Intelligent early warning system and bracelet for Parkinson / epilepsy recognition based on multi-modal sensor

Abstract

The invention discloses an intelligent pre-warning system for Parkinson / epilepsy recognition based on a multi-modal sensor and a bracelet, and relates to the technical field of biosensors. The intelligent early warning system for Parkinson / epilepsy recognition based on the multi-modal sensor comprises a multi-modal acquisition interference judgment module, a multi-modal interference processing optimization module and a multi-modal fusion input early warning module. According to the method, whether the multi-modal interference processing optimization is carried out or not is judged according to the multi-modal interference early warning result, if yes, the multi-modal fusion instruction is sent after the multi-modal interference processing optimization, and if not, multi-modal biological data fusion is directly executed, and the multi-modal fusion early warning result is obtained. Finally, whether multi-modal fusion input early warning is carried out or not is judged based on the multi-modal fusion early warning result, the effect of carrying out physical condition early warning more accurately is achieved, and the problem that in the prior art, in the process of carrying out physical condition early warning through wearable equipment, multi-modal biological data collection and fusion association are not sufficient is solved.

Description

Technical Field

[0001] This invention relates to the field of biosensor technology, and in particular to an early warning intelligent system and wristband for Parkinson's / epilepsy recognition based on multimodal sensors. Background Technology

[0002] With the development of modern medical technology, the early warning and monitoring of neurological diseases has become an important area of ​​medical research. Neurological diseases such as Parkinson's disease and epilepsy affect a large population worldwide, especially the elderly. Traditional methods of disease diagnosis and early warning often rely on patients' clinical symptoms and doctors' diagnoses, which suffer from problems such as delayed response and untimely diagnosis. However, with the development of multimodal biosensor technology, physiological signals such as motion acceleration, heart rate variability, and skin conductance levels have been proven to effectively reflect changes in physical condition and are closely related to the pathophysiological state of diseases such as Parkinson's disease and epilepsy. Therefore, by fusing this multimodal biosensor data for real-time monitoring and early warning, important technical support can be provided for the early detection, symptom management, and seizure warning of neurological diseases.

[0003] Existing technologies utilize multimodal biosensors embedded in wearable devices (such as smartwatches) to collect various types of biological data. For example, changes in skin conductivity measured by skin conductivity sensors reflect changes in sympathetic nerve activity. These changes can be used to monitor stress responses or abnormal activity, potentially aiding in the prediction of epileptic seizures. The biological data collected by wearable devices can be transmitted in real-time to mobile devices or cloud platforms for storage and processing via wireless communication technologies such as Bluetooth. Multimodal fusion technology then fuses the biological data from different modalities, and machine learning algorithms (such as support vector machines and neural networks) are used to further analyze the input multimodal fused data, thereby enabling early warning of the patient's physical condition.

[0004] For example, Chinese patent application CN120226994A discloses a wearable device for multimodal physiological data acquisition and intelligent analysis, comprising: a data acquisition module that collects raw physiological data such as heart rate and blood pressure through multiple sensors; a preprocessing module that normalizes and downsamples the data; a control analysis module that extracts and fuses feature vectors and obtains assessment results through a health assessment model; a communication module that transmits the data and results to a smart terminal; and a power management module that provides power and performs power status detection and low battery warning.

[0005] For example, Chinese invention patent CN118526162B discloses an intelligent dynamic brain function monitoring system and method based on multimodal data fusion, which includes: a data acquisition module for collecting data from the target object and acquiring multiple modal data; a data fusion module for fusing and processing the multiple modal data to acquire multiple connection matrices; and a model construction module for optimizing the initial brain function model through multiple connection matrices to acquire a target brain function model. The target brain function model is used to process the real-time fNIRS data of the target object and output the brain function monitoring index data of the target object.

[0006] The above-mentioned technology has at least the following technical problems:

[0007] Electromyographic interference (EMG) is a common and significant interfering factor in the acquisition of multimodal biological data, often leading to inaccurate data and consequently, inaccurate health condition warnings. In multimodal biological data acquisition, biosensors are often in direct contact with the skin or attached to the body surface. When muscles contract or move intentionally or unintentionally, the electric and magnetic fields generated on the muscle surface interact with the electric and magnetic fields of the sensors. This interaction, especially in patients with neurological disorders such as Parkinson's disease and epilepsy, often results in signal noise during the multimodal biological data acquisition process, thus failing to accurately reflect true physiological signals.

[0008] Furthermore, in the process of fusing multimodal biological data, if the data collected by multimodal biosensors is affected by electromyography (EMG), it often further affects the accuracy of the entire fusion process. For example, the fusion algorithm may make incorrect disease predictions or health condition warnings based on incorrect movement patterns, heart rate, or skin conductance characteristics, resulting in insufficient correlation between multimodal biological data collection and fusion during the health condition warning process through wearable devices. Summary of the Invention

[0009] To address the technical problem of insufficient multimodal biological data collection, fusion, and correlation in the process of health condition early warning through wearable devices in existing technologies, this invention provides an intelligent early warning system and wristband based on multimodal sensors for Parkinson's / epilepsy recognition. The technical solution is as follows:

[0010] On one hand, a pre-warning intelligent system for Parkinson's / epilepsy recognition based on multimodal sensors is provided, including: a multimodal acquisition interference judgment module, a multimodal interference processing optimization module, and a multimodal fusion input pre-warning module; wherein, the multimodal acquisition interference judgment module is used to obtain multimodal interference pre-warning results based on the multimodal biological data of the target object collected by the multimodal biosensors in the wearable device to quantitatively judge the degree of electromyographic interference in the multimodal biological data acquisition process; the multimodal interference processing optimization module is used to determine whether to perform multimodal interference processing optimization based on the obtained multimodal interference pre-warning results, and if so, to send the data after multimodal interference processing optimization for multimodal... The system executes a multimodal fusion command for biological data fusion, otherwise directly performs multimodal biological data fusion and obtains multimodal fusion early warning results to quantitatively judge the fusion effect of multimodal biological data. Multimodal interference processing optimization means combining the multimodal interference early warning results to perform filtering and acquisition adjustment on the multimodal biological data acquisition process. The multimodal fusion input early warning module is used to determine whether to perform multimodal fusion input early warning based on the multimodal fusion early warning results. If yes, it executes multimodal fusion input early warning; otherwise, it executes multimodal fusion processing optimization. Multimodal fusion processing optimization means combining the multimodal fusion early warning results to adjust the dimensionality reduction processing during the multimodal biological data fusion process.

[0011] On the other hand, a smart bracelet is provided, including: a bracelet body, a central processing unit, and a multimodal biosensor; the multimodal biosensor is configured in the bracelet body, and includes a 6-axis accelerometer, an optical heart rate sensor, and a skin conductance sensor; the central processing unit is located in the bracelet body; the central processing unit is used to obtain multimodal interference warning results based on the multimodal biological data collected by the multimodal biosensor, determine whether to perform multimodal interference processing optimization and send a multimodal fusion command, and determine whether to perform multimodal fusion input warning based on the obtained multimodal fusion warning results to output warning information.

[0012] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following:

[0013] 1. In the process of health condition early warning through wearable devices, electromyographic interference during the collection of multimodal biological data often leads to inaccurate multimodal biological data, which in turn affects the accuracy of the multimodal biological data fusion process, resulting in inaccurate health condition early warning. This invention uses multimodal biological data of the target object collected by multimodal biosensors in wearable devices to obtain multimodal interference early warning results. This helps to more accurately judge the degree of electromyographic interference in the multimodal biological data collection process, and then determine whether to perform multimodal interference processing optimization, thereby effectively reducing the impact of electromyographic interference on multimodal biological data collection and improving the reliability of multimodal biological data collection. Finally, based on the multimodal fusion early warning results, it determines whether to perform multimodal fusion input early warning, achieving a more accurate judgment of the multimodal biological data fusion effect. This enables the correlation analysis of the multimodal biological data collection process and the fusion process in the health condition early warning process, effectively solving the problem of insufficient correlation between multimodal biological data collection and fusion in the existing technology of health condition early warning through wearable devices.

[0014] 2. The system determines whether to perform multimodal interference processing optimization based on the multimodal interference warning results. When the multimodal interference warning result corresponds to normal interference limitation, multimodal interference processing optimization is not performed, and multimodal fusion warning results are obtained. This helps to more accurately perform correlation analysis of the multimodal biological data acquisition and fusion process. When the multimodal interference warning result corresponds to interference limitation warning, the system determines whether to perform time-domain-frequency domain interference filtering optimization based on the multimodal acquisition interference value. If so, an interference warning optimization command is sent after performing time-domain-frequency domain interference filtering optimization; otherwise, an interference warning optimization command for performing multimodal biological data warning optimization is sent directly. This effectively reduces electromyographic interference during the multimodal biological data acquisition process, thereby improving the reliability of multimodal biological data acquisition and warning during the body condition warning process using wearable devices.

[0015] 3. By comparing the multimodal fusion quality score with the preset fusion effect quality index to obtain the multimodal fusion early warning result, a more accurate judgment of the multimodal biological data fusion effect is achieved. When the multimodal fusion early warning result corresponds to an unqualified fusion effect, no multimodal fusion input early warning is issued, and multimodal fusion processing optimization is performed, which helps to improve the reliability of multimodal biological data fusion and thus achieves more accurate early warning of physical condition. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic diagram of the structure of an intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors provided in an embodiment of the present invention;

[0018] Figure 2 One of the homepage interface diagrams of the intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors provided in an embodiment of the present invention;

[0019] Figure 3 A flowchart illustrating the process of acquiring multimodal fusion early warning results for an intelligent early warning system based on multimodal sensors for Parkinson's / epilepsy identification, provided in an embodiment of the present invention.

[0020] Figure 4 A flowchart illustrating the multimodal biological data early warning optimization process of an intelligent early warning system for Parkinson's / epilepsy identification based on multimodal sensors, provided in an embodiment of the present invention.

[0021] Figure 5 This is one of the multimodal data management interface diagrams of an intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0023] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0024] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0025] This invention provides an intelligent early warning system and wristband based on multimodal sensors for Parkinson's / epilepsy recognition. This addresses the problem in existing technologies where multimodal biological data acquisition and fusion are insufficiently correlated during health condition early warning via wearable devices. By acquiring multimodal biological data of the target object using multimodal biosensors in the wearable device, a multimodal interference early warning result is obtained to quantify the degree of electromyographic interference during the multimodal biological data acquisition process. Then, based on the obtained multimodal interference early warning result, it is determined whether to perform multimodal interference processing optimization. If so, a multimodal fusion command is sent after multimodal interference processing optimization to perform multimodal biological data fusion; otherwise, multimodal biological data fusion is directly performed, and a multimodal fusion early warning result is obtained to quantify the fusion effect of the multimodal biological data. Finally, based on the multimodal fusion early warning result, it is determined whether to perform multimodal fusion input early warning. If so, the multimodal fusion input early warning is performed; otherwise, multimodal fusion processing optimization is performed, achieving more accurate health condition early warning.

[0026] The technical solution in this invention aims to address the problem of insufficient multimodal biological data collection, fusion, and correlation during the health status early warning process using wearable devices. The overall approach is as follows:

[0027] The system determines whether to perform multimodal interference processing optimization based on the multimodal interference warning results. If so, a multimodal fusion command is sent after the multimodal interference processing optimization. Otherwise, multimodal biological data fusion is performed directly and the multimodal fusion warning results are obtained. Finally, it determines whether to perform multimodal fusion input warning, thus achieving a more accurate effect of warning about physical condition.

[0028] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0029] like Figure 1 The diagram shown is a structural schematic of an intelligent early warning system for Parkinson's disease / epilepsy recognition based on multimodal sensors provided in an embodiment of the present invention. The intelligent early warning system for Parkinson's disease / epilepsy recognition based on multimodal sensors provided in an embodiment of the present invention includes: a multimodal acquisition interference judgment module, a multimodal interference processing optimization module, and a multimodal fusion input early warning module.

[0030] It's important to note that when designing the Parkinson's / epilepsy early warning intelligent system based on multimodal sensors, a dedicated multimodal physical condition early warning database was first created to store core configuration information. This database contains various reference values ​​necessary for system operation, such as reference values ​​for electromyography (EMG) signal intensity and peak EMG signal intensity. These initial settings are not arbitrarily assigned but calculated using a summation and averaging method based on a large amount of previously accumulated measured data in the multimodal physical condition early warning database, making the initial settings more objective and better reflective of general conditions. Of course, considering the complex and variable nature of real-world application environments and potential new problems during system debugging, these values ​​in the multimodal physical condition early warning database are not fixed. Technicians can manually set, adjust, or fine-tune them at any time based on the system's performance in actual testing, ensuring continuous system optimization to achieve optimal operating conditions.

[0031] like Figure 2 The image shown is one of the homepage interfaces of an intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors provided in an embodiment of the present invention; Figure 2 As can be seen, the navigation bar of the health monitoring system provided in this embodiment of the invention includes a homepage, real-time health monitoring, data aggregation and analysis, early warning and reminders, data management, personalized health management, user settings, and help and support. Data aggregation and analysis includes historical health data and health assessment reports; early warning and reminders include health early warnings, early warning management, and early warning history records; data management includes multimodal data management and data uploading and storage. Specifically, the homepage interface of the health monitoring system includes a core health indicator section, an early warning prompt section, and a health advice section. The core health indicator section displays data such as heart rate, blood pressure, blood oxygen, and body temperature. By monitoring key health indicators such as heart rate, blood pressure, blood oxygen, and body temperature in real time, the target user can understand their own health status in a timely manner and receive timely early warnings. The early warning prompt section displays early warning results and recent early warning records, and the health advice section provides health advice, such as suggesting reducing salt intake to no more than 5 grams per day. By providing personalized health advice, it further helps users better manage and improve their health.

[0032] The first module of the Parkinson's / Epilepsy Recognition Early Warning Intelligent System based on multimodal sensors is the multimodal acquisition interference judgment module. This module is used to collect multimodal biological data of the target object based on the multimodal biosensors in the wearable device. The target group is users of wearable devices for health monitoring and early warning, such as patients with epilepsy or Parkinson's disease. By collecting multimodal biological data through multimodal biosensors (such as a 6-axis accelerometer, optical heart rate sensor, and skin conductivity sensor), the physiological state of the target object can be monitored from multiple angles and in all directions, thereby helping to achieve comprehensive monitoring and early warning of the target object's health condition. Epileptic seizures are usually accompanied by muscle twitching, loss of consciousness, and abnormal electrical activity. Using a 6-axis accelerometer, the patient's movement patterns can be captured in real time to monitor for abnormal body movements (such as twitching). At the same time, the optical heart rate sensor can monitor changes in heart rate, and the skin conductivity sensor can measure changes in skin conductivity. These changes in physiological signals are sometimes precursors or reactions to epileptic seizures. Common movement disorders in Parkinson's disease patients include tremor, muscle rigidity, and bradykinesia. A 6-axis accelerometer can help monitor these motion abnormalities, an optical heart rate sensor can detect arrhythmia or other cardiovascular problems, and a skin conductance sensor can reflect skin responses related to emotions or stress. Both involve abnormalities of the brain and nervous system, and their conditions are often difficult to predict and monitor accurately. Therefore, real-time monitoring with external devices is necessary. By using multimodal sensors (such as accelerometers, optical heart rate sensors, and skin conductance sensors), the physiological signals of the target patient can be continuously and in real-time monitored, thereby better monitoring fluctuations and changes in physiological signals and enabling timely intervention. Next, multimodal interference warning results are obtained to quantify the degree of electromyographic interference during the multimodal biological data acquisition process. The specific steps for obtaining multimodal interference warning results are as follows:

[0033] A1. Obtain electromyographic interference (EMG) judgment indicators during the multimodal biological data collection process within the early warning collection period. The EMG judgment indicators include EMG time-domain indicators and EMG frequency-domain indicators.

[0034] Specifically, the time-domain index of electromyography (EMG) interference represents the result of coupling the integrated processing result of EMG signal intensity and the relative deviation processing result of EMG signal peak value with EMG interference time-domain influencing factors obtained from a preset database after compensation. The EMG signal intensity is acquired through a bioelectric sensor configured in a wearable device, and the integrated processing result of the EMG signal intensity is the specific constraint expression of the time-domain index of EMG interference. Part of the specific constraint expression for the relative deviation processing result of the electromyographic signal peak, i.e., the electromyographic interference time-domain index. The time-domain influence factors of electromyography (EMG) interference include the cumulative interference influence factor and the peak interference influence factor, which describe the degree of influence of the integrated processing result of EMG signal intensity and the relative deviation processing result of EMG signal peak on the time-domain index of EMG interference, respectively. The specific constrained expression for the time-domain index of EMG interference is as follows:

[0035]

[0036] In the formula, S represents the time-domain index of electromyography interference, S1 represents the cumulative interference factor, S2 represents the peak interference factor, t represents the preset time within the warning collection period, t∈[t0,t1], t0 represents the initial time within the warning collection period, t1 represents the end time within the warning collection period, X0 represents the reference value of electromyography signal intensity, x(t) represents the electromyography signal intensity at time t within the warning collection period, F represents the peak value of electromyography signal intensity, and F0 represents the reference peak value of electromyography signal intensity.

[0037] In addition, the electromyographic interference frequency domain index represents the comparison result of the deviation between the median frequency of the electromyographic signal and the median frequency reference value. That is, the result of taking the absolute value of the difference between the median frequency of the electromyographic signal and the median frequency reference value and then proportionally processing it with the median frequency reference value. The median frequency of the electromyographic signal is obtained by running a simplified Fourier algorithm through the MCU (Microcontroller Unit) built into the wearable device.

[0038] A2 determines whether the electromyography interference judgment index is 0. If it is, the multimodal acquisition interference value is marked as an outlier. When the multimodal acquisition interference value is an outlier, the target object is prompted to re-acquire multimodal biological data. Otherwise, the electromyography interference judgment index is harmonic averaged to obtain the multimodal acquisition interference value. The obtained multimodal acquisition interference value is compared with the electromyography interference limit index obtained from the preset database to obtain the multimodal interference warning result.

[0039] Specifically, the multimodal interference warning results include interference-limited normal and interference-limited warning. Interference-limited normal means that the multimodal acquisition interference value is not greater than the multimodal interference warning result corresponding to the electromyographic interference limit index. Interference-limited warning means that the multimodal acquisition interference value is greater than the multimodal interference warning result corresponding to the electromyographic interference limit index.

[0040] It is important to understand that as the indicators for judging EMG interference increase, the multimodal acquisition interference value also increases. The multimodal acquisition interference value quantifies the degree of EMG interference during the multimodal biological data acquisition process using both time-domain and frequency-domain EMG interference indicators. Furthermore, the multimodal acquisition interference value considers the correlation and mutual influence between various parameters. Specifically, the integrated processing result of the EMG signal intensity describes the overall intensity of the EMG signal over a time period, reflecting the continuous activity of the muscle, while the relative deviation processing result of the EMG signal peak value describes the instantaneous fluctuation of the EMG signal. If the intensity of the EMG signal increases within a certain period, it is usually accompanied by an increase in the peak value of the EMG signal. Furthermore, the two are interconnected: as the time-domain index of EMG interference increases, the multimodal acquisition interference value also increases; simultaneously, the larger the frequency-domain index of EMG interference, the more significant the change in the signal spectrum, which may indicate that the EMG signal has been subjected to strong interference, and the multimodal acquisition interference value also increases accordingly. The time-domain and frequency-domain characteristics of EMG signals are closely related. The time-domain index of EMG interference mainly reflects the persistence and fluctuation of the EMG signal intensity, while the frequency-domain index of EMG interference reflects the shift in the frequency characteristics of the EMG signal. As the time-domain index of EMG interference increases, the frequency-domain index of EMG interference also increases, indicating that the degree of EMG interference in the multimodal biological data acquisition process increases, that is, the multimodal acquisition interference value increases accordingly.

[0041] This invention obtains multimodal acquisition interference values ​​by using time-domain and frequency-domain indices of electromyography (EMG) interference, which helps to more accurately determine the degree of EMG interference during the multimodal biological data acquisition process. This improves the reliability of multimodal biological data acquisition and early warning during the health status early warning process using wearable devices.

[0042] The second module of the Parkinson's / Epilepsy Recognition Early Warning Intelligent System based on Multimodal Sensors is a multimodal interference processing optimization module. This module is used to determine whether to perform multimodal interference processing optimization based on the obtained multimodal interference early warning results. Multimodal interference processing optimization means combining the multimodal interference early warning results to perform filtering and acquisition adjustment on the multimodal biological data acquisition process. The specific steps are as follows:

[0043] B1, if the multimodal interference warning result corresponds to normal interference limitation, then no multimodal interference processing optimization is performed, a multimodal fusion command is sent, and the multimodal fusion warning result is obtained; if Figure 3The diagram shows a flowchart of the multimodal fusion early warning result acquisition process of the Parkinson's / epilepsy recognition early warning intelligent system based on multimodal sensors provided in this embodiment of the invention. The corresponding logic is as follows: Quantitatively evaluate the fusion effect of multimodal biological data based on the acquired fusion judgment quantification index to obtain a multimodal fusion quality score. Determine whether the multimodal fusion quality score is greater than the preset fusion effect quality index. If so, record the multimodal fusion early warning effect as qualified and perform multimodal fusion input early warning. Otherwise, record the multimodal fusion early warning effect as unqualified and perform multimodal fusion processing optimization.

[0044] Specifically, the multimodal fusion early warning results include fusion effect qualified and fusion effect unqualified; reference Figure 3 The specific steps to obtain multimodal fusion early warning results are as follows:

[0045] B11. Obtain quantitative indicators for fusion judgment during the multimodal biological data fusion process. These quantitative indicators include multimodal fusion information gain, multimodal fusion data loss, multimodal fusion processing time, and multimodal fusion memory consumption. Specifically, the multimodal fusion information gain and multimodal fusion data loss are obtained using the SciPy library in Python, the multimodal fusion processing time is obtained using the timeit library in Python, and the multimodal fusion memory consumption is obtained using the psuti l library in Python.

[0046] B12, based on the fusion judgment quantitative index, quantitatively evaluate the fusion effect of multimodal biological data to obtain a multimodal fusion quality score; the specific process for obtaining the multimodal fusion quality score is as follows:

[0047] B121, the fusion efficiency parameter is obtained by coupling the processing results of the relative proportion of fusion information gain deviation and the processing results of the deviation proportion of multimodal fusion lost data with the fusion efficiency compensation amount obtained from the preset database after compensation calculation; the specific constraint expression of the fusion efficiency parameter is as follows:

[0048]

[0049] In the formula, L represents the fusion efficiency parameter, L1 represents the fusion gain compensation, L2 represents the fusion loss compensation, Y0 represents the information gain reference limit, and Y... c Y represents the information gain reference deviation, D represents the multimodal fusion information gain, D represents the amount of data lost during multimodal fusion, and D0 represents the fusion loss reference limit.

[0050] The relative proportion of fusion information gain bias represents the result of proportionalizing the difference between the multimodal fusion information gain and the information gain reference limit with respect to the information gain reference bias, i.e., the ratio in the specific constraint expression of the fusion efficiency parameter. Partial; The result of the multimodal fusion data loss deviation ratio processing represents the result of processing the difference between the fusion loss reference limit and the multimodal fusion data loss ratio, and the fusion loss reference limit as a percentage, i.e., the ratio in the specific constraint expression of the fusion efficiency parameter. The fusion efficiency compensation includes fusion gain compensation and fusion loss compensation, which are used to describe the degree of influence of the fusion information gain deviation relative proportion processing result and the multimodal fusion lost data amount deviation ratio processing result on the fusion efficiency parameter.

[0051] B122, after performing compensation calculations on the multimodal fusion processing time deviation processing results and the multimodal fusion memory consumption deviation processing results, and coupling them with the fusion accuracy compensation amount obtained from the preset database, the fusion accuracy parameters are obtained. The multimodal fusion processing time deviation processing result represents the result of taking the absolute value of the difference between the fusion processing time limit and the multimodal fusion processing time and then proportionally processing it to the fusion processing time limit. The multimodal fusion memory consumption deviation processing result represents the result of taking the absolute value of the difference between the fusion memory consumption limit and the multimodal fusion memory consumption and then proportionally processing it to the fusion memory consumption limit. The fusion accuracy compensation amount includes the fusion processing time compensation amount and the fusion memory consumption compensation amount, which are used to describe the degree of influence of the multimodal fusion processing time deviation processing results and the multimodal fusion memory consumption deviation processing results on the fusion accuracy parameters, respectively.

[0052] B123, after weighting and coupling the obtained fusion efficiency parameters, fusion accuracy parameters and fusion quality compensation factors obtained from the preset database, a multimodal fusion quality score is obtained. The multimodal fusion quality score represents the quantitative data of the fusion judgment quantitative index on the fusion effect of multimodal biological data. The fusion quality compensation factor is used to describe the degree of influence of the fusion efficiency parameters and fusion accuracy parameters on the multimodal fusion quality score, including the fusion efficiency compensation factor and the fusion accuracy compensation factor.

[0053] The increase in fusion efficiency and fusion accuracy parameters indicates enhanced fusion performance of multimodal biological data, leading to a higher multimodal fusion quality score. The correlation and mutual influence between these parameters were considered quantitatively. For example, an increase in multimodal fusion information gain means the fused data contains more effective information and provides higher-quality biological signals; thus, an increase in fusion efficiency parameters results in a higher multimodal fusion quality score. Furthermore, fusion efficiency and fusion accuracy parameters are also interconnected and mutually influential. An increase in multimodal fusion processing time may indicate that the system fails to process data in a timely manner, leading to increased data loss or even an inability to provide high-quality information gain. In other words, a decrease in fusion efficiency parameters leads to a decrease in fusion accuracy parameters, consequently lowering the multimodal fusion quality score.

[0054] B13. Compare the multimodal fusion quality score with the preset fusion effect quality index obtained from the preset database, and determine whether the multimodal fusion quality score is greater than the preset fusion effect quality index. If so, record the multimodal fusion warning effect as qualified; otherwise, record the multimodal fusion warning effect as unqualified.

[0055] B2. If the multimodal interference warning result corresponds to an interference limitation warning, then determine whether to perform time-domain-frequency domain interference filtering optimization based on the multimodal acquired interference value. If so, send the interference warning optimization command after performing time-domain-frequency domain interference filtering optimization; otherwise, directly send the interference warning optimization command used to perform multimodal biological data warning optimization.

[0056] The specific steps for determining whether to perform time-domain-frequency domain interference filtering optimization based on multimodal acquisition interference values ​​are as follows:

[0057] B21. Determine whether the deviation of the multimodal acquisition interference value is within the preset interference deviation limit range obtained from the preset database. If not, do not perform time-domain-frequency domain interference filtering optimization and perform multimodal biological data early warning optimization. If yes, obtain the time-domain-frequency domain filtering adjustment amount to perform time-domain-frequency domain interference filtering optimization. The deviation of the multimodal acquisition interference value represents the result of the ratio processing of the difference between the multimodal acquisition interference value and the electromyographic interference limit index and the electromyographic interference limit index.

[0058] It should be noted that the time-domain to frequency-domain filtering adjustment includes the moving average window adjustment length obtained by arithmetically averaging the time-domain index mapping results of EMG interference and the deviation of multimodal acquisition interference values ​​from the time-domain mapping results, and the filter cutoff frequency adjustment obtained by arithmetically averaging the frequency-domain index mapping results of EMG interference and the deviation of multimodal acquisition interference values ​​from the frequency-domain mapping results.

[0059] Optimization of time-domain and frequency-domain interference filtering involves increasing the initial moving average window length by adjusting the moving average window size, while simultaneously decreasing the initial filter cutoff frequency by adjusting the filter cutoff frequency. Biological signals often exhibit high time-varying characteristics or occasional short-term fluctuations. Increasing the initial moving average window length effectively reduces the impact of these local fluctuations on the multimodal biological data acquisition process, thereby improving the reliability of health status early warnings. Lowering the filter cutoff frequency can more effectively remove high-frequency noise or interference from the signal, such as high-frequency components in electromyography (EMG) signals. The combination of increasing the moving average window length and lowering the filter cutoff frequency helps smooth the signal and reduce interference caused by high-frequency noise. The moving average provides temporal smoothing, while the low-pass filter removes unwanted high-frequency components in the frequency domain; their combined use achieves better signal denoising results.

[0060] The time-domain index mapping result of electromyography interference represents the result of mapping the time-domain index of electromyography interference from the electromyography time-domain interference mapping set constructed in the preset database. The frequency-domain index mapping result of electromyography interference represents the result of mapping the frequency-domain index of electromyography interference from the frequency-domain interference mapping set constructed in the preset database.

[0061] The multimodal acquisition interference value deviation time domain mapping result represents the result of mapping the degree of deviation of the multimodal acquisition interference value from the multimodal time domain interference mapping set constructed in the preset database. The multimodal acquisition interference value deviation frequency domain mapping result represents the result of mapping the degree of deviation of the multimodal acquisition interference value from the multimodal frequency domain interference mapping set constructed in the preset database.

[0062] In addition, such as Figure 4 The diagram shown is a flowchart of the multimodal biological data early warning optimization process of the Parkinson's / epilepsy recognition intelligent system based on multimodal sensors provided in this embodiment of the invention. The corresponding logic is as follows: First, increase the initial acquisition resolution based on the sampling resolution adjustment amount, and decrease the initial early warning period based on the early warning period adjustment amount. Then, determine whether the deviation of the multimodal acquisition interference value is greater than the midpoint of the deviation limit interval. If so, decrease the signal amplification gain according to the signal amplification gain up adjustment value; otherwise, increase the signal amplification gain according to the signal amplification gain down adjustment value. (Refer to...) Figure 4 The specific steps for implementing multimodal biological data early warning optimization are as follows:

[0063] First, the deviation of multimodal acquisition interference values ​​is input into the multimodal early warning-sampling resolution adjustment linear regression model, and the corresponding output is the sampling resolution adjustment amount used to improve the initial acquisition resolution in the process of acquiring multimodal biological data, that is, the initial acquisition resolution is increased according to the sampling resolution adjustment amount. Next, the deviation of multimodal acquisition interference values ​​is input into the multimodal early warning-early warning cycle adjustment linear regression model, and the corresponding output is the early warning cycle adjustment amount used to reduce the initial early warning cycle for obtaining multimodal interference early warning results, that is, the initial early warning cycle is reduced according to the early warning cycle adjustment amount.

[0064] The multimodal early warning-sampling resolution adjustment linear regression model is a pre-trained linear regression model used to fit the mapping relationship between the deviation of multimodal acquisition interference values ​​and the sampling resolution adjustment amount. The multimodal early warning-early warning cycle adjustment linear regression model is also a pre-trained linear regression model used to fit the mapping relationship between the deviation of multimodal acquisition interference values ​​and the early warning cycle adjustment amount. The model is obtained by inputting the deviation of multimodal acquisition interference values ​​over a historical period and the preset sampling resolution adjustment amount set by staff based on the deviation of multimodal acquisition interference values ​​into the linear regression model, and then training it using the least squares method through the scikit-learn framework. Similarly, the model is obtained by inputting the deviation of multimodal acquisition interference values ​​over a historical period and the preset early warning cycle adjustment amount set by staff based on the deviation of multimodal acquisition interference values ​​into the linear regression model, and then training it using the least squares method through the scikit-learn framework.

[0065] Secondly, it is determined whether the deviation of the multimodal acquisition interference value is greater than the midpoint of the deviation range. If so, the signal amplification gain is reduced by mapping the signal amplification gain upward from the pre-built signal amplification gain upward mapping set in the preset database based on the multimodal acquisition interference value deviation upward adjustment result. In other words, the signal amplification gain is reduced based on the signal amplification gain upward adjustment value. Otherwise, the signal amplification gain is increased by mapping the signal amplification gain downward from the pre-built signal amplification gain downward from the pre-built signal amplification gain downward from the multimodal acquisition interference value downward adjustment result. When the deviation of the multimodal acquisition interference value is large, it indicates that the acquired signal is affected by strong interference or noise. In this case, reducing the signal amplification gain can reduce the amplification of interference signals, avoiding excessive amplification of noise signals, which could lead to false alarms or inaccurate judgments. By reducing the signal amplification gain, the true state of the multimodal biological data acquisition process can be more accurately reflected, avoiding the impact of data distortion on health condition warnings. Furthermore, by mapping the signal amplification gain down-adjustment value obtained from the signal amplification gain down-adjustment mapping set constructed in the preset database based on the multimodal acquisition interference value deviation down-adjustment result, the signal amplification gain can be improved, which can enhance the amplification of effective biological signals and more accurately capture changes in the physical condition of the target object, especially in the small fluctuations of physiological parameters such as heart rate and respiration, thereby enhancing the early warning capability of physical condition.

[0066] The midpoint of the deviation limit interval is the average of the upper limit and lower limit of the deviation limit interval of the preset interference deviation limit interval. The preset interference deviation limit interval is the closed interval formed by the lower limit and upper limit of the deviation limit interval.

[0067] The result of the multimodal acquisition interference value deviation adjustment is the result of taking the absolute value of the difference between the degree of multimodal acquisition interference value deviation and the upper limit of the deviation limit interval, and then proportionally processing it with the upper limit of the deviation limit interval.

[0068] The result of the multimodal acquisition interference value deviation reduction is the result of taking the absolute value of the difference between the degree of multimodal acquisition interference value deviation and the lower limit of the deviation limit interval, and then proportionally processing it with the lower limit of the deviation limit interval.

[0069] This invention determines whether to perform multimodal interference processing optimization based on the obtained multimodal interference warning results, thereby achieving a more accurate judgment of the multimodal biological data fusion effect. When the multimodal fusion warning result corresponds to an unqualified fusion effect, multimodal fusion input warning is not performed, and multimodal fusion processing optimization is executed instead. This helps to improve the reliability of multimodal biological data fusion, thereby achieving more accurate warning of physical condition.

[0070] The third module of the intelligent early warning system for Parkinson's / Epilepsy recognition based on multimodal sensors is a multimodal fusion input early warning module. This module determines whether to issue a multimodal fusion input early warning based on the multimodal fusion early warning result. If so, the multimodal fusion input early warning is executed; otherwise, multimodal fusion processing optimization is performed. Multimodal fusion processing optimization means adjusting the dimensionality reduction process during multimodal biological data fusion based on the multimodal fusion early warning result. Specifically, the steps for determining whether to issue a multimodal fusion input early warning based on the multimodal fusion early warning result are as follows:

[0071] First, if the multimodal fusion warning result is that the fusion effect is qualified, then multimodal fusion input warning is carried out, while multimodal biological data of the target object is continuously collected and multimodal interference warning results are obtained.

[0072] Secondly, if the multimodal fusion warning result indicates that the fusion effect is unqualified, then the multimodal fusion input warning will not be issued, and multimodal fusion processing optimization will be performed. Multimodal fusion processing optimization includes batch processing data volume fusion optimization and batch update principal component processing optimization. Among them, batch processing data volume fusion optimization means adjusting the batch processing data volume for dimensionality reduction feature extraction during multimodal biological data fusion based on the multimodal fusion warning result; batch update principal component processing optimization means adjusting the batch update principal component period for dimensionality reduction feature extraction during multimodal biological data fusion based on the multimodal fusion warning result.

[0073] Specifically, the process for performing multimodal fusion processing optimization is as follows:

[0074] S1. The multimodal fusion quality score is input into the pre-constructed batch processing data volume mapping set and batch update principal component mapping set in the preset database to obtain batch processing data volume adjustment value and batch update principal component period adjustment value. The initial batch processing data volume is increased based on the obtained batch processing data volume adjustment value, and the initial batch update principal component period is decreased based on the batch update principal component period adjustment value. Specifically, if the batch processing data volume is too small each time, the updated principal components will be affected by small data fluctuations, leading to instability in the updated principal components. As the amount of multimodal biological data increases, the estimation of principal components may be inaccurate, thus affecting the dimensionality reduction effect and the quality of subsequent multimodal biological data fusion. Furthermore, if the batch update principal component period is too long, i.e., the principal components are not updated frequently, and the amount of data change in each update is small, the learning process of the principal components is too slow and may not be able to reflect the changes in multimodal biological data in a timely manner, thus affecting the fusion effect of multimodal biological data.

[0075] S2, re-execute multimodal biological data fusion, determine whether to issue a multimodal fusion input warning. If so, indicate that the multimodal fusion processing optimization has ended; otherwise, return to S1 until a multimodal fusion input warning is issued.

[0076] like Figure 5 The image shown is one of the multimodal data management interface diagrams of an intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors provided in an embodiment of the present invention; by Figure 5 As can be seen, the multimodal data management interface in the health monitoring system provided in this embodiment of the invention includes a multimodal data center, which is used to display data types, collection times, source devices, etc. The multimodal data center can centrally display and manage health data from different devices and of different types. This centralized data presentation method helps to quickly understand the data source, collection time, and data type, ensuring the comprehensiveness and reliability of the data and effectively supporting the overall analysis of the health monitoring system. In addition, the multimodal data management interface can also be used to display multimodal interference warning results, multimodal fusion warning results, and data cleaning and maintenance. When the multimodal fusion warning result is that the fusion effect is unqualified, it can also display the batch processing data volume and the adjustment status of the batch update principal component cycle, which helps to further understand the adjustment status of multimodal biological data processing, thereby improving the reliability of health status warning.

[0077] In the process of acquiring multimodal biological data through multimodal biosensors built into wearable devices to provide early warnings of the target object's physical condition, the electromyographic interference present during the data acquisition process often leads to inaccurate multimodal biological data, which in turn affects the accuracy of the multimodal biological data fusion process, resulting in inaccurate early warnings. This invention analyzes the correlation between the acquisition and fusion processes of multimodal biological data in the early warning process, enabling a more accurate assessment of the degree of electromyographic interference and the fusion effect during the acquisition process. This, in turn, improves the reliability of multimodal biological data acquisition, fusion, and early warning in the early warning process using wearable devices.

[0078] This application provides a smart bracelet, comprising: a bracelet body, a central processing unit (CPU), and a multimodal biosensor. The multimodal biosensor is disposed in the bracelet body and includes a 6-axis accelerometer, an optical heart rate sensor, and a skin conductance sensor. The CPU is located within the bracelet body. The CPU is used to acquire multimodal interference warning results based on multimodal biological data collected by the multimodal biosensor, determine whether to perform multimodal interference processing optimization and send a multimodal fusion command, and determine whether to perform multimodal fusion input warning based on the acquired multimodal fusion warning results to output warning information. The bracelet body includes a band and a ring disc connected together. Furthermore, the bracelet body also includes... The system comprises a communication component, a display screen, and a power supply. Specifically, the communication component is primarily responsible for data transmission between the bracelet and other devices (such as smartphones and cloud servers). The display screen serves as the main interface for interacting with the target audience, displaying alert information. The power supply includes a connected voltage power module and a voltage conversion module. The voltage conversion module converts the voltage provided by the voltage power supply into the operating voltage of the central processing unit. Additionally, the bracelet itself includes a charging module connected to the power supply. This charging module includes a charging current control module and a charging voltage control module, both connected to the power supply. The charging current control module is used to charge the voltage power module.

[0079] In summary, during the health condition early warning process using wearable devices, the electromyographic interference (EMG) during multimodal biological data collection often leads to inaccurate data, further affecting the accuracy of the multimodal biological data fusion process and consequently causing inaccurate health condition early warnings. This invention addresses this issue by using multimodal biological data of the target object collected by multimodal biosensors in wearable devices to obtain multimodal interference early warning results. This helps to more accurately determine the degree of EMG interference during the multimodal biological data collection process, and then determines whether to perform multimodal interference processing optimization. This effectively reduces the impact of EMG interference on multimodal biological data collection, improving the reliability of multimodal biological data collection. Finally, based on the multimodal fusion early warning results, it determines whether to perform multimodal fusion input early warning, achieving a more accurate assessment of the multimodal biological data fusion effect. This also enables the correlation analysis between the multimodal biological data collection and fusion processes during health condition early warning, effectively solving the problem of insufficient correlation between multimodal biological data collection and fusion in existing technologies for health condition early warning using wearable devices.

[0080] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0081] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0082] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0083] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0084] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0085] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0086] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors, characterized in that, include: Multimodal acquisition interference judgment module, multimodal interference processing optimization module, and multimodal fusion input early warning module; The multimodal acquisition interference judgment module is used to obtain multimodal interference warning results based on the multimodal biological data of the target object collected by the multimodal biosensor in the wearable device, so as to quantitatively judge the degree of electromyographic interference in the multimodal biological data acquisition process. The multimodal interference processing optimization module is used to determine whether to perform multimodal interference processing optimization based on the obtained multimodal interference warning results. If it is performed, a multimodal fusion command for performing multimodal biological data fusion is sent after the multimodal interference processing optimization. Otherwise, multimodal biological data fusion is performed directly, and multimodal fusion warning results are obtained to quantitatively judge the fusion effect of multimodal biological data. The multimodal interference processing optimization means combining the multimodal interference warning results to perform filtering and acquisition adjustment on the multimodal biological data acquisition process. The multimodal fusion input early warning module is used to determine whether to perform multimodal fusion input early warning based on the multimodal fusion early warning result. If so, multimodal fusion input early warning is performed; otherwise, multimodal fusion processing optimization is performed. The multimodal fusion processing optimization means adjusting the dimensionality reduction processing during the multimodal biological data fusion process in combination with the multimodal fusion early warning result.

2. The intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors according to claim 1, characterized in that, The specific steps for obtaining the multimodal interference early warning result are as follows: Acquire electromyographic interference (EMG) judgment indicators during multimodal biological data acquisition, wherein the EMG judgment indicators include EMG time-domain indicators and EMG frequency-domain indicators. The electromyographic interference time-domain index represents the result of coupling the integrated processing result of electromyographic signal intensity and the relative deviation processing result of electromyographic signal peak value with the electromyographic interference time-domain influence factor obtained from the preset database after compensation. The electromyographic interference frequency domain index represents the comparison result of the deviation between the median frequency of the electromyographic signal and the median frequency reference value. If the electromyography interference judgment index is 0, the multimodal acquisition interference value is marked as an outlier. Otherwise, the electromyography interference judgment index is processed by harmonic averaging to obtain the multimodal acquisition interference value. The obtained multimodal acquisition interference value is compared with the electromyography interference limit index obtained from the preset database to obtain the multimodal interference warning result. The multimodal interference warning results include interference-limited normal and interference-limited warning. Interference-limited normal indicates that the multimodal acquisition interference value is not greater than the multimodal interference warning result corresponding to the electromyographic interference limit index. Interference-limited warning indicates that the multimodal acquisition interference value is greater than the multimodal interference warning result corresponding to the electromyographic interference limit index.

3. The intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors according to claim 2, characterized in that, The specific steps for determining whether to perform multimodal interference processing optimization based on the obtained multimodal interference early warning results are as follows: If the multimodal interference warning result corresponds to normal interference limitation, then no multimodal interference processing optimization is performed, a multimodal fusion command is sent and the multimodal fusion warning result is obtained; If the multimodal interference warning result corresponds to an interference limitation warning, then determine whether to perform time-domain-frequency domain interference filtering optimization based on the multimodal acquisition interference value. If so, then send an interference warning optimization command after performing time-domain-frequency domain interference filtering optimization; otherwise, directly send an interference warning optimization command for performing multimodal biological data warning optimization.

4. The intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors according to claim 3, characterized in that, The specific steps for determining whether to perform time-domain-frequency domain interference filtering optimization based on multimodal acquisition interference values ​​are as follows: Determine whether the deviation of the multimodal acquisition interference value is within the preset interference deviation limit range obtained from the preset database. If not, do not perform time-domain-frequency domain interference filtering optimization and perform multimodal biological data early warning optimization. If yes, obtain the time-domain-frequency domain filtering adjustment amount to perform time-domain-frequency domain interference filtering optimization. The deviation of the multimodal acquisition interference value represents the result of proportional processing of the difference between the multimodal acquisition interference value and the electromyographic interference limit index and the electromyographic interference limit index. The time-domain-frequency-domain filtering adjustment includes the moving average window adjustment length obtained by arithmetically averaging the time-domain index mapping results of electromyography interference and the deviation of multimodal acquisition interference values ​​from the time-domain mapping results, and the filter cutoff frequency adjustment obtained by arithmetically averaging the frequency-domain index mapping results of electromyography interference and the deviation of multimodal acquisition interference values ​​from the frequency-domain mapping results. The time-domain-frequency domain interference filtering optimization means increasing the initial moving average window length according to the moving average window adjustment length, and simultaneously decreasing the initial filter cutoff frequency according to the filter cutoff frequency adjustment amount. The electromyographic interference time-domain index mapping result represents the result of mapping the electromyographic interference time-domain index from the electromyographic time-domain interference mapping set constructed in the preset database, and the electromyographic interference frequency-domain index mapping result represents the result of mapping the electromyographic interference frequency-domain index from the electromyographic frequency-domain interference mapping set constructed in the preset database. The multimodal acquisition interference value deviation time domain mapping result represents the result of mapping the degree of deviation of the multimodal acquisition interference value from the multimodal time domain interference mapping set constructed in the preset database, and the multimodal acquisition interference value deviation frequency domain mapping result represents the result of mapping the degree of deviation of the multimodal acquisition interference value from the multimodal frequency domain interference mapping set constructed in the preset database.

5. The intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors according to claim 4, characterized in that, The specific steps for performing multimodal biological data early warning optimization are as follows: The deviation of the multimodal acquisition interference value is input into the multimodal early warning-sampling resolution adjustment linear regression model, and the corresponding output is the sampling resolution adjustment amount used to improve the initial acquisition resolution in the process of acquiring multimodal biological data; The deviation of the multimodal acquisition interference value is input into the multimodal early warning-early warning cycle adjustment linear regression model, and the corresponding output is the early warning cycle adjustment amount used to reduce the initial early warning cycle for obtaining multimodal interference early warning results; If the deviation of the multimodal acquisition interference value is greater than the midpoint of the deviation limit interval, the signal amplification gain is reduced by mapping the signal amplification gain upward adjustment value obtained from the signal amplification gain upward adjustment mapping set constructed in the preset database according to the multimodal acquisition interference value deviation upward adjustment result. Otherwise, the signal amplification gain is increased by mapping the signal amplification gain downward adjustment value obtained from the signal amplification gain downward adjustment mapping set constructed in the preset database according to the multimodal acquisition interference value downward adjustment result. The midpoint of the deviation limit interval is the average of the upper limit value and the lower limit value of the deviation limit interval of the preset interference deviation limit interval. The preset interference deviation limit interval is a closed interval formed by the lower limit value and the upper limit value of the deviation limit interval. The multimodal acquisition interference value deviation adjustment result represents the result of taking the absolute value of the difference between the multimodal acquisition interference value deviation degree and the upper limit value of the deviation limit interval, and then proportionally processing it with the upper limit value of the deviation limit interval. The multimodal acquisition interference value deviation reduction result represents the result of taking the absolute value of the difference between the multimodal acquisition interference value deviation degree and the lower limit of the deviation limit interval, and then proportionally processing it with the lower limit of the deviation limit interval.

6. The intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors according to claim 3, characterized in that, The specific steps for obtaining the multimodal fusion early warning result are as follows: The quantitative indicators for fusion judgment during the multimodal biological data fusion process are obtained, including multimodal fusion information gain, multimodal fusion data loss, multimodal fusion processing time, and multimodal fusion memory consumption. The fusion effect of multimodal biological data is quantitatively evaluated based on fusion judgment quantitative indicators to obtain a multimodal fusion quality score; The multimodal fusion quality score is compared with the preset fusion effect quality index obtained from the preset database. It is determined whether the multimodal fusion quality score is greater than the preset fusion effect quality index. If so, the multimodal fusion warning effect is recorded as the fusion effect is qualified; otherwise, the multimodal fusion warning effect is recorded as the fusion effect is unqualified. The multimodal fusion early warning results include fusion effect qualified and fusion effect unqualified.

7. The intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors according to claim 6, characterized in that, The process for obtaining the multimodal fusion quality score is as follows: The fusion efficiency parameter is obtained by coupling the processing results of the relative proportion of fusion information gain deviation and the processing results of the deviation of multimodal fusion lost data with the fusion efficiency compensation amount obtained from the preset database after performing compensation calculation. The fusion information gain deviation relative proportion processing result represents the result of processing the difference between the multimodal fusion information gain and the information gain reference limit with the information gain reference deviation as a proportion; the multimodal fusion lost data amount deviation proportion processing result represents the result of processing the difference between the fusion lost reference limit and the multimodal fusion lost data amount with the fusion lost reference limit as a proportion. The multimodal fusion processing time deviation processing result and the multimodal fusion memory consumption deviation processing result are coupled with the fusion accuracy compensation amount obtained from the preset database after compensation calculation to obtain the fusion accuracy parameters. The multimodal fusion processing time deviation processing result represents the result of taking the absolute value of the difference between the fusion processing time limit and the multimodal fusion processing time and then proportionally processing it to the fusion processing time limit. The multimodal fusion memory consumption deviation processing result represents the result of taking the absolute value of the difference between the fusion memory consumption limit and the multimodal fusion memory consumption and then proportionally processing it to the fusion memory consumption limit. The multimodal fusion quality score is obtained by performing weighted operations and coupling processing on the obtained fusion efficiency parameters, fusion accuracy parameters and fusion quality compensation factors obtained from the preset database. The multimodal fusion quality score represents the quantitative data of the fusion judgment quantification index on the fusion effect of multimodal biological data. The fusion quality compensation factor is used to describe the degree of influence of fusion efficiency parameter and fusion accuracy parameter on the multimodal fusion quality score, including fusion efficiency compensation factor and fusion accuracy compensation factor.

8. The intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors according to claim 6, characterized in that, The specific steps for determining whether to issue a multimodal fusion input warning based on the multimodal fusion warning result are as follows: If the multimodal fusion warning result is that the fusion effect is qualified, then multimodal fusion input warning is issued, and multimodal biological data of the target object is continuously collected and multimodal interference warning results are obtained. If the multimodal fusion warning result is that the fusion effect is unqualified, then the multimodal fusion input warning will not be issued, and multimodal fusion processing optimization will be performed. The multimodal fusion processing optimization includes batch processing data volume fusion optimization and batch updating principal component processing optimization. The batch processing data volume fusion optimization means adjusting the batch processing data volume of dimensionality reduction feature extraction during the multimodal biological data fusion process by combining the multimodal fusion early warning results; The batch update principal component processing optimization means adjusting the batch update principal component cycle of dimensionality reduction feature extraction during multimodal biological data fusion by combining the multimodal fusion early warning results.

9. The intelligent early warning system for Parkinson's / epilepsy recognition based on multimodal sensors according to claim 8, characterized in that, The specific process for performing multimodal fusion processing optimization is as follows: S1, input the multimodal fusion quality score into the batch processing data volume mapping set and the batch update principal component mapping set that have been constructed in the preset database respectively, and map them to obtain the batch processing data volume adjustment value and the batch update principal component period adjustment value. Increase the initial batch processing data volume according to the obtained batch processing data volume adjustment value, and decrease the initial batch update principal component period according to the batch update principal component period adjustment value. S2, re-execute multimodal biological data fusion, determine whether to issue a multimodal fusion input warning. If so, indicate that the multimodal fusion processing optimization has ended; otherwise, return to S1 until a multimodal fusion input warning is issued.

10. A smart bracelet, employing the early warning smart system for Parkinson's / epilepsy recognition based on multimodal sensors as described in any one of claims 1-9, comprising: The bracelet itself, central processing unit, and multimodal biosensors; The multimodal biosensor is configured in the main body of the bracelet, and the multimodal biosensor includes a 6-axis accelerometer, an optical heart rate sensor, and a skin conductance sensor. The central processing unit is located in the main body of the bracelet. The central processing unit is used to obtain multimodal interference early warning results based on multimodal biological data collected by multimodal biosensors, determine whether to perform multimodal interference processing optimization and send multimodal fusion instructions, and determine whether to perform multimodal fusion input early warning based on the obtained multimodal fusion early warning results to output early warning information.

Images