Motion cognition evaluation system based on multi-source electrophysiological information coupling

Abstract

The invention provides a motion cognition evaluation system based on multi-source electrophysiological information coupling. The motion cognition evaluation system can be applied to the technical field of cognition evaluation. The motion cognition evaluation system comprises a multi-source electrophysiological signal synchronous acquisition module which is used for acquiring an electroencephalogram signal, an electromyographic signal and an electrocardiosignal of a target object in the same time period in the process that the target object sequentially carries out motion cognition tasks with various difficulties according to a motion instruction; the processor is electrically connected to the multi-source electrophysiological signal synchronous acquisition module and is used for extracting a first coupling feature of the electroencephalogram signal and the electromyographic signal, extracting a second coupling feature of the electroencephalogram signal and the electrocardiosignal and extracting a third coupling feature of the electrocardiosignal and the electromyographic signal; and processing the first coupling feature, the second coupling feature and the third coupling feature by using a predetermined model to obtain a motion cognition evaluation result of the target object.

Description

Technical Field

[0001] This disclosure relates to the field of cognitive assessment technology, and more specifically to a motor cognitive assessment system based on the coupling of multi-source electrophysiological information. Background Technology

[0002] During physical activity, motor cognition is a crucial ability for individuals to perceive, make decisions about, and regulate movement goals, action strategies, and environmental feedback. Therefore, assessment of motor cognition is of indispensable value in the fields of sports training optimization and neurorehabilitation diagnosis. However, some motor cognition assessment systems suffer from low accuracy in assessing motor cognitive function. Summary of the Invention

[0003] In view of the above problems, this disclosure provides a motor cognitive assessment system based on the coupling of multi-source electrophysiological information.

[0004] According to one aspect of this disclosure, a motor cognitive assessment system based on multi-source electrophysiological information coupling is provided, comprising: a multi-source electrophysiological signal synchronous acquisition module, used to acquire EEG, EMG, and ECG signals of the target object at the same time period during the process of the target object sequentially performing motor cognitive tasks of various difficulties according to motor instructions; and a processor, electrically connected to the multi-source electrophysiological signal synchronous acquisition module, and used to: extract a first coupling feature of EEG and EMG signals, extract a second coupling feature of EEG and ECG signals, and extract a third coupling feature of ECG and EMG signals; and process the first coupling feature, second coupling feature, and third coupling feature using a predetermined model to obtain the motor cognitive assessment result of the target object; wherein, the first coupling feature characterizes the degree of synchronization between the motor instructions generated by the target object's brain regions according to the motor instructions and the response status of the target object's muscle groups to the motor instructions; the second coupling feature characterizes the degree of influence of the cognitive regulation performed by the target object according to the changes in the difficulty of the motor cognitive task on the target object's heart rate; and the third coupling feature characterizes the degree of matching between the target object's physiological state and the energy consumed by the muscle groups.

[0005] According to embodiments of this disclosure, a motor cognitive assessment system is provided. In this system, a multi-source electrophysiological signal synchronous acquisition module acquires EEG, EMG, and ECG signals of the target subject during the same time period while the subject sequentially performs motor cognitive tasks of varying difficulty according to motor instructions. The processor can extract features from the EEG, EMG, and ECG signals to obtain a first coupling feature characterizing the response of the target subject's muscle groups to the motor commands generated by the target subject's brain regions according to the motor instructions; a second coupling feature characterizing the impact of the target subject's cognitive regulation on the target subject's heart rate as the difficulty of the motor task changes; and a third coupling feature characterizing the degree of matching between the target subject's physiological state and the energy consumed by the muscle groups. This fully integrates the "brain-heart-muscle" interaction mechanism of the target subject during the performance of motor cognitive tasks of varying difficulty, effectively overcoming the inherent limitations of single-modal signal analysis, realizing multi-dimensional quantitative representation of motor cognitive function assessment, and improving the accuracy of motor cognitive function assessment.

[0006] Furthermore, the motor cognition assessment system of this disclosure shows good applicability to both static motor tasks (such as fine motor skills of the fingers) and dynamic motor tasks (such as gait analysis), and can achieve accurate assessment and personalized analysis of motor cognitive function for different groups such as healthy people, professional athletes and neurorehabilitation patients.

[0007] Furthermore, based on the quantitative analysis results of objective neurophysiological indicators (i.e., the aforementioned motor cognitive assessment results), the motor cognitive assessment system of this disclosure can provide a scientific assessment basis for the field of neurorehabilitation (e.g., motor cognitive function reconstruction in stroke patients), helping to develop personalized rehabilitation plans. Moreover, in the field of sports training, the motor cognitive assessment system of this disclosure can provide data support for optimizing athlete training strategies, possessing significant clinical application value and sports science research significance. Attached Figure Description

[0008] The above-mentioned contents, other objects, features and advantages of this disclosure will become clearer from the following description of embodiments of this disclosure with reference to the accompanying drawings, which will be described in conjunction with the drawings.

[0009] Figure 1 A schematic diagram of a motor cognition assessment system according to an embodiment of the present disclosure is shown.

[0010] Figure 2 A flowchart illustrating a method for constructing a heat map according to an embodiment of the present disclosure is shown schematically.

[0011] Figure 3The illustration schematically shows a real-time assessment system for motor cognitive function (i.e., a motor cognitive assessment system capable of real-time assessment) and a quantitative diagram of real-time cognitive efficiency according to embodiments of the present disclosure.

[0012] Figure 4 A schematic diagram of a motor cognition assessment system according to another embodiment of the present disclosure is shown. Detailed Implementation

[0013] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0014] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0015] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0016] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0017] In the technical solution disclosed herein, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, and necessary measures have been taken to ensure that they do not violate public order and good morals. Corresponding operation entry points are provided for users to choose to authorize or refuse.

[0018] In scenarios involving automated decision-making using personal information, the methods, devices, and systems provided in this disclosure all offer users corresponding entry points for choosing to agree to or reject the automated decision-making results. If the user chooses to reject, the process proceeds to the expert decision-making stage. Here, "automated decision-making" refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests, or economic, health, and credit status through computer programs, and then making a decision. Here, "expert decision-making" refers to the activity of making decisions by personnel who specialize in a particular field, possess specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0019] During exercise, motor cognition is a crucial ability for individuals to perceive, make decisions, and regulate movement goals, action strategies, and environmental feedback. Therefore, the assessment of motor cognition is of indispensable value in the fields of exercise training optimization and neurorehabilitation diagnosis. The motor cognition assessment technology system mainly encompasses behavioral assessment methods and single-source electrophysiological assessment methods.

[0020] Behavioral assessment methods assess motor skills by quantifying behavioral indicators such as completion time and accuracy of motor tasks (typical applications include finger dexterity tests and gait control tasks). However, this method has significant limitations. Its assessment results are easily influenced by individual subjective states (including fatigue levels and emotional fluctuations), and it is difficult to deeply reveal the neurophysiological mechanisms behind motor cognition, making it difficult to achieve a deep analysis of the motor cognitive process.

[0021] Single-source electrophysiological assessment methods rely solely on a single electrophysiological signal for analysis, including studies based on electroencephalogram (EEG) signals (e.g., changes in motor cortex beta wave power) or electromyography (EMG) signals (e.g., action potential firing frequency). However, this approach neglects the synergistic mechanisms between the brain, heart, and muscle systems. Specifically, it fails to adequately consider the impact of information such as heart rate variability on cognitive resource allocation, as well as the synchronization between EMG signals and brain motor commands. These deficiencies directly lead to insufficient accuracy (e.g., accuracy below 75%) and poor real-time performance (signal processing delay exceeding 500ms) in the assessment results. Furthermore, they struggle to meet the practical needs of dynamic motion scenarios (e.g., complex gait analysis, multi-joint coordinated motion assessment), and their assessment effectiveness in complex motion environments urgently requires improvement.

[0022] Therefore, it is evident that the analytical paradigm based on single-modal physiological signals is limited by the incompleteness of neurophysiological representations, making it difficult to deconstruct the complex dynamic mechanisms of multi-system interactions in motor cognition. Furthermore, some assessment models have inherent defects in their time response characteristics and dynamic adaptability, making it difficult to meet the stringent requirements of real-time performance and accuracy in clinical diagnosis and scientific research.

[0023] In summary, there are core problems such as "single assessment dimensions, lack of research on multi-system coupling mechanisms, and weak adaptability to dynamic environments." There is an urgent need to develop a motion cognitive assessment system with multi-source collaborative analysis capabilities, high precision, and high real-time performance to break through technical bottlenecks and promote the theoretical and technological development of the field of motion cognitive assessment.

[0024] In view of this, embodiments of this disclosure provide a novel paradigm for motor cognitive assessment based on the coupling of multi-source electrophysiological information. By integrating multimodal physiological signals such as electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG), a spatiotemporally consistent cross-modal information coupling analysis framework is constructed to systematically analyze the nonlinear dynamic correlations and complementary representation mechanisms between different physiological signals. The following description is in conjunction with the accompanying drawings.

[0025] Figure 1 A schematic diagram of a motor cognition assessment system according to an embodiment of the present disclosure is shown.

[0026] like Figure 1 As shown, the motor cognition assessment system of this embodiment may include a multi-source electrophysiological signal synchronous acquisition module and a processor that are electrically connected.

[0027] The multi-source electrophysiological signal synchronous acquisition module can acquire the target object's electroencephalogram (EEG), electromyogram (EMG), and electrocardiogram (ECG) signals at the same time period during the target object's sequential performance of various motor cognitive tasks of different difficulties according to the motor instructions.

[0028] For example, motion instructions can be used to instruct a target object to move, such as at least one of voice broadcast instructions or visual display instructions. For the target object, the motion cognition task can be a task of moving according to motion instructions. It should be understood that since the target object needs to perform the motion task according to motion instructions during the motion cognition task, the motion cognition task in the embodiments of this disclosure is not a simple motion task, but is essentially a task of motion cognition for the target object. For example, the motion cognition task may include at least one of visually guided finger grasping tasks or dynamic balance control tasks.

[0029] Furthermore, the target audience can perform various levels of motor cognitive tasks from easy to difficult. For example, motor tasks of varying difficulty may include grasping a single object, grasping objects of different weights, and simultaneously performing object classification tasks during grasping. However, it should be understood that the embodiments disclosed herein are not limited to these. Based on this, each level of difficulty is designed following the principle of progressively increasing cognitive load.

[0030] During this process, the multi-source electrophysiological signal synchronous acquisition module can acquire EEG, EMG, and ECG signals generated by the target object at the same time period. For example, EEG signals may include electroencephalography (EEG), EMG signals may include electromyography (ECG), and ECG signals may include electrocardiography (EMG). It should be understood that acquiring EEG, EMG, and ECG signals at the same time period does not mean that the embodiments of this disclosure only acquired signals in one time period. For example, the embodiments of this disclosure may also acquire signals in multiple consecutive time periods, which will not be elaborated here.

[0031] Based on this, the processor can extract the first coupling feature, the second coupling feature, and the third coupling feature for EEG signals, EMG signals, and ECG signals at the same time period.

[0032] The first coupling feature can be the coupling feature between electroencephalogram (EEG) signals and electromyogram (EMG) signals. This first coupling feature characterizes the degree of synchronization between the motor commands generated by the target subject's brain regions based on motor cues and the response of the target subject's muscle groups to those commands. For example, a motor command can refer to an instruction generated by the target subject's brain regions based on the target subject's cognition of the motor cues. This motor command can be used to control the target subject's muscle groups to perform movements adapted to the motor cues.

[0033] The second coupling feature can be the coupling feature between electroencephalogram (EEG) and electrocardiogram (ECG) signals. This second coupling feature characterizes the degree to which the target subject's cognitive regulation, in response to changes in the difficulty of the motor cognitive task, affects the target subject's heart rate. It should be understood that as the difficulty of the motor cognitive task gradually increases, the target subject needs to utilize more of the brain's cognitive resources to process this change in order to control muscle groups for corresponding movements. Thus, the target subject will engage in cognitive regulation according to changes in the difficulty of the motor cognitive task.

[0034] The third coupling feature can be the coupling feature between electroencephalogram (EEG) signals and electrocardiogram (ECG) signals. This third coupling feature characterizes the degree of matching between the physiological state of the target object and the energy consumed by the muscle groups. For example, the physiological state can include the physiological state of nerves such as the autonomic nervous system, which are used to actively regulate the movement of the target object.

[0035] Thus, the processor can use a predetermined model to process the first coupling feature, the second coupling feature, and the third coupling feature to obtain the motion cognitive assessment result of the target object. The predetermined model may include a trained neural network model, etc., which will not be elaborated here. The motion cognitive assessment result can be used to characterize the target object's cognitive status regarding the motion cognitive task.

[0036] Based on this, in this embodiment, the multi-source electrophysiological signal synchronous acquisition module acquires the target object's electroencephalogram (EEG), electromyogram (EMG), and electrocardiogram (ECG) signals at the same time period during the process of the target object sequentially performing various motor cognitive tasks according to the motor instructions. The processor can extract features from the EEG, EMG, and ECG signals to obtain a first coupling feature that characterizes the response of the target object's muscle groups to the motor commands generated by the target object's brain regions according to the motor instructions; a second coupling feature that characterizes the degree of influence of the target object's cognitive regulation on the target object's heart rate according to the changes in the difficulty of the motor task; and a third coupling feature that characterizes the degree of matching between the target object's physiological state and the energy consumed by the muscle groups. This fully integrates the "brain-heart-muscle" interaction mechanism of the target object during the performance of various motor cognitive tasks, effectively overcoming the inherent limitations of single-modal signal analysis, realizing multi-dimensional quantitative representation of motor cognitive function assessment, and improving the accuracy of motor cognitive function assessment.

[0037] Furthermore, the motor cognition assessment system of this disclosure shows good applicability to both static motor tasks (such as fine motor skills of the fingers) and dynamic motor tasks (such as gait analysis), and can achieve accurate assessment and personalized analysis of motor cognitive function for different groups such as healthy people, professional athletes and neurorehabilitation patients.

[0038] Furthermore, based on the quantitative analysis results of objective neurophysiological indicators (i.e., the aforementioned motor cognitive assessment results), the motor cognitive assessment system of this disclosure can provide a scientific assessment basis for the field of neurorehabilitation (e.g., motor cognitive function reconstruction in stroke patients), helping to develop personalized rehabilitation plans. Moreover, in the field of sports training, the motor cognitive assessment system of this disclosure can provide data support for optimizing athlete training strategies, possessing significant clinical application value and sports science research significance.

[0039] The motor cognitive assessment system based on multimodal physiological signals disclosed in this embodiment mainly involves four core components: At the signal acquisition level, a multi-channel sensor array with high temporal resolution and high common-mode rejection ratio is employed, combined with a synchronous triggering mechanism to achieve spatiotemporal alignment of multimodal signals; in the signal preprocessing stage, noise suppression and feature enhancement are achieved through the synergistic effect of wavelet multi-resolution analysis and adaptive Kalman filtering algorithm; in the coupling analysis stage, a quantitative model is constructed based on the Dynamic Time Warping (DTW) algorithm and mutual information (MI) theory to reveal the time delay characteristics and information transmission paths between multimodal signals; in the assessment modeling stage, a multimodal feature fusion network is constructed based on deep learning, and objective quantification and real-time dynamic assessment of motor cognitive function are achieved through end-to-end supervised learning. This motor cognitive assessment system provides an innovative technical solution and theoretical support system for cognitive function assessment in interdisciplinary fields such as sports medicine and neurorehabilitation. The details are described below.

[0040] In this embodiment of the disclosure, the multi-source electrophysiological signal synchronous acquisition module may include a multimodal analog front-end acquisition unit. For example, the multimodal analog front-end acquisition unit may include an electroencephalogram (EEG) acquisition unit, an electromyography (EMG) acquisition unit, and an electrocardiogram (ECG) acquisition unit.

[0041] The EEG acquisition unit may include an EEG cap (e.g., a 32-channel EEG cap). For example, the reference electrodes of the EEG cap may be fixed to the mastoid process behind the ear of the target subject. Based on this, the EEG acquisition unit can acquire EEG signals at a sampling frequency of 2000 Hz based on the EEG activity of the motor cortex (e.g., C3 and C4 positions) and the prefrontal cortex (e.g., F3 and F4 positions).

[0042] The electromyography (EMG) acquisition unit may include surface EMG electrodes. For example, the EMG acquisition unit can record the EMG activity of movement-related muscle groups (such as the biceps brachii and tibialis anterior) at a sampling frequency of 2000 Hz, accurately capturing the temporal characteristics and intensity changes of muscle activation as EMG signals.

[0043] The ECG acquisition unit may include chest lead electrodes. For example, the ECG acquisition unit can acquire ECG signals related to heart rate variability (HRV) at a sampling rate of 500 Hz. This allows for the further extraction of information such as the standard deviation of normal sinus intervals (SDNN).

[0044] However, the embodiments disclosed herein are not limited thereto. Since the aforementioned electroencephalogram (EEG), electromyogram (EMG), and electrocardiogram (ECG) signals are weak and easily interfered with, an analog front-end (AFE) chip can be used for signal detection and preliminary conditioning. Based on this, differentiated circuit designs can be implemented for the aforementioned EEG acquisition unit, EMG acquisition unit, and ECG acquisition unit according to the characteristics of different signals.

[0045] For example, the EEG acquisition unit can be implemented based on a multi-channel high-precision analog front-end chip, which can be paired with the active electrodes of a 32-channel EEG cap to acquire EEG signals. Furthermore, this multi-channel high-precision analog front-end chip can incorporate a low-noise programmable gain amplifier (PGA) to amplify weak EEG signals to the range of an analog-to-digital converter, while reducing common-mode interference through a driving circuit (e.g., a driving circuit deployed on the target subject's right leg). For electrodes in target areas such as C3 and C4, monopolar or bipolar configurations can be used. The reference electrode signal can be connected to a dedicated reference terminal of the multi-channel high-precision analog front-end chip to match the reference electrode layout of the mastoid process behind the ear.

[0046] The electromyography (EMG) acquisition unit can be implemented based on surface EMG electrodes combined with a multi-channel signal conditioning circuit, and the chip can also be reused to expand the channels. For the EMG signal initially acquired by the EMG acquisition unit, a high-pass filter needs to be set to remove baseline drift, the EMG signal is amplified by a PGA, and then a low-pass anti-aliasing filter is used to suppress high-frequency noise, ensuring that the muscle activation signal is not distorted at the 2000 Hz sampling frequency, and accurately capturing the activity signals of target muscle groups such as the biceps brachii as EMG signals.

[0047] The ECG acquisition unit can utilize chest lead electrodes in conjunction with a multi-channel high-precision analog front-end chip. This chip can be configured with 8 channels and capable of 24-bit sampling. Based on this, the ECG acquisition unit can simultaneously acquire ECG signals and extract the RR interval sequence. Furthermore, the ECG acquisition unit can be synchronized with the EEG signal by configuring the sampling timing. For example, the RR interval sequence can refer to the time interval sequence between two adjacent QRS complex apexes (i.e., R waves) extracted from the ECG.

[0048] The processor can be electrically connected to the EEG, EMG, and ECG acquisition units to receive EEG, EMG, and ECG signals from these units. The processor can serve as the "central hub" of the multi-source electrophysiological signal synchronous acquisition module, responsible for unified scheduling of each acquisition channel, clock synchronization control, and preliminary data processing. It needs to have multiple peripheral interfaces and stable computing capabilities. For example, the processor can include a microcontroller. Specifically, the processor can be a processor with integrated edge computing capabilities. For instance, the processor can control the multi-channel sampling timing through configuration registers, adapt to a 12-bit built-in analog-to-digital converter, and support external high-precision analog-to-digital converters. Furthermore, the processor can be equipped with a neural network accelerator, enabling simultaneous sampling control and preliminary signal feature extraction, adapting to scenarios requiring edge-side artificial intelligence computing. Moreover, the processor can be connected to the same high-precision crystal oscillator to provide a unified clock signal to the EEG, EMG, and ECG acquisition units, ensuring consistent signal sampling timing from the source. In this way, the EEG, EMG, and ECG acquisition units share the same high-stability crystal oscillator clock signal as the control clock, thereby synchronously controlling the sampling rhythm of the analog-to-digital converters of each unit, ensuring synchronized sampling beats. Furthermore, the processor can distribute the aforementioned clock signal to the EEG, EMG, and ECG acquisition units via a low-skew buffer, reducing the time difference in clock signal edges reaching different units, thus solidifying the foundation for real-time consistency.

[0049] Furthermore, the multimodal simulation front-end acquisition unit may also include a synchronization triggering unit. The synchronization triggering unit can be electrically connected to the processor, the EEG acquisition unit, the EMG acquisition unit, and the ECG acquisition unit. Thus, under the control of a sampling start command from the processor, the synchronization triggering unit can simultaneously send trigger signals to the EEG acquisition unit, the EMG acquisition unit, and the ECG acquisition unit, enabling them to sample simultaneously and obtain EEG, EMG, and ECG signals at the same time.

[0050] Specifically, in this embodiment, to achieve a time synchronization accuracy of ±1 ms, a separate synchronization trigger unit needs to be designed as a "cooperative bridge" between the processor and the EEG, EMG, and ECG acquisition units. The synchronization trigger unit can be implemented based on an optocoupler or a dedicated synchronization trigger chip. Based on this, when the processor issues a sampling start command, the trigger signal can be simultaneously transmitted via the synchronization trigger unit to the aforementioned EEG, EMG, and ECG acquisition units through the EEG, ECG, and EMG acquisition channels, respectively, so that the EEG, EMG, and ECG acquisition units start sampling at the same time. Thus, for a synchronization trigger unit that enables high precision throughout the process, it can synchronously start the EEG, EMG, and ECG acquisition units through the trigger signal, thereby at least partially avoiding signal offset due to different start times. In this embodiment, this synchronization trigger mechanism covers the entire acquisition process, maintaining timing consistency at least partially from signal acquisition to data transmission, thereby at least partially avoiding the risk of losing synchronization midway.

[0051] Furthermore, the multimodal simulation front-end acquisition unit may also include a timestamp marking circuit. This timestamp marking circuit can be electrically connected to the EEG acquisition unit, EMG acquisition unit, and ECG acquisition unit. Thus, the timestamp marking circuit can receive EEG, EMG, and ECG signals at each moment from the EEG, EMG, and ECG acquisition units, and set corresponding timestamps for each EEG, EMG, and ECG signal according to the processor's clock signal. Therefore, by placing a timestamp marking circuit between the multimodal simulation front-end acquisition unit and the processor, each frame of data can be timestamped using a unified processor clock, at least partially improving the reliability of subsequent data alignment.

[0052] Furthermore, the timestamp marking circuit can also be electrically connected to the processor. Thus, the processor can receive EEG, EMG, and ECG signals via the timestamp marking circuit. Furthermore, based on the timestamps of the received EEG, EMG, and ECG signals, the processor can resample and align the signals, and fill in the sampling gaps caused by the resampling and alignment. This allows for time synchronization of the EEG, EMG, and ECG signals while preserving their changing trends, resulting in EEG, EMG, and ECG signals from the same time period.

[0053] For example, the processor can resample and align the received EEG, EMG, and ECG signals based on a unified timestamp from the original acquisition. On this basis, the frequencies of the EEG and EMG signals can be downsampled from 2000 Hz to 1000 Hz, and the frequency of the ECG signal can be upsampled from 500 Hz to 1000 Hz to at least partially ensure temporal consistency. Subsequently, linear interpolation can be used to fill the sampling gaps between the resampled EEG, EMG, and ECG signals, and at least partially preserve the original signal trends, achieving millisecond-level temporal synchronization of the EEG, EMG, and ECG signals, thus at least partially ensuring the reliability of subsequent multimodal fusion analysis.

[0054] Thus, in this embodiment, the synchronization triggering unit, processor, and shared clock signal work in deep coordination to achieve strict alignment of the sampling times of the aforementioned EEG acquisition unit, EMG acquisition unit, and ECG acquisition unit, thereby achieving a time synchronization accuracy of ±1ms. Based on this, by employing unified processor scheduling, signal delay deviations caused by multi-unit collaboration can be at least partially avoided. Furthermore, timestamps ensure that EEG, EMG, and ECG signals accurately correspond to the same physical event (such as the instant a motor cognitive task begins), at least partially guaranteeing the reliability of the phase relationship between signals and at least partially avoiding feature misjudgment caused by timing misalignment, providing a reliable data foundation for subsequent fusion operations such as cross-signal correlation analysis and coupling coefficient calculation. Moreover, this approach simplifies the time calibration steps in subsequent data preprocessing operations, improving the efficiency and accuracy of multimodal analysis.

[0055] Based on this, the multi-source electrophysiological signal synchronous acquisition module of this disclosure adopts a three-in-one design of "unified processor control + shared crystal oscillator clock + high-precision synchronous triggering" to achieve a high degree of consistency in the timestamps of EEG signals, EMG signals and ECG signals, so that the time synchronization accuracy reaches ±1 ms, providing a key guarantee for multimodal data fusion.

[0056] In addition, the multimodal simulation front-end acquisition unit may also include a data storage and transmission unit. This data storage and transmission unit can be used to temporarily store EEG, EMG, and ECG signals from the processor and transmit these three signals to the host computer, avoiding data loss and supporting subsequent analysis. The data storage and transmission unit may include a storage end and a transmission end. The storage end can be connected to an external pseudo-static random access memory (PSRAM) as a cache for temporarily storing high-speed acquired multi-channel EEG, EMG, and ECG signals. It can also be paired with a flash memory chip to achieve power-off retention of key data and save the EEG, EMG, and ECG signals in a timestamp format for easy traceability. The transmission end can integrate a Bluetooth Low Energy (BLE) transmission module or a Wi-Fi transmission module. The BLE transmission module is suitable for short-range, low-power transmission scenarios, while the Wi-Fi transmission module is suitable for scenarios requiring long-distance transmission. Both can communicate with the processor through a serial peripheral interface (SPI) or an integrated circuit bus (I2C) interface to transmit synchronized EEG, EMG, and ECG signals in real time.

[0057] Furthermore, the multi-source electrophysiological signal synchronous acquisition module may also include a power supply and safety isolation unit. This power supply and safety isolation unit ensures stable operation of the motor cognitive assessment system and complies with medical device safety specifications. Specifically, the power supply and safety isolation unit may include a medical-grade isolated power supply, paired with a dedicated power management chip (PMIC). Based on this, the power supply and safety isolation unit can generate multiple programmable voltage rails to provide stable power to the processor, multimodal analog front-end acquisition unit, and other devices, avoiding power interference between different units. In terms of safety, the power supply and safety isolation unit can achieve electrical isolation between the signal path and the power path through optocouplers or isolated analog-to-digital converters to control leakage current to less than 10 μA, meeting relevant medical safety standards. Additionally, a varistor can be added at the electrode interface to prevent electrostatic discharge or sudden current from damaging the circuit and affecting safety.

[0058] Furthermore, the processor can preprocess the received electroencephalogram (EEG), electromyogram (EMG), and electrocardiogram (ECG) signals.

[0059] For example, for EEG signals, the processor can perform layered noise and artifact removal to purify the effective frequency bands of the EEG, thereby obtaining high-quality EEG signals. Specifically, the processor can use wavelet transform technology based on the db4 wavelet basis to accurately filter out 50 Hz power frequency interference and low-frequency drift in the EEG signal through multi-scale decomposition and reconstruction, while separating electrooculography (EOG) artifacts to avoid damage to the details of the EEG signal caused by some filtering methods. Furthermore, the processor can use the Independent Component Analysis (ICA) algorithm to perform blind source separation on the EEG signal, efficiently removing motion artifacts caused by head movements and muscle contractions, and preserving the original true phase characteristics of the EEG signal. Furthermore, the processor can also perform noise reduction on the EEG signal based on wavelet soft thresholding. Specifically, dedicated thresholds can be set for the α, β, and γ wave bands to further suppress residual noise and improve the signal-to-noise ratio. Furthermore, wavelet packet decomposition technology is used to selectively extract signals from three core frequency bands: alpha waves (8–13 Hz), beta waves (13–30 Hz), and gamma waves (30–80 Hz), while eliminating irrelevant frequency interference. Thus, by reconstructing and integrating the filtered frequency band signals, a time-domain continuous and frequency-domain pure EEG signal is output, laying the foundation for the correlation analysis between cognitive function and EEG activity.

[0060] For ECG signals, the processor can perform targeted artifact suppression to accurately extract high-quality ECG signals. Specifically, a signal separator based on a least mean square adaptive filtering algorithm can be deployed in the processor. The processor can then use the delayed ECG signal as a reference input, leveraging the correlation difference between EMG interference and the ECG signal to specifically cancel EMG artifacts in the 2–2000 Hz frequency band. Furthermore, a 50 Hz adaptive notch filter can be superimposed to further eliminate power frequency interference in the ECG signal. The filter weights can be dynamically adjusted to adapt to real-time changes in signal noise, ensuring filtering stability. This results in a high-quality ECG signal.

[0061] For example, for electromyography (EMG) signals, the processor can optimize them based on baseline correction and noise suppression techniques. Specifically, a second-order Butterworth high-pass filter with a cutoff frequency of 10 Hz can be used to effectively remove baseline drift, combined with zero-phase IIR filtering to avoid signal phase distortion. Furthermore, a 50 Hz notch filter can be added to eliminate the superposition effect of power frequency interference on the EMG signal, improving the integrity of the signal in the 10–500 Hz effective frequency band. Further, after noise reduction, the EMG signal can be fully rectified and smoothed using a fourth-order Butterworth low-pass filter. Subsequently, a 50ms sliding window + overlapping window strategy can be used for processing to obtain the desired EMG signal. This improves the stability of the obtained EMG signal.

[0062] Based on this, and considering the noise characteristics, amplitude differences, and analysis objectives of EEG, EMG, and ECG signals, this disclosure constructs a differentiated preprocessing system of "targeted denoising + precise extraction + temporal alignment," which can output high-purity, correlated, and analytically valuable standardized data to support subsequent multimodal fusion analysis.

[0063] Furthermore, the processor can extract the beta wave of the target object from the EEG signal and calculate the phase synchronization index (PSI) between the beta wave and the electromyographic signal. Thus, the PSI can quantify the degree of synchronization between the brain's motor commands and muscle execution, facilitating the assessment of the phase coupling characteristics of the EEG and EMG signals. Based on this, the PSI can be defined as the first coupling characteristic. In addition, the first coupling characteristic may also include indicators such as the cross-correlation coefficient (CCC) between the EEG and EMG signals, which will not be elaborated upon here.

[0064] The processor can also determine the heart rate variability signal of the target object based on the electrocardiogram signal, and extract the high-frequency power and low-frequency power of the heart rate variability signal.

[0065] For example, the processor can detect and quantify the features of the electrocardiogram (ECG) signal. Specifically, the Pan-Tompkins algorithm can be used to accurately identify the start, peak, and end points of the QRS complex through differential operations and adaptive threshold adjustment. This method is more robust to interference than the traditional thresholding method. Subsequently, the core parameters of the heart rate variability signal can be calculated based on the identification results: time-domain indicators include the standard deviation of the normal sinus interval and the root mean square difference between adjacent normal intervals (RMSSD); time-frequency indicators include high-frequency power (HF) and low-frequency power (LF), comprehensively reflecting the state of autonomic nervous system function. In addition, time-frequency indicators may also include the power ratio (LF / HF), which will not be elaborated here.

[0066] Subsequently, the processor can extract the gamma wave power of the target object from the EEG signal and determine the second coupling feature based on the degree of influence of the gamma wave power on the high-frequency power.

[0067] For example, Granger causality analysis can be used to determine the causal relationship between the power of gamma waves and high-frequency power in the prefrontal cortex of electroencephalograms (EEGs), revealing the mechanism by which cognitive modulation affects heart rate.

[0068] Specifically, when the target object is in a resting state, the effect of gamma waves on the gain control (GC) value of the HF is not significant (p > 0.05), and the effect of HF on the GC value of gamma waves is not significant. Here, p represents the probability value. When the target object is performing a motor cognitive task, the effect of gamma waves on the GC value of HF is significant (p < 0.05, GC value = 0.3), while the effect of HF on gamma waves is not significant. Based on this, it can be concluded that when the target object is performing a motor cognitive task, gamma waves generated in the prefrontal cortex enhance parasympathetic activity through neural pathways (e.g., sequentially including the prefrontal cortex, hypothalamus, and dorsal nucleus of the vagus nerve), leading to an increase in HF power, supporting the mechanism that "cognitive regulation affects heart rate through brain-heart coupling." Based on this, the second coupling characteristic can be determined according to the change in GC value within the same time period.

[0069] The processor can also extract the average power frequency of the electromyographic signal, calculate the similarity distance between the low-frequency power and the average power frequency, and determine the similarity distance as the third coupling feature.

[0070] Specifically, the mean power frequency (MPF) can be extracted from electromyography (EMG) signals to reflect muscle fatigue status and support motor performance synergy analysis. Based on this, the dynamic time warping (DTW) algorithm can be used to calculate the similarity distance between low-frequency power and the mean power frequency, and this similarity distance is used as a third coupling feature to reflect the degree of matching between the target's physiological state and muscle energy consumption. Furthermore, integrated electromyography (iEMG) values ​​are calculated to quantify the total intensity of muscle activation; details of this will not be elaborated upon here.

[0071] Furthermore, in this embodiment of the disclosure, the electroencephalogram (EEG) signal includes EEG signals from multiple brain regions of the target object. For example, these multiple brain regions may be brain regions related to motor cognition. For example, the multiple brain regions may include brain regions such as the primary motor cortex and the prefrontal cortex. Correspondingly, the electromyographic (EMG) signal may include EMG signals from multiple muscle groups of the target object. Specifically, these multiple muscle groups may include muscle groups controlled by motor commands from the primary motor cortex, or muscle groups controlled by motor commands from the prefrontal cortex, etc. Based on this, the first coupling feature can be used to characterize the degree of synchronization between the motor commands of the multiple brain regions and the response status of the corresponding muscle groups in the multiple muscle groups.

[0072] Based on this, after inputting the first coupling feature into the predetermined model, the model can analyze the brain-muscle synergy of the target object according to the degree of synchronization between motor commands from multiple brain regions and corresponding muscle groups. Thus, compared to analyzing the first coupling feature based on motor commands from a single brain region and muscle groups, the predetermined model can more comprehensively analyze the brain-muscle synergy of the target object, improving the accuracy of motor cognitive assessment results.

[0073] In this embodiment of the disclosure, the predetermined model can be a neural network that incorporates an attention mechanism. This neural network can be a hybrid neural network built based on a convolutional neural network (CNN) and a long short-term memory network (LSTM). This will be described in detail below.

[0074] For example, the predetermined model includes a convolutional neural network (CNN). Based on this, the processor can use the CNN to process the first, second, and third coupling features to obtain convolutional coupling features. These convolutional coupling features can characterize the differences in response states between different brain regions. Furthermore, the CNN can use convolution operations to mine the spatial topological relationships of the coupling features, analyzing the differences in signal interactions between different brain regions and muscle groups. Subsequently, the processor can further process the convolutional coupling features using the predetermined model to obtain motor cognitive assessment results. This improves the accuracy of the motor cognitive assessment results.

[0075] Furthermore, the processor can also be equipped with a principal component analysis (PCA) algorithm. Before inputting the first, second, and third coupling features into the convolutional neural network, the processor uses PCA to reduce the dimensionality of these features. Then, the convolutional neural network processes the dimensionality-reduced features to obtain convolutional coupling features. Thus, by using PCA to reduce the dimensionality of these features, the processor can select principal component features with a cumulative contribution rate exceeding 90% as the effective feature set, thereby improving the accuracy of motion cognitive assessment results. It should be understood that the effective feature set here includes the dimensionality-reduced first, second, and third coupling features.

[0076] For example, the multi-source electrophysiological signal synchronous acquisition module can also acquire N sets of electrophysiological signals of the target object in N time periods. N is an integer greater than 1. Each set of electrophysiological signals includes EEG, EMG, and ECG signals in the same time period. The predetermined model may also include a long short-term memory network and an attention mechanism layer. Based on this, the processor can use the attention mechanism layer to process the convolutional coupling features to obtain attention coupling features. Subsequently, the processor can use the long short-term memory network to process the convolutional coupling features in N time periods to obtain spatiotemporal fusion features. This spatiotemporal fusion feature characterizes the changes in convolutional coupling features as the difficulty of the motor cognitive task changes. Subsequently, the processor can use the output layer to process the spatiotemporal fusion features to obtain the motor cognitive assessment results. Based on this, the embodiments of this disclosure utilize the temporal modeling capability of the long short-term memory network to capture the dynamic evolution of coupling features over time in the motor cognitive process, especially the feature transfer process when the task difficulty gradient changes.

[0077] Furthermore, in this embodiment, a feature weighting module for the attention mechanism layer is designed based on the self-attention mechanism, assigning differentiated weights to key indicators representing brain-muscle synergy (such as phase synchronization index and cross-correlation coefficient) to enhance the model's ability to perceive core information. Based on this, a labeled multi-dimensional dataset (i.e., a dataset including EEG signals, EMG signals, and ECG signals) can be used to train the predetermined model. This dataset covers the corresponding data of "coupled features - motor cognitive assessment results" for target objects at different levels of motor cognition. Supervised learning is used to achieve a quantitative assessment of motor cognitive function. The motor cognitive assessment results may include a motor cognitive efficiency index and a cognitive load level. The motor cognitive efficiency index characterizes the efficiency of the target object's cognition of various difficulty levels of motor cognitive tasks on muscle groups. Specifically, this embodiment constructs a two-dimensional evaluation index, where the motor cognitive efficiency index quantifies cognitive regulation effectiveness in the [0,1] interval, with higher values ​​indicating better cognitive resource utilization efficiency; the cognitive load level uses a five-level scale system to intuitively reflect the cognitive pressure level under task complexity.

[0078] It should be understood that the coupling features in the above dataset refer to the training data corresponding to the first coupling feature, the second coupling feature, and the third coupling feature. The motion cognitive evaluation results in the above dataset refer to the labels used for model training, which will not be elaborated upon here. Similarly, it should be understood that since the labels during model training include data such as the motion cognitive efficiency index and cognitive load level, the motion cognitive evaluation results during the application of the predetermined model can also include data such as the motion cognitive efficiency index and cognitive load level, which will not be elaborated upon here.

[0079] Based on this, the embodiments of this disclosure deeply explore the multi-source coupling characteristics and construct a predetermined model based on the attention mechanism by combining convolutional neural networks and long short-term memory networks, which can obtain accurate motion cognitive assessment results. Empirical studies have verified that the motion cognitive assessment system of the embodiments of this disclosure achieves an accuracy rate of over 92% in motion cognitive function assessment, with a response latency controlled within 200 ms, providing reliable technical support for real-time monitoring in dynamic motion scenarios.

[0080] Furthermore, this disclosure provides a visualization analysis framework. Specifically, a real-time dynamic visualization system can be developed to simultaneously present the time-series change curves of coupled features and the heatmap of evaluation results, and support longitudinal comparative analysis of historical data and multi-dimensional parameter linkage queries.

[0081] The following describes the motor cognition assessment system of this disclosure in conjunction with specific embodiments.

[0082] Figure 2 A flowchart illustrating a method for constructing a heat map according to an embodiment of the present disclosure is shown schematically.

[0083] like Figure 2 As shown, the thermal map construction method of this embodiment includes operations S210~S220.

[0084] During operation of S210, electroencephalogram (EEG), electromyography (EMG), and electrocardiogram (ECG) signals are acquired.

[0085] In operation S220, the EEG signal, EMG signal, and ECG signal are preprocessed, and the first coupling feature, second coupling feature, and third coupling feature are extracted.

[0086] In operation S230, the motion cognitive assessment results are obtained based on the first coupling feature, the second coupling feature, and the third coupling feature.

[0087] During operation of S240, a heat map is constructed based on the results of motion cognitive assessment.

[0088] For example, the target population could include 20 healthy adult participants (aged 20–30 years) and 10 patients in the recovery period after stroke (3–6 months post-stroke) with clinically confirmed mild motor cognitive impairment. It should be understood that all participants have signed informed consent forms, and this study protocol has been approved by the medical ethics committee.

[0089] For data acquisition equipment, a 32-lead EEG acquisition system (e.g., sampling accuracy up to 24 bits), a three-lead ECG acquisition module (e.g., signal conditioning circuitry can be integrated), or an 8-channel EMG acquisition system (e.g., operating bandwidth of 20~450 Hz) can be used, equipped with a synchronous triggering unit (time synchronization accuracy ±1 ms) to ensure the spatiotemporal consistency of multimodal EEG signals, EMG signals, and ECG signals.

[0090] For motor cognition tasks, a visually guided "step-by-step finger grasping task" paradigm can be constructed. This paradigm includes three difficulty levels: the beginner task involves grasping a single object; the intermediate task requires grasping objects of different weights; and the advanced task requires simultaneous object classification during the grasping process. Each difficulty level is designed following the principle of progressively increasing workload.

[0091] Based on this, after wearing the complete multi-source electrophysiological signal synchronous acquisition module, the subjects sequentially performed a three-level motor cognitive task. Throughout the experiment, electroencephalogram (EEG) signals (C3, C4, F3, F4 electrode leads), electrocardiogram (chest lead II), and electromyographic (EMG) signals (biceps brachii and radial wrist flexor muscles) were acquired synchronously. Acquisition lasted for 5 minutes at each task stage, with the sampling frequency strictly adhering to the technical parameters of each EEG, EMG, and ECG acquisition unit to ensure signal integrity.

[0092] Subsequently, for the processing of electroencephalogram (EEG), db4 wavelet transform can be used to suppress 50 Hz power frequency interference, combined with independent component analysis (ICA) to remove electrooculogram artifacts, and the power features of the β band (13~30 Hz) can be extracted.

[0093] For electrocardiogram (ECG) processing, an adaptive filtering algorithm can be used for noise reduction, and the Pan-Tompkins algorithm can be used to achieve accurate detection of QRS complexes, thereby calculating key indicators such as SDNN and high-frequency power (HF) in heart rate variability (HRV).

[0094] For electromyography (EMG) processing, a 10 Hz high-pass filter can be implemented to eliminate baseline drift. After full-wave rectification, a 50 ms moving average is performed to finally calculate the integrated electromyography value (iEMG) and the characteristic parameters of the mean power frequency (MPF).

[0095] Afterwards, brain-muscle coupling analysis, brain-heart coupling analysis, and heart-muscle coupling analysis can be performed.

[0096] For brain-muscle coupling analysis, the characteristics of neuromuscular coordinated activity can be quantified by calculating the cross-correlation coefficient between the C3 lead β-band EEG signal and the biceps brachii EMG signal.

[0097] For brain-heart coupling analysis, the Granger causality test can be used to assess the strength of the causal association between the F3 lead gamma band EEG signal and the HRV high-frequency power.

[0098] For cardiomyopathy coupling analysis, the similarity measure between HRV low-frequency power and electromyographic MPF characteristics can be calculated using the dynamic time warping (DTW) algorithm.

[0099] Next, feature dimensionality reduction can be performed. Specifically, principal component analysis (PCA) is used to reduce the dimensionality of the coupled feature matrix, and the top three principal components with a cumulative contribution rate of 92.3% are selected as effective features.

[0100] Next, an 8:2 data partitioning ratio can be used, with 80% of the data used for training the intended model and the remaining 20% ​​used to evaluate the model's generalization ability. For example, a learning rate of 0.001 can be set for 100 epochs of iterative training, with a batch size of 8.

[0101] Based on this, the trained pre-defined model achieved assessment accuracies of 95.2% and 90.5% in healthy population and stroke patient test sets, respectively, with an average response latency of 180 ms. Compared to single-modal EEG assessment methods (accuracy 78.3%, latency 350 ms), the motor cognitive assessment system of this disclosure demonstrates significant advantages in both accuracy and real-time performance.

[0102] Furthermore, the following section details the implementation process for constructing the motor cognition assessment model and its multi-dimensional output.

[0103] For example, a standardized dataset of preprocessed physiological signals from multiple sources can be selected, which includes three types of coupling features (such as phase synchronization index, cross-correlation coefficient, etc.) of EEG-EMG, EEG-ECG, and ECG-EMG, as well as corresponding subject annotation information (age, basic level of motor cognition, task completion quality score).

[0104] Based on this, a "coupled feature-cognitive ability level" labeling system can be constructed, which is divided into 5 levels according to cognitive regulation efficacy (level 1 is the lowest and level 5 is the highest) to ensure that the sample size of each level is balanced (sample ratio error ≤ 10%) and avoid model bias.

[0105] Subsequently, principal component analysis (PCA) can be used to reduce the dimensionality of high-dimensional coupled features, retaining principal components with a cumulative variance contribution rate of ≥90%, thereby reducing the interference of redundant features on model training.

[0106] By using Pearson correlation coefficient analysis, features strongly correlated with cognitive ability levels (absolute correlation coefficient ≥ 0.6) are selected. Combined with domain knowledge, core indicators of brain-muscle synergy (such as the phase synchronization index of gamma waves and electromyography, and the LF / HF ratio of HRV) are retained, and finally, the model input feature set is formed.

[0107] Furthermore, the training, validation, and test sets can be divided in a 7:2:1 ratio to ensure consistent dataset distribution (the difference in age and cognitive level distribution among subjects in each set is ≤5%). Subsequently, the input features can be Z-score standardized (mean = 0, variance = 1) to eliminate the influence of differences in feature amplitudes and improve model convergence speed.

[0108] Furthermore, for the predetermined model, the input feature dimension is the number of coupled features after dimensionality reduction (assumed to be N-dimensional), the designed time step is 50 (corresponding to the coupled feature sequence of a 5-second physiological signal), and the input format is [batch_size, time_step, feature_dim] (batch size × time step × feature dimension). The batch size (batch_size) is set to 32 to adapt to the parallel computing capabilities of the graphics processing unit, while avoiding training instability caused by excessively small batch sizes.

[0109] For a convolutional neural network with a predetermined model, it may include a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, and a flattened layer.

[0110] For the first convolutional layer: the number of convolutional kernels is 64, the size is (3,3), the stride is (1,1), the padding is "same", the activation function is ReLU (i.e., rectified linear unit), and the correlation of local spatial features is mined.

[0111] For the first pooling layer: it is used to achieve max pooling, with a pooling kernel size of (2,2) and a stride of (2,2), which reduces the feature dimension and preserves key information.

[0112] For the second convolutional layer: the number of convolutional kernels is 128, the size is (3,3), the stride is (1,1), the padding is "same", and the activation function is ReLU to deepen the extraction of spatial features.

[0113] For the second pooling layer: used to implement max pooling, with a pooling kernel size of (2,2), a stride of (2,2), and an output spatial feature map.

[0114] For the third convolutional layer: the number of convolutional kernels is 256, the size is (3,3), the stride is (1,1), the padding is "same", the activation function is ReLU, and the focus is on the interaction difference features between brain regions and muscle groups.

[0115] For flat layers: the two-dimensional feature maps output by the convolutional layers are converted into one-dimensional feature vectors to prepare for input into the long short-term memory network.

[0116] Furthermore, for the attention mechanism layer in the predefined model: a 4-head self-attention structure is adopted, and the feature attention weights are calculated through Query, Key, and Value matrix operations. The dimension of each attention head is equal to the output dimension of the flat layer / 4.

[0117] Higher weights (weight coefficient ≥ 1.5) are assigned to the feature vectors corresponding to core indicators such as phase synchronization index and cross-correlation coefficient, while lower weights (weight coefficient ≤ 0.8) are assigned to secondary features. The weight distribution is then normalized using the Softmax function.

[0118] The output features of the four attention heads are concatenated and integrated through a fully connected layer (512 hidden units) to output a weighted feature vector.

[0119] For a long short-term memory network in a predefined model, it may include a first bidirectional LSTM layer, a second bidirectional LSTM layer, a dropout layer, and a fully connected layer.

[0120] For the first layer bidirectional long short-term memory network module: the number of hidden layer units is 256, the activation function is tanh, the forget gate threshold is set to 0.8, the memory gate threshold is set to 0.5, and the short-term temporal evolution of the coupling features is captured.

[0121] For the second-layer bidirectional long short-term memory network module: the number of hidden layer units is 128, the activation function is tanh, and the focus is on tracking the feature transfer trajectory when the task difficulty gradient changes (such as feature changes from beginner to intermediate and then to advanced tasks).

[0122] Random dropout layer: The dropout rate is set to 0.3 to prevent the model from overfitting and improve generalization ability.

[0123] Fully connected layer: 64 hidden units, ReLU activation function, integrates temporal and spatial features, output spatiotemporal fusion features.

[0124] The output layer may include a dual-output branch structure, etc. For the dual-output branch structure: the first branch is used for predicting the motor cognitive efficiency index, using the Sigmoid activation function, and outputs continuous values ​​in the interval [0,1]; the second branch is used for cognitive load level classification, using the Softmax activation function, and outputs the probability distribution of multiple categories (e.g., 5 categories) (corresponding to load levels 1 to 5).

[0125] Loss function: A weighted loss function is adopted. The mean squared error (MSE) loss is used for predicting the motor cognitive efficiency index, and the cross-entropy loss is used for cognitive load level classification. The weights are 0.6 and 0.4, respectively, to match the importance of the two-dimensional assessment.

[0126] In addition, during the training of the predetermined model, parameters can be configured in the following ways.

[0127] For the optimizer: the Adam optimizer is selected, the initial learning rate is set to 0.001, and a dynamic learning rate adjustment strategy is adopted (the learning rate is halved every 5 epochs if the effect on the validation set improves).

[0128] Regarding the number of training epochs: the maximum number of epochs is set to 50, and an early stopping mechanism is introduced (training stops if the loss on the validation set does not decrease after 8 consecutive epochs) to avoid overfitting.

[0129] For regularization: Add L2 regularization (λ=0.001) to the fully connected layer to suppress overfitting caused by excessively large parameters.

[0130] Based on this, the loss value, accuracy (cognitive load classification accuracy), and root mean square error (RMSE, or motor cognitive efficiency index prediction) of the training / validation set can be monitored in real time, and training curves can be plotted.

[0131] Every 5 epochs, a model checkpoint is output, and the parameters of the model with the best performance on the validation set are saved (judged by "highest classification accuracy + lowest RMSE").

[0132] Based on this, test set verification, ablation experiments, and case verification can be carried out.

[0133] For test set validation: Calculate the cognitive load classification accuracy (target ≥ 85%) and the motor cognitive efficiency index prediction RMSE (target ≤ 0.05) on the test set to evaluate the model's generalization ability.

[0134] For ablation experiments: remove the attention mechanism layer, CNN, and LSTM respectively, compare the decline in model performance, and verify the effectiveness of each module (e.g., an accuracy decrease of ≤10% after removing the attention mechanism layer is considered acceptable).

[0135] For case validation: Select data from subjects with different cognitive levels (10 cases each of low, medium and high), and manually check the consistency between the model output results and the actual cognitive abilities. A consistency of ≥90% is considered passing.

[0136] Based on this, the output of the predetermined model can be quantized. Figure 3The illustration schematically shows a real-time assessment system for motor cognitive function (i.e., a motor cognitive assessment system capable of real-time assessment) and a quantitative diagram of real-time cognitive efficiency according to embodiments of the present disclosure.

[0137] refer to Figure 3 , Figure 3 The three sets of waveforms on the left side are used to illustrate EEG, EMG, and ECG signals, respectively. After the pre-defined model processes the first, second, and third coupling features of the EEG, EMG, and ECG signals, it can output the motor cognitive assessment results. Specifically, the first branch of the model outputs the Sigmoid-normalized value, which is directly used as the motor cognitive efficiency index (range [0,1]). For the motor cognitive efficiency index, 0.8~1.0 is "excellent" (efficient use of cognitive resources), 0.6-0.8 is "good" (reasonable allocation of cognitive resources), 0.4~0.6 is "moderate" (basically effective cognitive regulation), and 0~0.4 is "poor" (waste of cognitive resources or failure of regulation). For example, the second branch of the model outputs the probability distribution of 5 categories, and selects the category with the highest probability as the cognitive load level (level 1 "very low load" to level 5 "very high load"). Based on this, the level judgment threshold can be fine-tuned by combining the test set verification results (e.g., the probability threshold for level 3 load is set to 0.3 to ensure the accuracy of the classification). It should be noted that... Figure 3 The image has been blurred and does not show specific details of the overall interface. It should be understood that this disclosure does not limit this.

[0138] Furthermore, a data transmission interface can be established to receive preprocessed real-time coupling features (updated every 300ms, synchronized with signal acquisition) and call a pre-trained model to calculate the evaluation results in real time. Additionally, the coupling features and motion cognition evaluation results for the most recent 10 minutes can be cached, supporting time-series backtracking viewing.

[0139] Based on this, a line chart can be used to display the dynamic changes of core indicators (phase synchronization index, cross-correlation coefficient, LF / HF ratio), with the horizontal axis representing time (min:ss) and the vertical axis representing the normalized value of the indicator ([0,1]). Furthermore, interactive functions can be designed. For example, hovering the mouse can display the precise value, timestamp, and corresponding task stage; single indicator hiding / showing is supported, facilitating focus on key features.

[0140] In addition, a heatmap can be generated based on the motor cognitive assessment results and updated in real time. Specifically, the heatmap data can be refreshed every 30 seconds, highlighting the cells corresponding to the current task difficulty-cognitive load combination to intuitively present the assessment status.

[0141] Furthermore, it can provide time filtering controls (by date, test period), support the selection of multiple test periods (up to 5), and generate a trend chart of the motor cognitive efficiency index and a bar chart of the cognitive load level distribution.

[0142] Supports exporting comparison reports (e.g., portable document format), including trend analysis conclusions (such as "cognitive efficiency index increased by 0.15 in the past week, cognitive load level stabilized at level 2, and motor cognitive function improved").

[0143] Furthermore, it enables multi-dimensional parameter linkage queries. For example, filtering conditions (task type, cognitive load level, efficiency index range) can be designed, and users can accurately locate the coupling characteristic curve, evaluation results, and original signal segments for the corresponding scenario after selecting them. Additionally, it supports data export, allowing users to export filtered data (comma-separated value format) for subsequent scientific research analysis or adjustments to rehabilitation training programs.

[0144] Figure 4 A schematic diagram of a motor cognition assessment system according to another embodiment of the present disclosure is shown.

[0145] like Figure 4 As shown, in the motor cognitive assessment system, the multi-source electrophysiological signal synchronous acquisition module can be electrically connected to the human-computer interaction module to display the acquired electroencephalogram (EEG), electromyogram (EMG), and electrocardiogram (ECG) signals via the human-computer interface. Furthermore, the human-computer interface can also be electrically connected to the processor to display data from the processor; this will not be elaborated upon here. It should be noted that in... Figure 3 In this embodiment, the electrocardiogram (ECG) signal is collected from the left chest position. However, it should be understood that this is only for illustration. In other embodiments, the ECG signal may also be collected from the right chest position. The collection position is related to the location of the target object's heart, which will not be elaborated here.

[0146] Building upon this foundation, this publication focuses on the interdisciplinary field of motor cognitive assessment technology and electrophysiological information processing, proposing a motor cognitive assessment system based on the synergistic coupling analysis of multimodal electrophysiological signals, including electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG). This system achieves high-precision quantitative assessment of motor cognitive function by constructing a feature fusion model and dynamic assessment framework for multi-source electrophysiological signals. In practical applications, this technology can be effectively applied to cutting-edge research areas such as dynamic monitoring of athletes' specific training effects, scientific assessment of the rehabilitation process of patients with neurocognitive impairments, and real-time recognition of motor intentions in human-computer interaction modules, demonstrating significant theoretical value and application potential.

[0147] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0148] Those skilled in the art will understand that the features described in the various embodiments of this disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments of this disclosure can be combined and / or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0149] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A motion cognitive assessment system based on multi-source electro-physiological information coupling, characterized in that, The method comprises the following steps: a multi-source electrophysiological signal synchronous acquisition module is used to acquire electroencephalogram (EEG) signals, electromyogram (EMG) signals and electrocardiogram (ECG) signals of a target object in the same period during the process in which the target object performs a plurality of motion cognitive tasks of different difficulties according to a motion instruction; a processor is electrically connected to the multi-source electrophysiological signal synchronous acquisition module and is used to: extract first coupling features of the EEG signals and the EMG signals, second coupling features of the EEG signals and the ECG signals, and third coupling features of the ECG signals and the EMG signals; and process the first coupling features, the second coupling features and the third coupling features by using a predetermined model to obtain a motion cognitive evaluation result of the target object; wherein the first coupling features represent the synchronization degree of the motion instructions generated by the brain regions of the target object according to the motion instruction and the response conditions of the muscle groups of the target object to the motion instructions; the second coupling features represent the influence degree of the cognitive regulation of the target object according to the difficulty changes of the motion cognitive tasks on the heart rate of the target object; and the third coupling features represent the matching degree between the physiological state of the target object and the energy consumed by the muscle groups.

2. The motion cognitive assessment system of claim 1, wherein, The EEG signals comprise EEG signals of a plurality of brain regions of the target object; and the EMG signals comprise EMG signals of a plurality of muscle groups of the target object. The first coupling features represent the synchronization degree of the motion instructions of the plurality of brain regions and the response conditions of the corresponding muscle groups in the plurality of muscle groups.

3. The motion cognitive assessment system of claim 2, wherein, The predetermined model comprises a convolutional neural network. The processor is further used to: process the first coupling features, the second coupling features and the third coupling features by using the convolutional neural network to obtain convolutional coupling features; wherein the convolutional coupling features are used to represent the difference conditions between the response conditions corresponding to different brain regions; further process the convolutional coupling features by using the predetermined model to obtain the motion cognitive evaluation result.

4. The motion cognitive assessment system of claim 3, wherein, The processor is further used to: perform dimension reduction on the first coupling features, the second coupling features and the third coupling features respectively by using a principal component analysis algorithm; process the dimension-reduced first coupling features, the dimension-reduced second coupling features and the dimension-reduced third coupling features by using the convolutional neural network to obtain the convolutional coupling features.

5. The motion cognitive assessment system of claim 3, wherein, The multi-source electrophysiological signal synchronous acquisition module is further used to acquire N groups of electrophysiological signals of the target object in N periods; wherein each group of the electrophysiological signals comprises EEG signals, EMG signals and ECG signals in the same period; the predetermined model further comprises a long short-term memory network and an attention mechanism layer; and N is an integer greater than 1. The processor is further used to: process the convolutional coupling features by using the attention mechanism layer to obtain attention coupling features; process the attention coupling features in the N periods by using the long short-term memory network to obtain a spatiotemporal fusion feature; and the spatiotemporal fusion feature represents the change conditions of the convolutional coupling features with the difficulty changes of the motion cognitive tasks. The output layer processes the spatio-temporal fusion features to obtain the motion cognitive evaluation result.

6. The motion cognitive assessment system of claim 5, wherein, The multi-source electrophysiological signal synchronous acquisition module comprises: an electroencephalogram acquisition unit, an electromyogram acquisition unit, an electrocardiogram acquisition unit, and a synchronous triggering unit electrically connected to the processor, the electroencephalogram acquisition unit, the electromyogram acquisition unit, and the electrocardiogram acquisition unit, and configured to: under the control of a sampling start instruction from the processor, send a triggering signal to the electroencephalogram acquisition unit, the electromyogram acquisition unit, and the electrocardiogram acquisition unit simultaneously, so that the electroencephalogram acquisition unit, the electromyogram acquisition unit, and the electrocardiogram acquisition unit sample simultaneously to obtain the electroencephalogram signal, the electromyogram signal, and the electrocardiogram signal in the same time period.

7. The motion cognitive assessment system of claim 6, wherein, The multi-source electrophysiological signal synchronous acquisition module further comprises a time stamp marking circuit electrically connected to the electroencephalogram acquisition unit, the electromyogram acquisition unit, and the electrocardiogram acquisition unit, and configured to: receive the electroencephalogram signal, the electromyogram signal, and the electrocardiogram signal at each time point from the electroencephalogram acquisition unit, the electromyogram acquisition unit, and the electrocardiogram acquisition unit, and set a corresponding time stamp for the electroencephalogram signal, the electromyogram signal, and the electrocardiogram signal at each time point according to a clock signal of the processor.

8. The motion cognitive assessment system of claim 7, wherein, The processor is electrically connected to the time stamp marking circuit and is further configured to: receive the electroencephalogram signal, the electromyogram signal, and the electrocardiogram signal from the time stamp marking circuit; based on the time stamps of the received electroencephalogram signal, electromyogram signal, and electrocardiogram signal, resample and align the electroencephalogram signal, electromyogram signal, and electrocardiogram signal, and fill in the sampling gaps of the electroencephalogram signal, electromyogram signal, and electrocardiogram signal due to resampling and alignment, to time-synchronously synchronize the electroencephalogram signal, electromyogram signal, and electrocardiogram signal while retaining the change trend of the electroencephalogram signal, electromyogram signal, and electrocardiogram signal, to obtain the electroencephalogram signal, electromyogram signal, and electrocardiogram signal in the same time period.

9. The motion cognitive assessment system of any one of claims 1-8, wherein, The processor is further configured to: extract a beta wave of the target object from the electroencephalogram signal, calculate a phase synchronization index of the beta wave and the electromyogram signal, and determine the phase synchronization index as the first coupling feature; determine a heart rate variability signal of the target object according to the electrocardiogram signal, extract a high-frequency power of the heart rate variability signal, extract a gamma wave power of the target object from the electroencephalogram signal, and determine the second coupling feature according to the degree of influence of the gamma wave power on the high-frequency power; extract a low-frequency power of the heart rate variability signal, extract an average power frequency of the electromyogram signal, calculate a similarity distance of the low-frequency power and the average power frequency, and determine the similarity distance as the third coupling feature.

10. The motion cognitive assessment system of any one of claims 1-8, wherein, The motion cognitive evaluation result comprises a motion cognitive efficiency index and a cognitive load level. The motion cognitive efficiency index represents the influence efficiency of the cognitive of the target object on the muscle group to the motion cognitive tasks of the multiple difficulties.

Images