matrix, the feature value of each column can be obtained, i.e., I feature values. By concatenating the I feature values, the baseline feature can be obtained.

[0061] For example, taking the data in the first column of the initial feature matrix (i.e., i=1), the median of the six sub-feature data in the first column is taken to obtain the feature value m1. Following this method, I feature values ​​can be obtained. Concatenating these I feature values ​​yields the baseline feature [m1, m2, ..., m]. I-1 ,m I ].

[0062] In the above embodiments, when processing the sub-feature data of multiple initial feature data, the median is used. Compared with the mean, the median avoids the instability of the baseline features caused by encountering extremely large or small values, and is therefore more stable. Of course, in addition to the median, other feature values ​​can also be used, such as the weighted average, median, etc.

[0063] Figure 3 This is a schematic diagram illustrating the acquisition of multiple physiological signal data provided in an embodiment of the present invention.

[0064] The reference physiological data includes N sets of data, where N is an integer greater than or equal to 1. Each set of data is obtained by collecting data from the corresponding subjects. Each set of data includes M sequence data, where M is an integer greater than or equal to 1. Each sequence data is collected through the corresponding data acquisition channel. Multiple data segments are extracted from the reference physiological data, including: for M×N sequence data, a portion of the sequence data is extracted from each sequence data to obtain M×N sub-sequence data, and the M×N sub-sequence data form a data segment.

[0065] For example, such as Figure 3 As shown in the figure, the horizontal axis represents the time of physiological data collection, and the vertical axis represents the channels. The reference physiological data in the figure consists of N groups, i.e., N subjects. Each subject includes M sequence data, which are collected from M channels respectively. Taking the extraction of data segment 01 as an example, partial sequence data are extracted from subject 1, subject 2, ... subject 01. For example, M partial sequence data are extracted from subject 1 to obtain M sub-sequence data, M partial sequence data are extracted from subject 2 to obtain M sub-sequence data, and so on, to obtain M×N sub-sequence data. The M×N sub-sequence data constitute data segment 01.

[0066] In the above embodiments, multiple sequence data from multiple subjects can be extracted and analyzed. This method is more suitable for use in real-time operations, such as when the same subject is at different time periods or when there are multiple subjects, and it is more stable. Furthermore, by increasing the physiological state data of multiple subjects, the sample size for obtaining baseline features can be increased, thereby avoiding the occurrence of inaccurate baseline features due to errors in one person's signal.

[0067] Figure 4 A flowchart of a physiological state model training method provided for another embodiment of the present invention.

[0068] like Figure 4 As shown, the physiological state model training method 400 includes steps S410 to S450.

[0069] Step S410: Obtain physiological sample data, wherein the physiological sample data includes status labels.

[0070] For example, the status labels may include physiological states such as fatigue, excitement, and happiness. The status labels can be represented by 0, 1, 2, etc., for example, 0 represents fatigue, 1 represents excitement, and 2 represents happiness.

[0071] Step S420: Extract features from physiological sample data to obtain sample feature data.

[0072] For example, partial sequence data is extracted from each sequence data of physiological sample data to obtain multiple sub-sequence data. The multiple sub-sequence data form a data segment. Feature extraction is performed on the multiple data segments to obtain sample feature data.

[0073] Step S430: Standardize the sample feature data based on the baseline features to obtain standardized sample feature data, wherein the baseline features are obtained based on any of the above-mentioned physiological signal processing methods based on human factors intelligence.

[0074] For example, Figure 5 A schematic diagram of brain task-state physiological signal feature extraction provided for another embodiment of the present invention, as shown below. Figure 5 As shown, based on the obtained sample feature data (denoted by X) and the baseline features (denoted by m), the sample feature data is divided by the baseline features (i.e., standardization) to obtain the standardized sample feature data.

[0075] Step S440: Input the standardized sample feature data into the physiological state model to be trained for state prediction and obtain the physiological state prediction result.

[0076] Step S450: Based on the physiological state prediction results and state labels, adjust the model parameters of the physiological state model to be trained.

[0077] For example, the physiological state prediction result can be represented as a sequence of 0, 1, or a combination of 0 and 1. Based on the deviation between the physiological state prediction result and the state label, the model parameters of the physiological state model to be trained are adjusted in reverse so that the deviation between the physiological state prediction result obtained in subsequent predictions and the state label is reduced. Iterative training is performed in sequence. When the deviation is less than a preset threshold or other training convergence conditions are met, training is stopped and a trained physiological state model is obtained.

[0078] In the above embodiments, the sample feature data is standardized based on baseline features. The standardized sample feature data is then input into the physiological state model to be trained for state prediction, resulting in a physiological state prediction. In this approach, because the baseline features accurately reflect the characteristics of the physiological data, the model training accuracy is also improved.

[0079] Figure 6 A flowchart of a physiological state prediction method provided in another embodiment of the present invention.

[0080] like Figure 6 As shown, the physiological state prediction method 600 includes steps S610 to S640.

[0081] Step S610: Obtain the physiological data to be predicted.

[0082] Step S620: Extract features from the physiological data to be predicted to obtain target feature data.

[0083] For example, partial sequence data is extracted from each sequence of physiological data to be predicted to obtain multiple sub-sequence data. Multiple sub-sequence data form a data segment. Feature extraction is performed on multiple data segments to obtain target feature data.

[0084] Step S630: Standardize the target feature data based on the baseline features to obtain standardized target feature data, wherein the baseline features are obtained based on any of the above-mentioned human-based intelligent physiological signal processing methods.

[0085] For example, based on the obtained target feature data and baseline features, the target feature data can be divided by the baseline features (i.e., standardization) to obtain the standardized target feature data.

[0086] Step S640: Input the standardized target feature data into the trained physiological state model to predict the state and obtain the physiological state prediction result.

[0087] In the above embodiments, the target feature data is standardized based on baseline features, and the standardized target feature data is input into a trained physiological state model for state prediction to obtain the physiological state prediction result. In this method, since the baseline features can accurately reflect the characteristics of physiological data, the accuracy of the physiological state model in state prediction is also improved.

[0088] Figure 7 A block diagram of a physiological signal processing device based on human factors intelligence, provided for another embodiment of the present invention.

[0089] This invention provides a physiological signal processing device 700 based on human-centric intelligence. Please refer to [link / reference]. Figure 7 The physiological signal processing device 700 based on human factors intelligence includes: a first acquisition module 710, a first extraction module 720, and a fusion module 730.

[0090] For example, the first acquisition module 710 is used to acquire reference physiological data.

[0091] For example, the first extraction module 720 is used to extract multiple data segments from reference physiological data, and to perform feature extraction on the multiple data segments respectively to obtain multiple initial feature data.

[0092] For example, the fusion module 730 is used to fuse multiple initial feature data to obtain baseline features, wherein the baseline features are used as reference features for predicting physiological states.

[0093] It is understood that for a detailed description of the physiological signal processing device 700 based on human factors intelligence, please refer to the description of the physiological signal processing method based on human factors intelligence above, and will not be repeated here.

[0094] Figure 8 A block diagram of a training device for a physiological state model provided in another embodiment of the present invention.

[0095] This invention provides a training device 800 for a physiological state model. Please refer to [link / reference]. Figure 8 The training device 800 for the physiological state model includes: a second acquisition module 810, a second extraction module 820, a first processing module 830, a first prediction module 840, and an adjustment module 850.

[0096] For example, the second acquisition module 810 is used to acquire physiological sample data, wherein the physiological sample data includes status labels.

[0097] For example, the second extraction module 820 is used to extract features from physiological sample data to obtain sample feature data.

[0098] For example, the first processing module 830 is used to standardize the sample feature data based on the baseline features to obtain standardized sample feature data, wherein the baseline features are obtained based on the above-mentioned human-based intelligent physiological signal processing device.

[0099] For example, the first prediction module 840 is used to input the standardized sample feature data into the physiological state model to be trained to perform state prediction and obtain the physiological state prediction result.

[0100] For example, the adjustment module 850 is used to adjust the model parameters of the physiological state model to be trained based on the physiological state prediction results and state labels.

[0101] It is understood that a detailed description of the training device 800 for the physiological state model can be found in the description of the training method for the physiological state model above, and will not be repeated here.

[0102] Figure 9 A block diagram of a physiological state prediction device provided for another embodiment of the present invention.

[0103] This invention provides a physiological state prediction device 900. Please refer to [link / reference]. Figure 9 The physiological state prediction device 900 includes: a third acquisition module 910, a third extraction module 920, a second processing module 930, and a second prediction module 940.

[0104] For example, the third acquisition module 910 is used to acquire physiological data to be predicted.

[0105] For example, the third extraction module 920 is used to extract features from the physiological data to be predicted to obtain target feature data.

[0106] For example, the second processing module 930 is used to standardize the target feature data based on the baseline features to obtain standardized target feature data, wherein the baseline features are obtained based on the above-mentioned human-based intelligent physiological signal processing device.

[0107] For example, the second prediction module 940 is used to input the standardized target feature data into the trained physiological state model to predict the state and obtain the physiological state prediction result.

[0108] It is understood that a detailed description of the physiological state prediction device 900 can be found in the description of the physiological state prediction method above, and will not be repeated here.

[0109] An embodiment of the present invention provides an edge device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the method of any of the above embodiments.

[0110] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method of any of the above embodiments.

[0111] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this invention, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0112] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0113] In the description of this invention, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this invention, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0114] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0115] Furthermore, the terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying relative importance, or implicitly specifying the number of technical features indicated in this embodiment. Therefore, features defined with terms such as "first" and "second" in the embodiments of this invention can explicitly or implicitly indicate that the embodiment includes at least one of those features. In the description of this invention, the word "multiple" means at least two or more, such as two, three, four, etc., unless otherwise explicitly specified in the embodiments.

[0116] In this invention, unless otherwise explicitly specified or limited in the embodiments, the terms "installation," "connection," "joining," and "fixing" appearing in the embodiments should be interpreted broadly. For example, a connection can be a fixed connection, a detachable connection, or an integral part; it can also be a mechanical connection, an electrical connection, etc. Of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two components, or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific implementation.

[0117] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0118] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A physiological signal processing method based on human-centric intelligence, characterized in that, The method includes: Obtain reference physiological data; Multiple data segments were extracted from the reference physiological data; Feature extraction is performed on the multiple data segments respectively to obtain multiple initial feature data; The multiple initial feature data are fused to obtain baseline features, which are used as reference features for predicting physiological states. Each initial feature data includes I sub-feature data, where I is an integer greater than or equal to 1; The process of fusing the multiple initial feature data to obtain baseline features includes: The median value of the i-th sub-feature data corresponding to the multiple initial feature data is obtained by performing a median operation, where 1≤i≤1; The baseline features are obtained by concatenating I feature values.

2. The method according to claim 1, characterized in that, The reference physiological data includes data corresponding one-to-one with K acquisition times, where K is an integer greater than 1; the extraction of multiple data segments from the reference physiological data includes: Each data segment is extracted from the reference physiological data according to a preset data length k, where 1≤k≤K, the data length of each data segment is the preset data length k, and there may be or may not be duplicates between any two data segments among the plurality of data segments.

3. The method according to claim 1, characterized in that, The reference physiological data includes M sequence data, where M is an integer greater than or equal to 1, and each sequence data is acquired through a corresponding data acquisition channel; The extraction of multiple data segments from the reference physiological data includes: Extract a portion of the sequence data from each sequence data to obtain M sub-sequence data, and the M sub-sequence data constitute a data segment.

4. The method according to claim 1, characterized in that, The reference physiological data includes N sets of data, where N is an integer greater than or equal to 1. Each set of data is obtained by collecting data from the corresponding subject. Each set of data includes M sequence data, where M is an integer greater than or equal to 1. Each sequence data is collected through a corresponding data acquisition channel. The extraction of multiple data segments from the reference physiological data includes: For M×N sequence data, a portion of the sequence data is extracted from each sequence data to obtain M×N sub-sequence data, and the M×N sub-sequence data constitute a data segment.

5. A training method for a physiological state model, characterized in that, The method includes: Acquire physiological sample data, wherein the physiological sample data includes status labels; Feature extraction is performed on the physiological sample data to obtain sample feature data; The sample feature data is standardized based on the baseline features to obtain standardized sample feature data, wherein the baseline features are obtained based on the method of any one of claims 1-4; The standardized sample feature data is input into the physiological state model to be trained for state prediction, and the physiological state prediction result is obtained. Based on the physiological state prediction results and the state labels, the model parameters of the physiological state model to be trained are adjusted.

6. A method for predicting physiological states, characterized in that, The method includes: Obtain the physiological data to be predicted; Feature extraction is performed on the physiological data to be predicted to obtain target feature data; The target feature data is standardized based on the baseline features to obtain standardized target feature data, wherein the baseline features are obtained based on the method of any one of claims 1-4; The standardized target feature data is input into the trained physiological state model to predict the state and obtain the physiological state prediction result.

7. A physiological signal processing device based on human-centric intelligence, characterized in that, The apparatus for implementing the method of any one of claims 1-6 comprises: The first acquisition module is used to acquire reference physiological data; The first extraction module is used to extract multiple data segments from the reference physiological data, and perform feature extraction on the multiple data segments to obtain multiple initial feature data. The fusion module is used to fuse the multiple initial feature data to obtain baseline features, wherein the baseline features are used as reference features for physiological state prediction.

8. An edge device comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1-6.

Images