Blood glucose signal processing method for blood glucose detection

Abstract

The invention discloses a blood glucose signal processing method for blood glucose detection, and relates to the technical field of blood glucose signal processing.The method comprises the steps that an optical blood glucose sensor and an electrochemical blood glucose sensor are used for synchronously and continuously detecting blood glucose of a target user in a prediction detection window, obtaining an optical blood glucose signal detection value sequence and an electrochemical blood glucose signal detection value sequence; obtaining an optical blood glucose signal detection value sub-sequence set and an electrochemical blood glucose signal detection value sub-sequence set; performing multi-modal interactive fusion analysis to obtain blood glucose fusion features; and calling a historical blood glucose feature set of the target user to authenticate the blood glucose fusion feature to obtain an authentication result, and if the authentication result is successful, taking the blood glucose fusion feature as a blood glucose signal processing result. The technical problems that in the prior art, the blood glucose signal processing accuracy is low, and the blood glucose condition of a user cannot be effectively reflected are solved. The technical effect of improving the reliability of blood glucose signal processing is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of blood sugar signal processing, and in particular to a blood sugar signal processing method for blood sugar detection. Background Art

[0002] Diabetes is a common metabolic disease with an increasing incidence rate year by year, affecting a large number of people worldwide. Therefore, the management of diabetes requires the ability to detect patients' blood sugar levels in real time and accurately, so that treatment plans can be adjusted in time according to blood sugar fluctuations. Traditional blood sugar testing methods mainly rely on regular blood sampling and measurement with a blood glucose meter. Although this intermittent method is simple and easy to use, it cannot provide information on the dynamic changes of blood sugar, which limits the precise treatment of diabetes.

[0003] The existing technology has the technical problem of low accuracy in processing blood glucose signals and failure to effectively reflect the user's blood glucose status. Summary of the invention

[0004] The present application provides a blood glucose signal processing method for blood glucose detection, which is used to solve the technical problem in the prior art that the blood glucose signal processing accuracy is low and cannot effectively reflect the user's blood glucose status.

[0005] In view of the above problems, the present application provides a blood sugar signal processing method for blood sugar detection, the method comprising:

[0006] Using an optical blood glucose sensor and an electrochemical blood glucose sensor to synchronously and continuously detect the blood glucose of a target user in a predicted detection window, and obtaining an optical blood glucose signal detection value sequence and an electrochemical blood glucose signal detection value sequence;

[0007] Dividing the optical blood glucose signal detection value sequence and the electrochemical blood glucose signal detection value sequence according to a preset division scale to obtain an optical blood glucose signal detection value subsequence set and an electrochemical blood glucose signal detection value subsequence set;

[0008] Performing multimodal interactive fusion analysis on the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set to obtain a blood glucose fusion feature;

[0009] The historical blood glucose feature set of the target user is called to authenticate the blood glucose fusion feature to obtain an authentication result. If the authentication result is passed, the blood glucose fusion feature is used as the blood glucose signal processing result.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] The present application uses an optical blood glucose sensor and an electrochemical blood glucose sensor to synchronously and continuously detect the target user's blood glucose in the predicted detection window, obtains an optical blood glucose signal detection value sequence and an electrochemical blood glucose signal detection value sequence, and then divides the optical blood glucose signal detection value sequence and the electrochemical blood glucose signal detection value sequence according to a preset division scale, obtains an optical blood glucose signal detection value subsequence set and an electrochemical blood glucose signal detection value subsequence set, and then performs multimodal interactive fusion analysis on the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set to obtain a blood glucose fusion feature, and then calls the target user's historical blood glucose feature set to authenticate the blood glucose fusion feature to obtain an authentication result. If the authentication result is passed, the blood glucose fusion feature is used as the blood glucose signal processing result. The technical effect of improving the quality and reliability of blood glucose signal processing is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Attached Figure 1 It is a flow chart of a blood glucose signal processing method for blood glucose detection provided by an embodiment of the present invention.

[0013] Attached Figure 2 It is a flow chart of obtaining blood glucose fusion features in a blood glucose signal processing method for blood glucose detection provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0015] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0016] Embodiment, as attached Figure 1 As shown, the present application provides a blood sugar signal processing method for blood sugar detection, wherein the method comprises:

[0017] S1: Using an optical blood glucose sensor and an electrochemical blood glucose sensor to synchronously and continuously detect the blood glucose of the target user in the predicted detection window, and obtaining an optical blood glucose signal detection value sequence and an electrochemical blood glucose signal detection value sequence;

[0018] In a possible embodiment, the optical blood glucose sensor is a device that measures blood glucose levels by detecting light absorption, scattering or reflection characteristics, and is usually capable of non-invasive or minimally invasive, continuous detection. The electrochemical blood glucose sensor is a device that measures blood glucose concentration by the current or voltage signal generated by glucose in an electrochemical reaction, and is characterized by high accuracy. The target user is an individual who undergoes blood glucose testing, usually a diabetic patient or a person who needs blood glucose testing.

[0019] The predicted detection window refers to the time range for blood glucose data collection and detection within a preset time period, which can provide continuous data support for dynamic blood glucose analysis, wherein the preset detection window can be set by a person skilled in the art. The optical blood glucose signal detection value sequence and the electrochemical blood glucose signal detection value sequence are blood glucose data recorded by optical and electrochemical sensors, respectively, organized in the form of time series.

[0020] By using optical blood glucose sensors and electrochemical blood glucose sensors to simultaneously start the target user's blood glucose level and continuously collect data in the same time period, the error caused by time offset is reduced and the data consistency is ensured. The optical sensor provides continuous blood glucose fluctuation information without invasiveness, while the electrochemical sensor supplements the accuracy of the optical signal with high-precision data. This step ensures that the collection of blood glucose data can cover the dynamic changes of the time axis, while taking into account non-invasiveness and high precision, providing high-quality data input for subsequent processing steps (such as data segmentation and fusion analysis).

[0021] By utilizing the characteristic advantages of optical blood glucose sensors and electrical blood glucose sensors, the technical effect of compensating for the possible defects of a single sensor is achieved, and the goal of reliable blood glucose detection and providing data support for the target users' blood glucose fluctuation trend analysis is achieved.

[0022] S2: dividing the optical blood glucose signal detection value sequence and the electrochemical blood glucose signal detection value sequence according to a preset division scale to obtain an optical blood glucose signal detection value subsequence set and an electrochemical blood glucose signal detection value subsequence set;

[0023] In one embodiment, the preset division scale is a time period between two adjacent divisions preset by a person skilled in the art, such as every 5 minutes or 10 minutes. The optical blood glucose signal detection value sequence and the electrochemical blood glucose signal detection value sequence are divided according to the preset division scale, thereby obtaining the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set.

[0024] Each subsequence in the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set contains continuous data within a time period of a preset division scale, which is convenient for subsequent processing and analysis. By performing subsequence division, data support is provided for subsequent more detailed analysis, so that it is convenient to extract the local change trend of blood glucose, such as fluctuation amplitude, short-term fluctuation and other information, thereby improving the accuracy of blood glucose signal processing.

[0025] S3: performing multimodal interactive fusion analysis on the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set to obtain a blood glucose fusion feature;

[0026] In a possible embodiment, the optical blood glucose sensor and the electrochemical blood glucose sensor have different advantages and disadvantages respectively. The optical blood glucose sensor has good continuous detection performance, but is easily affected by external factors and has low detection accuracy, while the electrochemical blood glucose sensor has high detection accuracy, but poor continuous detection performance. Therefore, firstly, each subsequence in the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set is respectively screened with a representative signal detection value to obtain an optical blood glucose signal screening value sequence and an electrochemical blood glucose signal screening value sequence, so as to achieve the goal of reducing the dimension of the data and improving the signal processing efficiency. Then, feature extraction is performed based on the screened data to determine the optical blood glucose screening feature and the electrochemical blood glucose screening feature within the preset detection window. The goal of feature extraction of blood glucose signals is achieved. Then, multimodal interactive fusion analysis is performed on the optical blood glucose screening feature and the electrochemical blood glucose screening feature, and the optical blood glucose screening feature and the electrochemical blood glucose screening feature are enhanced through interactive fusion analysis, and the blood glucose fusion feature is obtained by taking the mean. The technical effect of providing more reliable data support for blood glucose detection and improving the accuracy of blood glucose signal processing is achieved.

[0027] S4: calling the historical blood glucose feature set of the target user to authenticate the blood glucose fusion feature to obtain an authentication result. If the authentication result is passed, the blood glucose fusion feature is used as the blood glucose signal processing result.

[0028] In one embodiment, the historical blood sugar feature set reflects the blood sugar changes of the target user in the historical time. These features generally include the volatility, trend, mutation point, and blood sugar concentration of the blood sugar level, reflecting the blood sugar change pattern of the target user in history. After obtaining the blood sugar fusion feature, the blood sugar fusion feature is authenticated by using the historical blood sugar feature set of the target user to determine whether the blood sugar fusion feature is reliable. If the authentication result is passed, the blood sugar fusion feature is used as the blood sugar signal processing result. Thus, the technical effect of improving the accuracy of blood sugar signal processing is achieved.

[0029] Further, such as Figure 2 As shown, multimodal interactive fusion analysis is performed on the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set to obtain blood glucose fusion features. Step S3 of the embodiment of the present application also includes:

[0030] Performing concentrated iterative screening of signal detection values ​​within the subsequences of the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set, and sorting the screening results in chronological order of detection to obtain an optical blood glucose signal screening value sequence and an electrochemical blood glucose signal screening value sequence;

[0031] Using a blood glucose signal extraction network layer to extract blood glucose signal features from the optical blood glucose signal screening value sequence and the electrochemical blood glucose signal screening value sequence, to obtain optical blood glucose screening features and electrochemical blood glucose screening features;

[0032] A multimodal interactive fusion analysis is performed on the optical blood glucose screening feature and the electrochemical blood glucose screening feature to determine the blood glucose fusion feature.

[0033] In a possible embodiment, for each subsequence of the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set, noise data or abnormal values ​​are removed through centralized iterative screening, and the screening value that best represents the subsequence detection value situation in each subsequence is retained. Then, the screened screening values ​​are rearranged in chronological order to generate optical signal screening value sequences and electrochemical signal screening value sequences, thereby ensuring data integrity and timeline consistency.

[0034] In order to improve the efficiency of blood glucose signal processing, a blood glucose signal extraction network layer is used to extract features from the optical blood glucose signal screening value sequence and the electrochemical blood glucose signal screening value sequence. The network layer can capture the time dependence, nonlinear relationship and deep dynamic characteristics of the signal, thereby generating the optical blood glucose screening features and electrochemical blood glucose screening features. The optical blood glucose screening features reflect the blood glucose characteristics of the target user in a preset detection window after the target user's blood glucose is tested with an optical blood glucose sensor, including blood glucose concentration characteristics, blood glucose change rate characteristics, blood glucose fluctuation amplitude characteristics and blood glucose trend characteristics. The electrochemical blood glucose screening features reflect the blood glucose characteristics of the target user in a preset detection window after the target user's blood glucose is tested with an electrochemical blood glucose sensor, including blood glucose concentration characteristics, blood glucose change rate characteristics, blood glucose fluctuation amplitude characteristics and blood glucose trend characteristics.

[0035] Furthermore, by combining the continuity of the optical signal and the accuracy of the electrochemical signal, that is, performing multimodal interactive fusion analysis on the optical blood glucose screening feature and the electrochemical blood glucose screening feature, a more representative blood glucose fusion feature that can fully reflect the dynamic changes in the user's blood glucose is obtained. In this way, the complete process from data screening, feature extraction to fusion analysis is completed, achieving the technical effect of improving the quality and reliability of the blood glucose fusion feature.

[0036] Further, the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set are respectively subjected to concentrated iterative screening of signal detection values ​​within the subsequences, and the screening results are respectively sorted in order of detection time to obtain an optical blood glucose signal screening value sequence and an electrochemical blood glucose signal screening value sequence. Step S3 of the embodiment of the present application further includes:

[0037] Randomly extracting a first optical blood glucose signal detection value subsequence from the optical blood glucose signal detection value subsequence set;

[0038] Calculating the mean of the first optical blood glucose signal detection value subsequence to obtain a first optical blood glucose signal detection mean value;

[0039] Taking the first optical blood glucose signal detection mean value as an initial screening value, constructing an initial neighborhood in the first optical blood glucose signal detection value subsequence according to a preset centralized screening step;

[0040] Calculating the average of the first optical blood glucose signal detection values ​​in the initial neighborhood to obtain an initial neighborhood average;

[0041] Performing iterative screening authentication based on the difference between the initial neighborhood mean and the initial screening value, and if the iterative screening authentication result is passed, randomly extracting a first optical blood glucose signal detection value from the neighborhood edge of the initial neighborhood as the iterative screening value;

[0042] Iterating in the first optical blood glucose signal detection value subsequence based on the iterative screening value until the iterative screening authentication result is failure, and taking the neighborhood mean of the iterative neighborhood corresponding to the last iteration as the first optical signal screening value;

[0043] Similarly, the optical blood glucose signal detection value subsequence set is traversed to perform centralized screening of signal detection values ​​in the subsequences to obtain Q optical blood glucose signal screening values, and the optical blood glucose signal screening values ​​are arranged in order from front to back according to the detection time of the optical blood glucose signal detection value subsequences corresponding to the Q optical blood glucose signal screening values ​​to obtain the optical blood glucose signal screening value sequence, wherein Q is the number of optical blood glucose signal detection value subsequences in the optical blood glucose signal detection value subsequence set;

[0044] The electrochemical blood glucose signal detection value subsequence is subjected to centralized screening of signal detection values ​​within the subsequence to obtain the electrochemical blood glucose signal screening value sequence.

[0045] Further, iterative screening authentication is performed based on the difference between the initial neighborhood mean and the initial screening value. If the iterative screening authentication result is passed, a first optical blood glucose signal detection value is randomly extracted from the neighborhood edge of the initial neighborhood as the iterative screening value. Step S3 of the embodiment of the present application also includes:

[0046] Calculate the difference between the initial neighborhood mean and the initial screening value, and determine whether the difference is greater than or equal to a preset difference. If so, the iterative screening authentication result is passed;

[0047] If not, the iterative screening authentication result is failed, and the initial neighborhood mean is used as the first optical signal screening value.

[0048] In a possible embodiment, an optical blood glucose signal detection value subsequence is randomly selected from the optical blood glucose signal detection value subsequence set as the first optical blood glucose signal detection value subsequence. The first optical blood glucose signal detection value subsequence is iteratively screened for signal detection values ​​in the subsequence to remove noise and abnormal points, and a representative first optical blood glucose signal detection value subsequence is screened. The mean of the detection values ​​in the first optical blood glucose signal detection value subsequence is traversed and calculated to obtain the first optical blood glucose signal detection mean.

[0049] Furthermore, the first optical blood glucose signal detection mean is used as the initial screening value to construct the initial neighborhood. Preferably, the first optical blood glucose signal detection value in the first optical blood glucose signal detection value subsequence whose difference with the first optical blood glucose signal detection mean is within the preset centralized screening step is added to an initially empty set to obtain the initial neighborhood. The preset screening step is the amplitude of a single centralized iterative screening pre-set by a technician in this field.

[0050] Then, the average of the first optical blood glucose signal detection values ​​in the initial neighborhood is calculated to obtain the initial neighborhood average, wherein the initial neighborhood average reflects the average level of the optical blood glucose signal detection values ​​in the initial neighborhood.

[0051] Preferably, the difference between the initial neighborhood mean and the initial screening value is calculated. The size of the difference reflects the degree of discreteness of the data in the initial neighborhood. The larger the difference, the more discrete the data in the initial neighborhood. The data that can reflect the general situation of the first optical blood glucose signal detection value subsequence is more likely to be not in the initial neighborhood. It is further determined whether the difference is greater than or equal to the preset difference (the minimum difference for centralized iterative screening is pre-set by those skilled in the art). If so, the iterative screening authentication result is passed, indicating that centralized iterative screening needs to be continued at this time; if not, the iterative screening authentication result is not passed, indicating that centralized iterative screening should be stopped at this time. The data distribution in the initial neighborhood is relatively concentrated and has a high representativeness. At this time, the initial neighborhood mean is used as the first optical signal screening value.

[0052] Preferably, if the iterative screening authentication result is passed, a first optical blood glucose signal detection value is randomly extracted from the neighborhood edge of the initial neighborhood as the iterative screening value. Then, the iterative screening value is used as the new neighborhood center to construct the iterative screening neighborhood. Then, based on the same principle as the iterative screening authentication based on the difference between the initial neighborhood mean and the initial screening value, calculate whether the difference between the iterative screening neighborhood mean and the iterative screening value is greater than or equal to the preset difference. If so, a first optical blood glucose signal detection value is randomly extracted from the edge of the iterative screening neighborhood to update the iterative screening value, and then continue to iterate in the first optical blood glucose signal detection value subsequence according to the updated iterative screening value until the iterative screening authentication result is failed, stop the iteration, and use the neighborhood mean of the iterative neighborhood corresponding to the last iteration as the first optical signal screening value.

[0053] By analogy, based on the same principle, the signal detection values ​​in the subsequence are concentratedly screened for each optical blood glucose signal detection value subsequence set in the optical blood glucose signal detection value subsequence set to obtain Q optical blood glucose signal screening values, and the optical blood glucose signal screening values ​​are arranged in order from front to back according to the detection time of the optical blood glucose signal detection value subsequences corresponding to the Q optical blood glucose signal screening values ​​to obtain the optical blood glucose signal screening value sequence, wherein Q is the number of optical blood glucose signal detection value subsequences in the optical blood glucose signal detection value subsequence set. Furthermore, based on the same principle of obtaining the optical blood glucose signal screening value sequence, the signal detection values ​​in the subsequence are concentratedly screened for the electrochemical blood glucose signal detection value subsequence to obtain the electrochemical blood glucose signal screening value sequence.

[0054] Through centralized iterative screening, noise or outliers in optical and electrochemical signals can be effectively removed to ensure the quality and representativeness of the screening values. The refinement of the screening process is controlled by the neighborhood mean and screening step, making the screened signal sequence smoother and reducing errors caused by data fluctuations. Furthermore, by using the screening certification process to verify the rationality of the signal detection value through step-by-step iteration, it is possible to stably extract representative values ​​with statistical significance and improve the reliability of the signal analysis results. The technical effect of laying the foundation for the accuracy of blood glucose signal processing has been achieved.

[0055] Furthermore, step S3 of the embodiment of the present application also includes:

[0056] Acquire multiple sample optical blood glucose signal screening value sequences and corresponding multiple sample optical blood glucose screening features as optical training data, use the optical training data to perform supervised training on a framework built based on a convolutional neural network, learn a one-to-one mapping relationship between the optical blood glucose signal screening value sequence and the sample optical blood glucose screening features during training, until the training converges, and obtain a trained optical blood glucose signal extraction branch;

[0057] A plurality of sample electrochemical blood glucose signal screening value sequences and a plurality of corresponding sample electrochemical blood glucose screening features are obtained as electrochemical training data, and a framework based on a convolutional neural network is supervised and trained using the optical training data. During the training, a one-to-one mapping relationship between the electrochemical blood glucose signal screening value sequence and the sample electrochemical blood glucose screening features is learned until the training converges, thereby obtaining a trained electrochemical blood glucose signal extraction branch;

[0058] The optical blood glucose signal extraction branch and the electrochemical blood glucose signal extraction branch are connected in parallel to obtain the blood glucose signal extraction network layer.

[0059] In a possible embodiment, the plurality of sample optical blood glucose signal screening value sequences are sequences extracted from a plurality of sample optical blood glucose signal detection value subsequences after screening processing, and are used to train and test a neural network model. The plurality of sample optical blood glucose screening features refer to feature data corresponding to the sample optical blood glucose signal screening value sequences. The plurality of sample optical blood glucose signal screening value sequences and the corresponding plurality of sample optical blood glucose screening features are used as optical training data.

[0060] Furthermore, the convolutional neural network (CNN) framework is used for model construction. CNN automatically learns and extracts spatial and temporal features in data through multi-layer structures such as convolutional layers, pooling layers, and fully connected layers. The input of the network is an optical blood glucose signal screening value sequence, and the output is a blood glucose screening feature corresponding to each sequence. Preferably, the optical blood glucose screening features of the multiple samples are labeled. During the training process, the network is trained by calculating the difference (i.e., error) between the network output and the true label according to the input optical blood glucose signal screening value sequence (training data) and its corresponding optical blood glucose screening feature (label data). Through the back propagation algorithm, the error is back-propagated to each level in the network, and the network parameters (such as convolution kernel weights, etc.) are adjusted to minimize the error. In each iteration, the convolutional neural network updates the model parameters through an optimization algorithm (such as gradient descent) to gradually reduce the prediction error. After each round of training, the network will gradually adjust its own weights and biases so that the mapping relationship between the input optical blood glucose signal screening value sequence and the output optical blood glucose screening feature becomes more and more accurate. After multiple rounds of training, the model error tends to stabilize and reaches a minimum value or convergence state, indicating that the network has learned the optimal mapping relationship from the optical blood glucose signal screening value sequence to the blood glucose screening feature. When the training is completed and converged, the obtained model is the trained optical blood glucose signal extraction branch. The optical blood glucose signal extraction branch can automatically extract the key features in the optical blood glucose signal sequence and map them one-to-one with the blood glucose screening features.

[0061] Based on the same construction principle as the optical blood glucose signal extraction branch, multiple sample electrochemical blood glucose signal screening value sequences and corresponding multiple sample electrochemical blood glucose screening features are used as electrochemical training data, and the framework constructed based on the convolutional neural network is supervised and trained using the optical training data. During the training, the one-to-one mapping relationship between the electrochemical blood glucose signal screening value sequence and the sample electrochemical blood glucose screening features is learned until the training converges, and a trained electrochemical blood glucose signal extraction branch is obtained. The electrochemical blood glucose signal extraction branch can automatically extract the key features in the electrochemical blood glucose signal sequence and map them one-to-one with the blood glucose screening features.

[0062] Furthermore, the trained optical blood glucose signal extraction branch and the electrochemical blood glucose signal extraction branch are combined to form a multi-input network layer (blood glucose signal extraction network layer), which can process two different types of signals at the same time and extract their blood glucose-related features.

[0063] Further, a multimodal interactive fusion analysis is performed on the optical blood glucose screening feature and the electrochemical blood glucose screening feature to determine the blood glucose fusion feature. Step S3 of the embodiment of the present application further includes:

[0064] Calculating the similarity between the optical blood glucose screening feature and the electrochemical blood glucose screening feature to obtain a screening feature similarity set;

[0065] Normalizing the screening feature similarity set, and adding the processing result into an initially empty matrix to obtain a multimodal interaction fusion matrix;

[0066] Performing convolution calculation on the optical blood glucose screening feature and the multimodal interactive fusion matrix to obtain an optical blood glucose fusion feature;

[0067] Performing convolution calculation on the electrochemical blood glucose screening feature and the multimodal interactive fusion matrix to obtain an electrochemical blood glucose fusion feature;

[0068] The average of the optical blood glucose fusion feature and the electrochemical blood glucose fusion feature is calculated to obtain the blood glucose fusion feature.

[0069] Furthermore, step S3 of the embodiment of the present application also includes:

[0070] Obtain a normalization function, wherein the normalization function is:

[0071]

[0072] Among them, Mix[lim(x i ,y i )] is the normalized value corresponding to the i-th screening feature similarity in the screening feature similarity set, e is the base of the natural logarithm, n is the total number of screening feature similarities in the screening feature similarity set, lim(x i ,y i ) is the i-th screening feature similarity in the screening feature similarity set;

[0073] The normalization function is used to perform normalization processing on the screening feature similarity set to obtain the processing result.

[0074] In a possible embodiment, in order to perform multimodal interactive fusion analysis on the optical blood glucose screening feature and the electrochemical blood glucose screening feature, the optical blood glucose screening feature and the electrochemical blood glucose screening feature are mutually enhanced. First, the similarities of the corresponding features in the optical blood glucose screening feature and the electrochemical blood glucose screening feature are calculated respectively using the cosine calculation formula to obtain the screening feature similarity set. Among them, the screening feature similarity set reflects the similarity between the optical blood glucose screening feature and the electrochemical blood glucose screening feature.

[0075] Then, the screening feature similarity set is normalized using a normalization function to eliminate dimension differences, and the processing result is added to the initially empty matrix to obtain the multimodal interaction fusion matrix. The multimodal interaction fusion matrix reflects the interaction relationship between the optical blood glucose screening feature and the multimodal interaction fusion matrix.

[0076] Convolution operation is performed on the optical blood glucose screening feature and the multimodal interaction fusion matrix to extract the high-order interaction features between the two and generate the optical blood glucose fusion feature. Preferably, a set of sample optical blood glucose screening features, a sample multimodal interaction fusion matrix and a sample optical blood glucose fusion feature are obtained as training data, and the training data is used to perform supervised training on a framework built based on a convolutional neural network until the training converges to obtain a trained optical convolutional network layer. The optical convolutional network layer is used to perform convolution analysis on the optical blood glucose screening feature and the multimodal interaction fusion matrix to obtain the optical blood glucose fusion feature.

[0077] Similarly, the electrochemical blood glucose screening feature is convolved with the multimodal interactive fusion matrix to extract its interactive features and generate an electrochemical blood glucose fusion feature. Preferably, a set of sample electrochemical blood glucose screening features, a sample multimodal interactive fusion matrix and a sample electrochemical blood glucose fusion feature are obtained as training data, and the training data is used to perform supervised training on a framework built based on a convolutional neural network until the training converges to obtain a trained electrochemical convolutional network layer. The electrochemical convolutional network layer is used to perform convolution analysis on the electrochemical blood glucose screening feature and the multimodal interactive fusion matrix to obtain the electrochemical blood glucose fusion feature. Through convolution calculation, deep interactive relationships are extracted from the input features and the interaction matrix to provide a more accurate and comprehensive feature expression for blood glucose detection and prediction.

[0078] Further, the historical blood glucose feature set of the target user is called to authenticate the blood glucose fusion feature to obtain an authentication result. If the authentication result is passed, the blood glucose fusion feature is used as the blood glucose signal processing result. Step S4 of the embodiment of the present application also includes:

[0079] Calculate similarity between the blood glucose fusion feature and the historical blood glucose feature set to obtain a historical similarity set;

[0080] The historical similarity set is authenticated using a preset historical similarity threshold. When the ratio of the number of historical similarities in the historical similarity set that is less than the preset historical similarity threshold to the total number of historical similarities in the historical similarity set is less than a preset ratio, the authentication result is passed.

[0081] In a possible embodiment, the historical blood glucose feature set is accumulated from the target user's past blood glucose test data, and includes blood glucose features at several historical time points. Each historical blood glucose feature is composed of data of multiple dimensions, representing the user's blood glucose status at a certain moment. The blood glucose fusion feature is the result of processing by fusing optical and electrochemical blood glucose signals, representing the current blood glucose status of the target user. This feature set is usually also multidimensional data.

[0082] The distance between the blood glucose fusion feature and each historical blood glucose feature in the historical blood glucose feature set is calculated using the Euclidean distance. The smaller the distance, the higher the similarity. The inverse of the calculation result is used as the historical similarity set. The historical similarity set reflects the similarity between the blood glucose fusion feature and the historical blood glucose status of the target user.

[0083] Furthermore, a preset historical similarity threshold (e.g., 0.6) is set by a person skilled in the art, and the value is used to determine whether the similarity between the current blood glucose fusion feature and the historical blood glucose feature is high enough. If the similarity is higher than the threshold, it means that the current blood glucose state is very similar to the historical blood glucose state, and if it is lower than the threshold, it means that the blood glucose state has changed significantly, and there may be a blood glucose signal processing error.

[0084] When the ratio of the number of similarities in the historical similarity set that is less than the preset historical similarity threshold to the total number of historical similarities in the historical similarity set is less than the preset ratio, the authentication result is passed. Preferably, the ratio calculation formula can be: ratio = the number of similarities in the historical similarity set that is less than the preset historical similarity threshold / the total number of historical similarity sets. The ratio represents the degree of matching between the current blood glucose fusion feature and the historical blood glucose feature.

[0085] If the ratio is less than the preset ratio (e.g. 0.6), the authentication result is passed, indicating that the current blood glucose fusion feature matches the historical blood glucose feature sufficiently well and meets the processing requirements. If the ratio is greater than the preset ratio, the authentication fails, indicating that the current blood glucose fusion feature differs greatly from the historical blood glucose feature and requires further processing.

[0086] In summary, the embodiments of the present application have at least the following technical effects:

[0087] 1. This application utilizes dual detection of optical blood glucose sensor and electrochemical blood glucose sensor, which can complement each other and reduce the errors that may be caused by a single sensor. By performing multimodal interactive fusion analysis, useful information in multi-source data can be better extracted, thereby achieving the technical effect of further improving the accuracy of blood glucose signal processing.

[0088] 2. This application authenticates the blood glucose fusion feature by calling the target user's historical blood glucose feature set to ensure the consistency of the current blood glucose signal with the historical data. This process can effectively avoid erroneous analysis caused by emergencies or abnormal fluctuations, ensure the accurate reflection of blood glucose signals, and make blood glucose processing results more reliable.

[0089] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0090] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0091] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A blood sugar signal processing method for blood sugar detection, characterized in that: The method comprises: Using an optical blood glucose sensor and an electrochemical blood glucose sensor to synchronously and continuously detect the blood glucose of a target user in a predicted detection window, and obtaining an optical blood glucose signal detection value sequence and an electrochemical blood glucose signal detection value sequence; Dividing the optical blood glucose signal detection value sequence and the electrochemical blood glucose signal detection value sequence according to a preset division scale to obtain an optical blood glucose signal detection value subsequence set and an electrochemical blood glucose signal detection value subsequence set; Performing multimodal interactive fusion analysis on the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set to obtain a blood glucose fusion feature; The historical blood glucose feature set of the target user is called to authenticate the blood glucose fusion feature to obtain an authentication result. If the authentication result is passed, the blood glucose fusion feature is used as the blood glucose signal processing result.

2. A blood sugar signal processing method for blood sugar detection according to claim 1, characterized in that: Performing multimodal interactive fusion analysis on the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set to obtain blood glucose fusion features, including: Performing concentrated iterative screening of signal detection values ​​within the subsequences of the optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set, and sorting the screening results in chronological order of detection to obtain an optical blood glucose signal screening value sequence and an electrochemical blood glucose signal screening value sequence; Using a blood glucose signal extraction network layer to extract blood glucose signal features from the optical blood glucose signal screening value sequence and the electrochemical blood glucose signal screening value sequence, to obtain optical blood glucose screening features and electrochemical blood glucose screening features; A multimodal interactive fusion analysis is performed on the optical blood glucose screening feature and the electrochemical blood glucose screening feature to determine the blood glucose fusion feature.

3. A blood sugar signal processing method for blood sugar detection as claimed in claim 2, characterized in that: The optical blood glucose signal detection value subsequence set and the electrochemical blood glucose signal detection value subsequence set are respectively subjected to concentrated iterative screening of signal detection values ​​within the subsequences, and the screening results are respectively sorted in order of detection time to obtain an optical blood glucose signal screening value sequence and an electrochemical blood glucose signal screening value sequence, including: Randomly extracting a first optical blood glucose signal detection value subsequence from the optical blood glucose signal detection value subsequence set; Calculating the mean of the first optical blood glucose signal detection value subsequence to obtain a first optical blood glucose signal detection mean value; Taking the first optical blood glucose signal detection mean value as an initial screening value, constructing an initial neighborhood in the first optical blood glucose signal detection value subsequence according to a preset centralized screening step; Calculating the average of the first optical blood glucose signal detection values ​​in the initial neighborhood to obtain an initial neighborhood average; Performing iterative screening authentication based on the difference between the initial neighborhood mean and the initial screening value, and if the iterative screening authentication result is passed, randomly extracting a first optical blood glucose signal detection value from the neighborhood edge of the initial neighborhood as the iterative screening value; Iterating in the first optical blood glucose signal detection value subsequence based on the iterative screening value until the iterative screening authentication result is failure, and taking the neighborhood mean of the iterative neighborhood corresponding to the last iteration as the first optical signal screening value; Similarly, the optical blood glucose signal detection value subsequence set is traversed to perform centralized screening of signal detection values ​​in the subsequences to obtain Q optical blood glucose signal screening values, and the optical blood glucose signal screening values ​​are arranged in order from front to back according to the detection time of the optical blood glucose signal detection value subsequences corresponding to the Q optical blood glucose signal screening values ​​to obtain the optical blood glucose signal screening value sequence, wherein Q is the number of optical blood glucose signal detection value subsequences in the optical blood glucose signal detection value subsequence set; The electrochemical blood glucose signal detection value subsequence is subjected to centralized screening of signal detection values ​​within the subsequence to obtain the electrochemical blood glucose signal screening value sequence.

4. A blood sugar signal processing method for blood sugar detection as claimed in claim 3, characterized in that: Iterative screening authentication is performed based on the difference between the initial neighborhood mean and the initial screening value. If the iterative screening authentication result is passed, a first optical blood glucose signal detection value is randomly extracted from the neighborhood edge of the initial neighborhood as the iterative screening value, including: Calculate the difference between the initial neighborhood mean and the initial screening value, and determine whether the difference is greater than or equal to a preset difference. If so, the iterative screening authentication result is passed; If not, the iterative screening authentication result is failed, and the initial neighborhood mean is used as the first optical signal screening value.

5. A blood sugar signal processing method for blood sugar detection as claimed in claim 2, characterized in that: include: Acquire multiple sample optical blood glucose signal screening value sequences and corresponding multiple sample optical blood glucose screening features as optical training data, use the optical training data to perform supervised training on a framework built based on a convolutional neural network, learn a one-to-one mapping relationship between the optical blood glucose signal screening value sequence and the sample optical blood glucose screening features during training, until the training converges, and obtain a trained optical blood glucose signal extraction branch; A plurality of sample electrochemical blood glucose signal screening value sequences and a plurality of corresponding sample electrochemical blood glucose screening features are obtained as electrochemical training data, and a framework based on a convolutional neural network is supervised and trained using the optical training data. During the training, a one-to-one mapping relationship between the electrochemical blood glucose signal screening value sequence and the sample electrochemical blood glucose screening features is learned until the training converges, thereby obtaining a trained electrochemical blood glucose signal extraction branch; The optical blood glucose signal extraction branch and the electrochemical blood glucose signal extraction branch are connected in parallel to obtain the blood glucose signal extraction network layer.

6. A blood sugar signal processing method for blood sugar detection as claimed in claim 2, characterized in that: Performing multimodal interactive fusion analysis on the optical blood glucose screening feature and the electrochemical blood glucose screening feature to determine the blood glucose fusion feature includes: Calculating the similarity between the optical blood glucose screening feature and the electrochemical blood glucose screening feature to obtain a screening feature similarity set; Normalizing the screening feature similarity set, and adding the processing result into an initially empty matrix to obtain a multimodal interaction fusion matrix; Performing convolution calculation on the optical blood glucose screening feature and the multimodal interactive fusion matrix to obtain an optical blood glucose fusion feature; Performing convolution calculation on the electrochemical blood glucose screening feature and the multimodal interactive fusion matrix to obtain an electrochemical blood glucose fusion feature; The average of the optical blood glucose fusion feature and the electrochemical blood glucose fusion feature is calculated to obtain the blood glucose fusion feature.

7. A blood sugar signal processing method for blood sugar detection according to claim 6, characterized in that: include: Obtain a normalization function, wherein the normalization function is: Among them, Mix[lim(x i ,y i )] is the normalized value corresponding to the i-th screening feature similarity in the screening feature similarity set, e is the base of the natural logarithm, n is the total number of screening feature similarities in the screening feature similarity set, lim(x i ,y i ) is the i-th screening feature similarity in the screening feature similarity set; The normalization function is used to perform normalization processing on the screening feature similarity set to obtain the processing result.

8. A blood sugar signal processing method for blood sugar detection according to claim 1, characterized in that: The target user's historical blood glucose feature set is called to authenticate the blood glucose fusion feature to obtain an authentication result. If the authentication result is passed, the blood glucose fusion feature is used as a blood glucose signal processing result, including: Calculate similarity between the blood glucose fusion feature and the historical blood glucose feature set to obtain a historical similarity set; The historical similarity set is authenticated using a preset historical similarity threshold. When the ratio of the number of historical similarities in the historical similarity set that is less than the preset historical similarity threshold to the total number of historical similarities in the historical similarity set is less than a preset ratio, the authentication result is passed.

Images