Neurofeedback reaction time assessment and cognitive training method and system

Abstract

This invention relates to the field of data processing technology and discloses a method and system for evaluating neural feedback reaction time and cognitive training. The method includes: initializing a set of basic microstate parameters; acquiring pairs of long-range stationary and short-range impulsive features; implementing two-dimensional nonlinear smoothing coefficient modulation; updating the moving average center estimate in conjunction with the center estimate and the smoothing coefficient; and constructing a covariance threshold for variance and mean based on the center estimate and the smoothing coefficient. Compared to existing technologies, in neural feedback training scenarios requiring rapid tracking of cognitive state transitions, traditional change point detection algorithms suffer from excessive detection lag due to fixed time delays during accelerated state transitions, especially under conditions of accelerated microstate transitions, failing to achieve constant low-latency online detection. This application achieves constant low-latency online microstate detection through bimodal feature extraction and adaptive modulation, improving the timeliness and reliability of neural feedback intervention.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for assessing neurofeedback reaction time and cognitive training. Background Technology

[0002] Currently, in rapid neurofeedback cognitive training scenarios, it is necessary to monitor the subject's reaction time sequence in real time and trigger neural modulation intervention based on micro-state transitions. Existing technologies typically employ online change point detection algorithms to continuously monitor the mean or variance drift of the reaction time sequence, such as exponentially weighted moving average control charts based on a fixed forgetting factor, Bayesian online change point detection, or sliding statistical methods with a fixed window length. These algorithms calculate detection statistics using a preset global smoothing coefficient or a fixed-width sliding window and compare them with a constant threshold. Once the threshold is exceeded, a state transition is determined, and a stimulus command is issued, attempting to capture the change point at the instant of cognitive state transition. However, in rapid neurofeedback closed loops, the intervention window is often only a few hundred milliseconds, making it extremely sensitive to detection delays.

[0003] For example, during the micro-state transition from high concentration to sudden distraction, reaction time sequences often exhibit a sharp increase in mean accompanied by a dramatic amplification of variance. Fixed forgetting factor change point detection algorithms, due to their constant smoothing coefficient, do not adequately increase the weight of recently observed values ​​when the state transition speed accelerates, resulting in a response lag and a delay of several seconds between the change point and the trigger signal. Bayesian online methods require maintaining the run-length probability distribution; the computational complexity and memory usage per step increase linearly with the sequence length, making it difficult to provide a constant low-latency response. This detection lag far exceeds the optimal neural intervention window, causing interventions such as transcranial electrical stimulation or auditory feedback to miss long-term enhancement opportunities, and even reducing training effectiveness due to feedback misalignment. Furthermore, fixed control limits cannot adapt to non-stationary drift of the reaction time baseline; gradual increases in reaction time caused by subject fatigue or habituation are prone to triggering false alarms and interrupting the training process.

[0004] Therefore, there is an urgent need for an online change point detection scheme that can still provide constant computational complexity and controllable latency even when the speed of cognitive state switching changes dynamically and the reaction time baseline is not stationary. This would allow the system to complete change point confirmation and issue a trigger signal within the minimum delay budget after the micro-state transition of the reaction time sequence, thereby ensuring that neural feedback intervention accurately locks the optimal time window and improves the timeliness of neural modulation and the reliability of cognitive training. Summary of the Invention

[0005] To address the aforementioned technical shortcomings, the purpose of this invention is to propose a neurofeedback reaction time assessment and cognitive training method. This method aims to solve the technical problem in existing technologies where, in neurofeedback training scenarios requiring rapid follow-up on cognitive state switching, traditional change point detection algorithms suffer from excessive detection lag due to fixed time delays during accelerated state transitions. This lag is particularly pronounced under conditions where micro-state switching accelerates and delays exceed the optimal intervention window, making it impossible to achieve constant low-latency online detection.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a method for assessing neurofeedback reaction time and cognitive training.

[0007] The aforementioned neurofeedback reaction time assessment and cognitive training method includes: Step S10: Obtain the initial configuration parameter set formed by the reaction time acquisition unit and the EEG synchronous acquisition unit, perform the parameter initialization task based on the initial configuration parameter set, and output the microstate tracking benchmark parameter set; Step S20: Obtain the current trial reaction time observation value, and perform the fluctuation feature extraction task based on the microstate tracking benchmark parameter set and the current trial reaction time observation value using the bistate coefficient of variation decomposition method, and output the bistate coefficient of variation pair consisting of long-range stationary coefficient of variation and short-range impact coefficient of variation; Step S30: Based on the dual-state variation coefficient pair, a two-dimensional nonlinear smoothing coefficient modulation method is used to perform a real-time adjustment task of tracking agility, and output the dynamic smoothing coefficient at the current moment; Step S40: Based on the observed value of the current reaction time and the dynamic smoothing coefficient at the current moment, perform the real-time tracking task of the micro-state center using a time-weighted update method, and output the updated reaction time series index weighted value; Step S50: Construct a covariance threshold for variance and mean based on the updated reaction time index weighted value, the current trial reaction time observation value, and the current time dynamic smoothing coefficient to obtain an adaptive detection threshold bound; determine whether a micro-state switch has occurred in the current trial based on the adaptive detection threshold bound, and output the current time micro-state switch detection result and neural feedback trigger control command.

[0008] Preferably, step S10 specifically includes: Step S101: Obtain the calibration reaction time sample set, delay budget parameter, and threshold bias parameter, wherein the calibration reaction time sample set includes multiple benchmark reaction time observations continuously collected before training begins, and the delay budget parameter includes the minimum delay budget. and maximum allowed detection delay The threshold bias parameter includes an adaptive threshold baseline offset coefficient. ; Step S102: Set the basic smoothing coefficient Global coefficient of variation smoothing factor and local variation coefficient smoothing factor and make ; Step S103: Determine the initial center estimate based on the calibration reaction time sample set. and initial local variance estimate and initialize the global coefficient of variation. and initial value of local coefficient of variation Set all parameters to 0 to output the microstate tracking reference parameter set.

[0009] Preferably, step S20 specifically includes: Step S201: Obtain the observed reaction time of the current trial in the i-th trial. ,when When calculating the reaction time difference between adjacent trials. ,in, For the first Observed values ​​of reaction time in each trial; Step S202: Based on the observed values ​​of the current reaction time. The previous time-center estimate and the global coefficient of variation at the previous time step Update the global coefficient of variation used to reflect long-range stationarity. The calculation formula is:

[0010] in, The global coefficient of variation at the current moment. To prevent positive numbers with a denominator of zero; Step S203: Based on the reaction time difference between adjacent trials The estimated center value at the previous time step Local variation coefficient at the previous moment Update the local coefficient of variation used to reflect short-range impact. The calculation formula is:

[0011] The final output is from and The two-state coefficient of variation pairs.

[0012] Preferably, in step S202, the global coefficient of variation smoothing factor The value is less than the local variation coefficient smoothing factor in step S203. .

[0013] Preferably, step S30 specifically includes: Step S301: Based on the global coefficient of variation and the local coefficient of variation Constructing unlimited smoothing coefficients The calculation formula is:

[0014] in, The current unlimited smoothing coefficient. As a long-range stability penalty factor, It is a short-range impact enhancement factor; Step S302: Based on the minimum delay budget and the maximum allowable detection delay Determine the lower limit of the smoothing coefficient With upper limit of smoothing coefficient The calculation formula is:

[0015] in, Used to limit slow updates to the microstate center Used to limit the microstate center from updating too quickly; Step S303: Adjust the unlimited smoothing coefficient Limited to the lower limit of the smoothing coefficient and the upper limit of the smoothing coefficient Within the closed interval formed, the dynamic smoothing coefficient at the current time is obtained. The calculation formula is: .

[0016] Preferably, step S40 specifically includes: Step S401: Obtain the observed value of the reaction time for the current trial. Dynamic smoothing coefficient at the current moment and the center estimate of the previous time step ; Step S402: According to the dynamic smoothing coefficient at the current time Observed values ​​for the current trial reaction time and the center estimate of the previous time step Perform a time-weighted update and output the updated reaction time series index weighted value. , .

[0017] Preferably, step S50 specifically includes: Step S501: Obtain the adaptive threshold baseline offset coefficient The updated reaction time series index weighted value The current dynamic smoothing coefficient The observed value of the current reaction time. Local variance estimate at the previous time step ; Step S502: Based on the dynamic smoothing coefficient at the current time The local variance estimates are updated using an exponential weighting method, calculated as follows:

[0018] in, This is the local variance estimate at the current time. Step S503: Based on the updated reaction time series index weighted value The estimated local variance at the current time. and the adaptive threshold reference offset coefficient The adaptive detection threshold bound for the current time step is constructed using the following formula:

[0019] in, The upper control limit is set at the current moment. The lower control limit at the current moment, The output is the local standard deviation estimate at the current time. and The system is composed of an adaptive detection threshold bound for the current moment; finally, it determines whether a micro-state switch has occurred in the current trial based on the adaptive detection threshold bound, and outputs the micro-state switch detection result and neural feedback trigger control command for the current moment.

[0020] The present invention also provides a neurofeedback reaction time assessment and cognitive training system, comprising: The parameter initialization module is used to obtain the initial configuration parameter set formed by the reaction time acquisition unit and the EEG synchronous acquisition unit, perform the parameter initialization task based on the initial configuration parameter set, and output the microstate tracking benchmark parameter set. The fluctuation feature extraction module is used to obtain the current trial reaction time observation value. Based on the microstate tracking benchmark parameter set and the current trial reaction time observation value, the fluctuation feature extraction task is performed using the bistate coefficient of variation decomposition method, and the output is a bistate coefficient of variation pair consisting of long-range stationary coefficient of variation and short-range impact coefficient of variation. The smoothing coefficient modulation module is used to perform a real-time adjustment task of tracking agility based on the dual-state variation coefficient pair using a two-dimensional nonlinear smoothing coefficient modulation method, and outputs the dynamic smoothing coefficient at the current moment. The microstate center tracking module is used to perform real-time microstate center tracking tasks based on the observed reaction time of the current trial and the dynamic smoothing coefficient at the current moment using a time-weighted update method, and output the updated reaction time series index weighted value. The threshold boundary generation module is used to construct an adaptive detection threshold boundary by performing variance and mean covariance threshold based on the updated reaction time index weighted value, the current trial reaction time observation value, and the current time dynamic smoothing coefficient; it determines whether a micro-state switch of reaction time has occurred in the current trial based on the adaptive detection threshold boundary, and outputs the micro-state switch detection result and neural feedback trigger control command at the current time.

[0021] The present invention also provides a neurofeedback reaction time assessment and cognitive training device, the neurofeedback reaction time assessment and cognitive training device comprising: a memory, a processor, and a neurofeedback reaction time assessment and cognitive training program stored in the memory and executable on the processor, wherein the neurofeedback reaction time assessment and cognitive training program implements the above method when executed by the processor.

[0022] The present invention also provides a computer program product, the computer program product including a neurofeedback reaction time assessment and cognitive training program, which implements the above method when executed by a processor.

[0023] The beneficial effects of this invention are as follows: This invention achieves feature separation and extraction of long-range stationary drift and short-range instantaneous impact in reaction time series by separately initializing the forgetting factors of global variation and local shock, and independently updating the global variation coefficient and the first-order difference local variation coefficient. This enables the micro-state tracking system to simultaneously capture the slow evolution and instantaneous switching of cognitive states, providing accurate dual-modal fluctuation feature input for subsequent dynamic agility adjustment.

[0024] Based on a two-dimensional nonlinear smoothing coefficient modulation function, the aforementioned bimodal variability coefficient is transformed into a real-time dynamic smoothing coefficient. The smoothing coefficient is automatically reduced according to the local impact intensity to improve tracking agility. This drives the center estimate to converge rapidly to the new state center within the minimum delay budget after a cognitive micro-state transition, avoiding the tracking lag caused by traditional fixed smoothing coefficients during accelerated state transitions and ensuring that neural feedback intervention can be triggered within the effective time window. Simultaneously, by constructing a covariant threshold using variance and mean, the adaptively generated detection threshold bound can be adjusted synchronously with local variability. This prevents false alarms after state switching using fixed thresholds and maintains high detection sensitivity for small drifts during the stable period, thus improving the overall reliability of online micro-state detection. Attached Figure Description

[0025] Figure 1This is a flowchart illustrating the first embodiment of a neurofeedback reaction time assessment and cognitive training method according to the present invention. Detailed Implementation

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0027] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Example 1: As Figure 1 The diagram shown is a flowchart of a first embodiment of a neurofeedback reaction time assessment and cognitive training method according to the present invention, which presents a first embodiment of a neurofeedback reaction time assessment and cognitive training method according to the present invention.

[0029] In the first embodiment, the neurofeedback reaction time assessment and cognitive training method includes: Step S10: Obtain the initial configuration parameter set, perform the parameter initialization task based on the initial configuration parameter set, and output the micro-state tracking benchmark parameter set; The initial configuration parameter set obtained in step S10 includes the basic smoothing coefficient. Minimum delay budget, maximum allowable detection delay, adaptive threshold base offset And a global coefficient of variation smoothing factor for bimodal coefficient of variation decomposition. and first-order difference local variation coefficient smoothing factor . The default exponentially weighted moving average memory length is determined, typically between 0.1 and 0.2; Dmin and Dmax define the lower and upper limits of the detection latency, respectively, based on the number of trials, and are used to verify whether the detection timing falls within the allowable range in subsequent processes. The scaling factor used to construct adaptive control limits is generally set around 2.5 to 3.0. Smaller values, such as 0.01 to 0.03, are assigned to perform slow smoothing on the global coefficient of variation to characterize the steady fluctuations of the reaction time series over a longer time scale. A larger value, such as 0.1 to 0.2, is used to quickly respond to changes in the first-order difference between adjacent trials and capture the instantaneous impact of micro-state switching.

[0030] By employing a dual forgetting factor of global variation and local shock for initialization, the system possesses dual perception capabilities from the outset. The slower global forgetting rate ensures that subsequent updates to the global variation coefficient are largely unaffected by occasional fluctuations in individual trials, robustly reflecting the overall stability of the cognitive state. The faster local forgetting rate makes the first-order difference local variation coefficient highly sensitive to small jumps between adjacent trials. This differentiation of initial parameters means that the basic parameter set for microstate tracking is not a set of undifferentiated steady-state initial values, but rather pre-embeds the differentiated dynamic response characteristics necessary for subsequent bimodal decomposition. Based on this, step S20 does not require tentative separation of the two variation components at startup; it can directly and continuously update according to their respective smooth paths, avoiding confusion of fluctuation characteristics caused by initial parameter homogenization. Meanwhile, the upper and lower limits of the delay budget are pre-written into the parameter set as hard constraints, so that no matter how the subsequent dynamic smoothing coefficient changes, the detection timing of micro-state switching is always constrained within a known and engineerable range. This provides a time boundary for subsequent steps to limit the detection lag to within the maximum allowable delay. The temporal behavior of the entire tracking process is built into a framework that is conducive to the timeliness of neural feedback intervention from the beginning.

[0031] In traditional reaction time-varying point detection initialization, only a single smoothing coefficient or a fixed window length is often set. The resulting mean and variance estimates can only operate on the same time scale from the outset. If the smoothing strength is set too high to ensure long-term stability, it becomes difficult to react quickly to sudden state transitions in the initial stage, increasing detection latency. Conversely, if a shorter window or a smaller forgetting factor is used to pursue faster response, initial spurious fluctuations are easily amplified into false changes. Especially at the beginning of neurofeedback training, when subjects are in the task adaptation period and reaction times are often accompanied by a warm-up effect, this trade-off caused by single-scale parameter settings can lead to either severe lag in microstate detection during the initial training phase or frequent false alarms, thus missing the opportunity for early intervention. This invention explicitly specifies and stores two sets of forgetting factors during initialization, eliminating the detection bias caused by single-scale initialization from the source. This ensures that the system, whether in the initial stage or during operation, has the ability to simultaneously observe long-range drift and short-range shocks. This allows all subsequent dynamic adjustments to be based on matching the actual structure of cognitive fluctuations, helping to reduce the detection latency of microstate transitions to near the minimum budget.

[0032] Step S20: Based on the microstate tracking reference parameter set and the current trial reaction time observation, the fluctuation feature extraction task is performed using the bistate coefficient of variation decomposition method, and the bistate coefficient of variation pair is output, consisting of long-range stationary coefficient of variation and short-range impact coefficient of variation. After bimodal coefficient of variation decomposition, the microstate tracker no longer relies solely on a single, general fluctuation metric, but instead possesses two decoupled variance measures. During periods of stable cognitive operation, reaction time fluctuates randomly around the center, with small differences between adjacent trials, and both coefficients of variation remain low. This allows the system to maintain a high level of smoothness and avoid over-responding to noise. When cognitive state undergoes a slow drift, such as a gradual increase in reaction time due to fatigue, the global coefficient of variation slowly accumulates and rises, while the local coefficient of variation remains low because there are no drastic successive jumps. This differentiated response allows the system to distinguish between slow trends and sudden shocks. Once a microstate switch occurs, such as a step increase in reaction time after inattention, the sudden increase in the first-order difference causes a momentary spike in the local coefficient of variation, while the global coefficient of variation, although lagging, also shows an upward trend. This clear separation provides rich operating condition information for the two-dimensional modulation function in step S30, enabling the dynamic smoothing coefficient to be adjusted as needed according to the true cause of the fluctuation. This ensures the efficiency of micro-state switching detection while suppressing false triggering caused by confusion drift and impact, and allows the number of tests required for detection to be constrained within the delay budget.

[0033] Existing online change point detection schemes often use metrics such as the standard deviation within a fixed window or variance estimation using a single forgetting factor to measure the degree of fluctuation. These metrics cannot decompose the sources of variation in observed values ​​into background fluctuations and instantaneous shocks. When slow cognitive state drift coexists with occasional operational errors, a single variation metric can exhibit ambiguous fluctuations, making it difficult for the detection algorithm to correctly determine whether a switching signal should be issued. If a low threshold is set to capture shocks, normal fluctuations during the drift period are easily misjudged as change points; if the threshold is relaxed to suppress false alarms, genuine state transitions may be missed, resulting in detection lag exceeding the allowable range. This invention updates the coefficients of variation of two different time constants simultaneously in each iteration, which is equivalent to performing dual-channel characteristic separation on the reaction time series, distinguishing between low-frequency neural trends and high-frequency cognitive shocks. This allows the system to be free from mixed fluctuations when generating subsequent thresholds, enabling it to more reliably characterize short-range shocks as micro-state transitions that require rapid follow-up, and long-range drifts as background evolutions that require moderate tracking but do not require immediate alarms, thus providing feature-level separation assurance for constant low-latency detection.

[0034] Step S30: Based on the dual-state variation coefficient pair, a two-dimensional nonlinear smoothing coefficient modulation method is used to perform a real-time adjustment task of tracking agility, and output the dynamic smoothing coefficient at the current moment; In traditional change point detection or reaction time baseline tracking, the smoothing coefficient is usually a fixed constant, or it is simply switched in stages based on a single-dimensional criterion. For example, temporarily increasing the smoothing coefficient after several consecutive trials exceeding a fixed threshold results in a response lag, and it cannot distinguish between background drift and true state transitions. It is prone to premature acceleration during drift, leading to violent oscillations of the tracking center and subsequent instability in detection. The two-dimensional modulation mechanism proposed in this invention actually constructs a nonlinear cost function for the cause of fluctuations: long-range stationarity deterioration only requires an exponential decay cost, and the smoothing coefficient decreases moderately, ensuring smooth following of slow trends such as fatigue; short-range shocks produce near-saturated gains, resulting in a sharp increase in the smoothing coefficient, and the stronger the shock, the more decisive the increase. This design allows the adjustment of the tracking speed to precisely match the nature and intensity of cognitive fluctuations, ensuring the shortest delay detection while avoiding noise amplification problems caused by unnecessary high-agility operation. Thus, without increasing external information, it improves the timeliness and robustness of micro-state transition detection in neural feedback scenarios.

[0035] Step S40: Based on the observed value of the current reaction time and the dynamic smoothing coefficient at the current moment, perform the real-time tracking task of the micro-state center using a time-weighted update method, and output the updated reaction time series index weighted value; In traditional EWMA baseline tracking based on a fixed smoothing coefficient, a small smoothing coefficient requires a longer average number of trials to adequately respond to step changes, causing detection delays to exceed the optimal intervention window. Conversely, a large smoothing coefficient, while accelerating the response, makes the estimated value susceptible to excessive manipulation by single outliers, leading to frequent baseline fluctuations and consequently frequent false breakpoint alarms. This contradiction is particularly pronounced in neurofeedback applications, where the subject's reaction time sequence itself exhibits high inter-trial variability. This step directly benefits from the dynamic data provided in step S30. This allows the center estimate to intelligently follow the bimodal variation characteristics: remaining robust during stable periods and accelerating decisively during shocks, thus eliminating the need for manually pre-setting stage parameters and trade-offs between response speed and estimation variance. The center estimate thus becomes a micro-state indicator signal that combines robustness and agility, providing a reliable localization reference for subsequent adaptive threshold bounds and helping to reliably limit detection latency within a preset maximum allowable range during accelerated state transitions.

[0036] Step S50: Construct a variance and mean covariance threshold based on the updated reaction time index weighted value, the current trial reaction time observation value, and the current time dynamic smoothing coefficient, and determine whether a microstate switch has occurred in the current trial based on the constructed adaptive detection threshold boundary, and output the current time microstate switch detection result and neural feedback trigger control command.

[0037] Because the local variance estimate and the central estimate share the same dynamic smoothing coefficient, the covariance mechanism of variance and mean is achieved. When a cognitive microstate switches, the central estimate moves rapidly, and the local variance also changes accordingly. The rapid expansion due to the increase in size causes the threshold bound to widen automatically, adapting to the potentially higher variability of the new state and thus avoiding misreporting normal fluctuations in the new state as another switch. Once the state stabilizes again, the central estimate tends to plateau, and the local variance decreases accordingly. As the threshold gradually narrows due to the decline in amplitude, the threshold boundary tightens synchronously, restoring high sensitivity to subtle drifts. Thus, the system achieves a dynamic balance between detection sensitivity and false alarm suppression, eliminating the need for manual setting of state-related threshold parameters. Simultaneously, because variance and center updates are driven synchronously, the trigger signal for micro-state switching is generated promptly when the center has just stabilized to the new level and the variance has reflected the new characteristics. Detection latency is reliably limited to within the maximum budget, meeting the timing requirements of the fast neural feedback loop.

[0038] Traditional control charts or threshold detection methods often set fixed upper and lower control limits based on historical training data. These fixed limits cannot be adjusted as variance evolves during online training. In neurofeedback training, the standard deviation of reaction time can vary significantly under different cognitive states. Fixed limits either lead to numerous false alarms when volatility increases, interfering with training immersion, or they are insufficient in detection when volatility decreases, missing true state transitions and causing excessive detection delays. This invention integrates the updating of variance estimation with the dynamics of center estimation. By binding the threshold together and recalculating it at each moment based on the latest estimated center and standard deviation, the detection boundary becomes an instantaneous function of the cognitive state. This design allows the threshold boundary to adjust synchronously with the state change when the subject's cognitive state transitions accelerate. It avoids misjudging new high-variability responses as new change points by continuously using the low-variability threshold of the old state, and also avoids missing subsequent micro-state transitions due to an overly wide threshold. This keeps the detection latency within a range that matches the maximum allowable latency, improving the timeliness and reliability of online detection.

[0039] Example 2: Furthermore, the present invention provides a neurofeedback reaction time assessment and cognitive training system, employing a neurofeedback reaction time assessment and cognitive training method from the above embodiments, which can solve a technical problem related to neurofeedback reaction time assessment and cognitive training. The beneficial effects of the neurofeedback reaction time assessment and cognitive training system provided by the present invention are the same as those of the neurofeedback reaction time assessment and cognitive training method provided in the above embodiments, and other technical features of the neurofeedback reaction time assessment and cognitive training system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0040] Example 3: This invention provides a neurofeedback reaction time assessment and cognitive training device. The device includes at least one processor and a memory communicatively connected to the processor. The memory stores instructions executable by the processor, which are then executed to enable the processor to perform a neurofeedback reaction time assessment and cognitive training method as described in Example 1. The neurofeedback reaction time assessment and cognitive training device in this invention may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. This neurofeedback reaction time assessment and cognitive training device is merely an example and should not limit the functionality or scope of the invention. The device may also include a processing unit (e.g., a central processing unit, a graphics processing unit), which can perform various appropriate actions and processes based on a program stored in a read-only memory or a program loaded from a storage device into a random access memory. The random access memory also stores various programs and data required for the operation of a neurofeedback reaction time assessment and cognitive training device. The processing unit, read-only memory, and random access memory are interconnected via a bus. An I / O interface is also connected to the bus. Typically, the following systems can be connected to the I / O interface: input devices including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices including liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices including magnetic tapes, hard disks, etc.; and communication devices. The communication device allows a neurofeedback reaction time assessment and cognitive training device to communicate wirelessly or wiredly with other devices to exchange data. While a neurofeedback reaction time assessment and cognitive training device with various systems has been described, it should be understood that it is not required to implement or possess all the systems described. Alternatively, more or fewer systems may be implemented.

[0041] Example 4: This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the neurofeedback reaction time assessment and cognitive training method described above. The computer program product provided by this invention can solve a technical problem related to neurofeedback reaction time assessment and cognitive training. Compared with the prior art, the beneficial effects of the computer program product provided by this invention are the same as those of the neurofeedback reaction time assessment and cognitive training method provided in the above embodiments, and will not be repeated here.

[0042] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a read-only memory. When the computer program is executed by a processing device, it performs the functions defined in the methods of the embodiments disclosed in this invention.

[0043] It should be understood that the various parts disclosed in this invention can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0044] 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 present invention and its equivalents, the present invention also intends to include these modifications and variations.

Claims

1. A method for assessing neurofeedback reaction time and cognitive training, characterized in that, A method for applying neurofeedback cognitive training devices, including reaction time acquisition units and EEG synchronization acquisition units, includes: Step S10: Obtain the initial configuration parameter set formed by the reaction time acquisition unit and the EEG synchronous acquisition unit, perform the parameter initialization task based on the initial configuration parameter set, and output the microstate tracking benchmark parameter set; Step S20: Obtain the current trial reaction time observation value, and perform the fluctuation feature extraction task based on the microstate tracking benchmark parameter set and the current trial reaction time observation value using the bistate coefficient of variation decomposition method, and output the bistate coefficient of variation pair consisting of long-range stationary coefficient of variation and short-range impact coefficient of variation; Step S30: Based on the dual-state variation coefficient pair, a two-dimensional nonlinear smoothing coefficient modulation method is used to perform a real-time adjustment task of tracking agility, and output the dynamic smoothing coefficient at the current moment; Step S40: Based on the observed value of the current reaction time and the dynamic smoothing coefficient at the current moment, perform the real-time tracking task of the micro-state center using a time-weighted update method, and output the updated reaction time series index weighted value; Step S50: Construct a covariance threshold for variance and mean based on the updated reaction time index weighted value, the current trial reaction time observation value, and the current time dynamic smoothing coefficient to obtain an adaptive detection threshold bound; determine whether a micro-state switch has occurred in the current trial based on the adaptive detection threshold bound, and output the current time micro-state switch detection result and neural feedback trigger control command.

2. The neurofeedback reaction time assessment and cognitive training method as described in claim 1, characterized in that, Step S10 specifically includes: Step S101: Obtain the calibration reaction time sample set, delay budget parameter, and threshold bias parameter, wherein the calibration reaction time sample set includes multiple benchmark reaction time observations continuously collected before training begins, and the delay budget parameter includes the minimum delay budget. and maximum allowed detection delay The threshold bias parameter includes an adaptive threshold baseline offset coefficient. ; Step S102: Set the basic smoothing coefficient Global coefficient of variation smoothing factor and local variation coefficient smoothing factor and make ; Step S103: Determine the initial center estimate based on the calibration reaction time sample set. and initial local variance estimate and initialize the global coefficient of variation. and initial values ​​of local coefficient of variation Set all parameters to 0 to output the microstate tracking baseline parameter set.

3. The neurofeedback reaction time assessment and cognitive training method as described in claim 2, characterized in that, Step S20 specifically includes: Step S201: Obtain the observed reaction time of the current trial in the i-th trial. ,when When calculating the reaction time difference between adjacent trials. ,in, For the first Observed values ​​of reaction time in each trial; Step S202: Based on the observed values ​​of the current reaction time. The previous time-center estimate and the global coefficient of variation at the previous time step Update the global coefficient of variation used to reflect long-range stationarity. The calculation formula is: in, The global coefficient of variation at the current moment. To prevent positive numbers with a denominator of zero; Step S203: Based on the reaction time difference between adjacent trials The estimated center value at the previous time step Local variation coefficient at the previous moment Update the local coefficient of variation used to reflect short-range impact. The calculation formula is: The final output is from and The two-state coefficient of variation pairs.

4. The neurofeedback reaction time assessment and cognitive training method as described in claim 3, characterized in that, In step S202, the global coefficient of variation smoothing factor The value of is less than the local variation coefficient smoothing factor mentioned in step S203. .

5. The neurofeedback reaction time assessment and cognitive training method as described in claim 1, characterized in that, Step S30 specifically includes: Step S301: Based on the global coefficient of variation and the local coefficient of variation Constructing unlimited smoothing coefficients The calculation formula is: in, The current unlimited smoothing coefficient. As a long-range stability penalty factor, It is a short-range impact enhancement factor; Step S302: Based on the minimum delay budget and the maximum allowable detection delay Determine the lower limit of the smoothing coefficient With upper limit of smoothing coefficient The calculation formula is: in, Used to limit slow updates to the microstate center Used to limit the microstate center from updating too quickly; Step S303: Adjust the unlimited smoothing coefficient Limited to the lower limit of the smoothing coefficient and the upper limit of the smoothing coefficient Within the closed interval formed, the dynamic smoothing coefficient at the current time is obtained. The calculation formula is: .

6. The neurofeedback reaction time assessment and cognitive training method as described in claim 1, characterized in that, Step S40 specifically includes: Step S401: Obtain the observed value of the reaction time for the current trial. Dynamic smoothing coefficient at the current moment and the center estimate of the previous time step ; Step S402: According to the dynamic smoothing coefficient at the current time Observed values ​​for the current reaction time and the center estimate of the previous time step Perform a time-weighted update and output the updated reaction time series index weighted value. , .

7. The neurofeedback reaction time assessment and cognitive training method as described in claim 6, characterized in that, Step S50 specifically includes: Step S501: Obtain the adaptive threshold baseline offset coefficient The updated reaction time series index weighted value The current dynamic smoothing coefficient The observed value of the current reaction time. Local variance estimate at the previous time step ; Step S502: Based on the dynamic smoothing coefficient at the current time The local variance estimates are updated using an exponential weighting method, calculated as follows: in, This is the local variance estimate at the current time. Step S503: Based on the updated reaction time series index weighted value The estimated local variance at the current time. and the adaptive threshold reference offset coefficient The adaptive detection threshold bound for the current time step is constructed using the following formula: in, The upper control limit is set at the current moment. The lower control limit at the current moment, The output is the local standard deviation estimate at the current time. and The system is composed of an adaptive detection threshold bound for the current moment; finally, it determines whether a micro-state switch has occurred in the current trial based on the adaptive detection threshold bound, and outputs the micro-state switch detection result and neural feedback trigger control command for the current moment.

8. A neurofeedback reaction time assessment and cognitive training system, applied to the neurofeedback reaction time assessment and cognitive training method according to any one of claims 1 to 7, characterized in that, The system includes: The parameter initialization module is used to obtain the initial configuration parameter set formed by the reaction time acquisition unit and the EEG synchronous acquisition unit, perform the parameter initialization task based on the initial configuration parameter set, and output the microstate tracking benchmark parameter set. The fluctuation feature extraction module is used to obtain the current trial reaction time observation value. Based on the microstate tracking benchmark parameter set and the current trial reaction time observation value, the fluctuation feature extraction task is performed using the bistate coefficient of variation decomposition method, and the output is a bistate coefficient of variation pair consisting of long-range stationary coefficient of variation and short-range impact coefficient of variation. The smoothing coefficient modulation module is used to perform a real-time adjustment task of tracking agility based on the dual-state variation coefficient pair using a two-dimensional nonlinear smoothing coefficient modulation method, and outputs the dynamic smoothing coefficient at the current moment. The microstate center tracking module is used to perform real-time microstate center tracking tasks based on the observed reaction time of the current trial and the dynamic smoothing coefficient at the current moment using a time-weighted update method, and output the updated reaction time series index weighted value. The threshold boundary generation module is used to construct an adaptive detection threshold boundary by performing variance and mean covariance threshold based on the updated reaction time index weighted value, the current trial reaction time observation value, and the current time dynamic smoothing coefficient; it determines whether a micro-state switch of reaction time has occurred in the current trial based on the adaptive detection threshold boundary, and outputs the micro-state switch detection result and neural feedback trigger control command at the current time.

9. A neurofeedback reaction time assessment and cognitive training device, characterized in that, The neurofeedback reaction time assessment and cognitive training device includes: a memory, a processor, and a neurofeedback reaction time assessment and cognitive training program stored in the memory and executable on the processor. When the neurofeedback reaction time assessment and cognitive training program is executed by the processor, it implements a neurofeedback reaction time assessment and cognitive training method according to any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a neurofeedback reaction time assessment and cognitive training program, which, when executed by a processor, implements a neurofeedback reaction time assessment and cognitive training method according to any one of claims 1 to 7.

Images