MCI intervention system based on virtual reality cognitive motion training difficulty dynamic adjustment

Abstract

The invention discloses an MCI intervention system based on virtual reality cognitive motion training difficulty dynamic adjustment. The MCI intervention system comprises an upper limb motion task prototype module, an upper limb motion task dynamic difficulty adjustment module, a virtual cognitive task prototype module and a virtual cognitive task dynamic difficulty adjustment module. A dynamic difficulty adjustment algorithm based on Pareto frontier is introduced in the system, modeling is carried out on the relation between task intensity and participant performance, possible combinations of various task difficulties and corresponding tradeoffs are evaluated from the relation, and therefore dynamic optimization adjustment of the task difficulties is achieved. Finally, through random contrast research carried out in cooperative hospitals, the obvious effect of the system in the aspect of improving the cognitive ability and the upper limb movement function is verified. The invention provides an innovative system which combines cognition and exercise training and performs personalized intervention through a dynamic difficulty adjustment mechanism, and an effective path is provided for health management and cognitive function improvement of old people.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual reality intervention for elderly cognitive impairment in non-drug intervention, and in particular to an MCI intervention system based on dynamic adjustment of virtual reality cognitive motor training difficulty. Background Art

[0002] Alzheimer's disease is a global challenge and economic burden. The number of people with Alzheimer's disease is increasing rapidly in aging societies. The World Health Organization calls on countries to develop strategies to reduce the risk of Alzheimer's disease, including promoting ongoing research to optimize brain health and specialized prevention of brain diseases. Fortunately, there is growing evidence that about 40% of Alzheimer's disease can be attributed to lifestyle-related risk factors, so Alzheimer's disease is considered preventable and reducing the risk now seems feasible.

[0003] At present, in addition to traditional psychological cognitive training and physical training, virtual reality technology, as an innovative auxiliary therapy, provides a more intuitive and interesting intervention method for elderly patients with cognitive impairment by constructing a multi-sensory, personalized, and interactive virtual environment, which can effectively enhance the participation enthusiasm of patients. In recent years, studies have explored the combination of virtual environments with a variety of non-drug intervention strategies (such as cognitive games, recollection therapy, music therapy, etc.), and simulated daily life and sports activities with the help of Microsoft Kinect, large screens, computers, safety belts, treadmills and other equipment, aiming to improve patients' cognitive function and physical health. Although these studies have initially shown positive effects, they have not been widely used in MCI patients due to problems such as high training time requirements, insufficient personalized adaptation, and complex equipment configuration. In addition, existing virtual reality intervention programs are mostly based on linear difficulty adjustment based on the interactive performance of participants, ignoring the possibility of enhancing the ability of participants through more challenging settings, resulting in a relatively single dynamic task adjustment mechanism. Therefore, it is of great practical significance to develop an intervention program with a flexible task difficulty adjustment mechanism, which is simple, efficient, highly personalized, and easy to promote, to improve cognitive and physical functions. Summary of the invention

[0004] The present invention proposes a virtual reality cognitive motor intervention system based on dynamic difficulty adjustment, which enhances the intervention effect and the ability to adapt to individual differences through fine task design, real-time difficulty adjustment and integration of portable devices. In this system, the intervention system is designed based on sitting operation and has wide applicability in multiple scenarios such as home, community and nursing home. The intervention system designed by the present invention is designed based on sitting operation and has wide applicability in multiple scenarios such as home, community and nursing home.

[0005] The technical system of the present invention is as follows:

[0006] An MCI intervention system based on dynamic adjustment of virtual reality cognitive motor training difficulty, comprising an upper limb motor task prototype module, an upper limb motor task dynamic difficulty adjustment module, a virtual cognitive task prototype module and a virtual cognitive task dynamic difficulty adjustment;

[0007] Upper limb movement task prototype module: Determine the upper limb movement task based on the requirements. The design process of this task covers the upper limb movement action design, interaction mode and virtual environment evaluation;

[0008] The upper limb movement task dynamic difficulty adjustment module adopts a dynamic difficulty adjustment algorithm based on the Pareto optimal theory. It generates a set of Pareto optimal solutions by modeling the difficulty of the upper limb movement task, i.e., the intervention intensity, and the movement performance of the participants, i.e., the intervention performance. Then, by analyzing the trade-off characteristics of each system in the generated solution set, the most suitable solution is selected so that the task parameters or difficulty can be adjusted accordingly.

[0009] Virtual cognitive task prototype module: Determine the virtual cognitive task based on the requirements. The design process of this task covers the cognitive domain task design, interaction mode and style layout evaluation;

[0010] The dynamic difficulty adjustment of virtual cognitive tasks adopts a dynamic difficulty adjustment algorithm based on Pareto optimal theory to improve the efficiency of difficulty adjustment of cognitive tasks. This algorithm only has different objective functions in each cognitive task, and the rest of the operation process is consistent with the upper limb movement task.

[0011] The system and the upper limb movement task prototype: the upper limb movement tasks include elbow flexion and extension, upper limb raising and shoulder rotation; elbow flexion and extension are performed in a four-beat cycle, while upper limb raising and shoulder rotation are designed in an eight-beat cycle.

[0012] The system and the dynamic difficulty adjustment of the upper limb movement task are as follows: two decision objective functions are defined: intervention intensity and intervention performance, wherein the intervention intensity is related to the difficulty of the task, and the intervention performance is closely related to the action performance of the participant. By optimizing these two objectives, effective adaptation and improvement of the participant's ability can be achieved; the intervention intensity formula is:

[0013]

[0014] Among them, I i is the standardized value of the i-th intervention intensity, v xi 、v yi and v zi are the speeds of the ball in the x, y and z axes at the i-th time, min(I) and max(I) are the minimum and maximum intervention intensities, I i The larger the value, the greater the difficulty;

[0015] The intervention performance formula is:

[0016] P i =1-[λ×(1-Accuracy i )+(1-λ)×Time i ]

[0017] Among them, P i is the intervention performance at the ith time, Accuracy i and Time i are the accuracy and standardized time of the participants in the upper limb task at the i-th time, P i The larger the value, the better the participant’s performance;

[0018] Secondly, a set of candidate solutions is generated in the search space by using a random algorithm. Each candidate solution is a point in the search space and consists of a set of values ​​of decision variables.

[0019] Next, for each candidate solution, its mapping value in the objective function definition is calculated, and at the same time, in each practice phase before the intervention, the intervention intensity generated by each candidate solution is applied to the participant's practice to generate the target value of the intervention performance;

[0020] Then, in order to ensure the effectiveness of sorting solutions within the Pareto front and reduce computational complexity, an efficient non-dominated sorting method is introduced, in which all solutions in the feasible solution set are sorted in ascending order according to the intervention intensity of the first objective value. If the first objective values ​​of two or more solutions are the same, they are sorted in ascending order according to the intervention performance of the second objective value. If all objective values ​​of two or more solutions are the same, their sorting can be arbitrary. Through this sorting process, the efficient non-dominated sorting can gradually allocate solutions to the Pareto front, thereby optimizing the decision-making process.

[0021] Finally, a method to measure the trade-off information of a pair of non-dominated solutions is introduced. The trade-off information of each solution on the frontier is calculated to select one of the solutions as the new game level. The formula is:

[0022]

[0023] Among them, F m (v i ) and F m (v j ) are respectively the solution system v i and v j The value of the solution of the mth objective function, and is the maximum and minimum value of the mth objective function value in the candidate solution system;

[0024] In addition, any solution v i The trade-off value can be given by the following formula:

[0025]

[0026] Among them, v i and v j are mutually non-dominant solutions in the non-dominant solution set S, μ(v i ,S) indicates that v i Replace any solution v in the solution set S j The minimum improvement resulting from a decrease in an indicator.

[0027] The virtual cognitive tasks included three virtual cognitive tasks, namely, the box-block test, used to train executive functions, the working memory test, used to improve memory ability, and the depth perception test, used to train spatial positioning ability;

[0028] Box-block test: Participants are required to move a specified number of virtual blocks from one compartment to another, requiring their hands to pass through the middle partition; two extended versions of the test are designed: the first requires participants to place the grabbed virtual blocks in the corresponding area according to color matching; the second requires participants to place the grabbed blocks in the white area in order, row by row, to improve their task organization and execution ability;

[0029] Working memory test: This task is designed to assess the participant's ability to process, manipulate and store information in a short period of time. The participant must first remember the location of a series of objects, which will then disappear. The system will randomly display one of the objects, and the participant must put it back to its original location accurately. This process will be repeated until all objects are correctly placed. If the participant's error rate exceeds 2 / 3 for three consecutive times in the test, the test will be terminated.

[0030] Depth perception test: This test uses multiple white balls in a virtual environment to stimulate the subject's spatial positioning and execution ability; the participant needs to click on the balls in order of far and near or near and far, and the system records the order of clicks and reaction time. The reaction time is composed of the interval between each click of the ball, which is ultimately used to evaluate the subject's cognitive performance.

[0031] The system and the dynamic difficulty adjustment of the virtual cognitive task: during the intervention process of the virtual cognitive task, the dynamic adjustment of the task difficulty is similar to that of the upper limb movement task, and both need to be optimized in real time according to the individual performance of the participants; a dynamic difficulty adjustment algorithm based on the Pareto optimal theory is adopted to improve the efficiency of the difficulty adjustment of the cognitive task. The algorithm only has different objective functions in each cognitive task, and the rest of the operation process is consistent with the upper limb movement task.

[0032] In the system, the objective function settings of the three cognitive tasks include the objective function of the virtual box block test, the objective function of the working memory test and the objective function of the depth perception test.

[0033] The objective function of the system and the virtual box test is:

[0034] In this task, participants are required to move a specified number of blocks from one side of the compartment to the other as efficiently as possible without a time limit; the number of blocks in the task will be dynamically adjusted based on the completion time of the participant in the previous task to ensure that the difficulty matches the ability level of the participant; therefore, the objective function is defined as:

[0035]

[0036] in, is the number of blocks dragged and dropped for the ith time, ranging from [30,31,...,80], and are the start and end time of the i-th drag and drop.

[0037] The system, the objective function of the working memory test:

[0038] In this task, intervention intensity is defined as a composite measure of the number of items and the correlation between items. At the same time, intervention performance is evaluated based on two key performance indicators: the accuracy of item recall and the time to complete the task. The objective function is defined as:

[0039]

[0040] in, is the number of items in the i-th time, and its value range is [3,4,...,8], is the correlation between items at the i-th time, N i To correctly recall the number of items, is the normalized time for item recall at the ith time.

[0041] The objective function of the system and the depth perception test is:

[0042] In this task, the adjustment of ball density and depth distance will directly affect the difficulty level of participants searching and selecting balls in the virtual environment; at the same time, the evaluation of intervention performance is based on accuracy and completion time; therefore, the objective function is defined as:

[0043]

[0044] in, is the number of balls in the ith time, ranging from [3,4,...,8], is the depth distance of the ith time, and its value range is [10,11,...,60], To click the ball accuracy, Normalize the time for clicking the ball.

[0045] The beneficial effects of the present invention are:

[0046] The present invention proposes a virtual reality cognitive-motor intervention strategy based on dynamic difficulty adjustment, aiming to improve the intervention effect and enhance adaptability to individual differences. The strategy combines three cognitive trainings with three upper limb movement tasks to comprehensively train the cognitive and motor functions of the elderly. By introducing a dynamic difficulty adjustment algorithm based on Pareto optimality theory, the intervention process can be comprehensively evaluated based on the performance of the participants and the intensity of the intervention, dynamically optimize the training system, and select the best system from the Pareto optimal solution set. This method effectively realizes the personalized adjustment of training difficulty, which not only improves the enthusiasm and participation of the participants, but also maximizes the intervention effect. The results of the present invention provide an important reference for the prevention and management of cognitive impairment in the elderly, and demonstrate the application potential of personalized intervention in improving the welfare of the elderly. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of the co-design and development process for the intervention system of the present invention;

[0048] Figure 2 It is a schematic diagram of the decomposition of upper limb movement actions based on sitting posture and the execution sequence thereof of the present invention;

[0049] Figure 3 Schematic diagram of virtual environments based on different styles of the present invention; (a) urban environment; (b) hospital environment; (c) natural environment; (d) fitness environment;

[0050] Figure 4 Schematic diagram of the Pareto-based dynamic difficulty adjustment algorithm of the present invention; (a) defining the objective function; (b) generating a candidate solution set; (c) evaluating the objective function; (d) non-dominated sorting; (e) weighing the value of non-dominated solutions;

[0051] Figure 5 Schematic diagram of three different arrangements of the target area based on the box block test of the present invention; (a) 4×4 arrangement distribution; (b) 3×3 arrangement distribution; (c) 2×2 arrangement distribution;

[0052] Figure 6 Schematic diagram of the participants' performance on different tasks during the 12-week intervention process under the intervention system proposed in the present invention; (a) elbow flexion and extension; (b) upper limb raising; (c) shoulder wrap; (d) box-block test; (e) working memory; and (f) depth perception. DETAILED DESCRIPTION

[0053] The present invention is described in detail below in conjunction with specific embodiments.

[0054] An MCI intervention system based on dynamic adjustment of virtual reality cognitive motor training difficulty, characterized by including a demand determination module, an upper limb motor task prototype design module, a virtual cognitive task prototype design module and a field research module. The overall process is as follows Figure 1 shown.

[0055] A1 In order to design an intervention task that is easy to perform and efficient, the present invention deeply understands the intervention training program suitable for elderly people with cognitive decline, the psychology of the elderly, and their acceptance and preference for virtual reality interaction methods from the two perspectives of task requirements and system requirements.

[0056] The specific steps of step A1 are as follows:

[0057] A11.Task requirements.

[0058] When exploring the cognitive field, the research team emphasized the significant advantages of virtual reality technology in spatial perception and executive function training compared to traditional devices (such as tablets, smartphones and large-screen displays), especially in providing a more realistic and intuitive interactive experience. Experts in the field of human-computer interaction recommend migrating cognitive tasks to virtual environments to provide a more immersive and realistic training experience for the elderly. At the same time, clinical observations of neurologists show that many elderly patients (whether in outpatient clinics or nursing institutions) often lack the motivation to exercise and feel uncomfortable with high-intensity exercise. To address this issue, the focus is on the design of seated upper limb movement tasks, aiming to create a safe and comfortable training environment for the elderly to reduce the risk of injury and increase their enthusiasm and persistence in participation.

[0059] A12.System requirements.

[0060] Combined with rich clinical experience with the elderly, specific expectations for the system are proposed: first, the task design should be concise and clear to ensure that users can easily understand and operate it; second, the accessibility of the system should be highly convenient so that elderly users can use it easily; then, the system's engagement should be enhanced by integrating visual, auditory, tactile feedback and reward mechanisms to optimize the user experience; finally, the intervention content should cover a wide range of cognitive and motor training, aiming to comprehensively improve the rehabilitation effect of the elderly.

[0061] A2 is an upper limb movement task determined based on needs. The design process of this task covers upper limb movement action design, interaction methods and virtual environment evaluation, as well as dynamic difficulty adjustment of upper limb movement tasks.

[0062] The specific steps of step A2 are as follows:

[0063] A21. Upper limb movement design.

[0064] The upper limb movement tasks were designed to target specific upper limb movements, ensuring that the tasks effectively activated key muscle groups and covered an adequate range of motion, thereby enhancing the effectiveness of the training. Upper limb movement patterns suitable for sitting were identified. Figure 2 As shown in the figure, the designed upper limb movements include elbow flexion and extension, upper limb lifting and shoulder circle movement. These movements have been carefully designed, with elbow flexion and extension movement taking four beats as a cycle, while upper limb lifting and shoulder circle use eight beats as a cycle to ensure the scientific rhythm of the movement and the effectiveness of the training.

[0065] A22. Interaction methods and virtual environment evaluation.

[0066] A pilot evaluation of the system prototype was conducted based on the selected upper limb motor training, involving 15 elderly participants from two nursing homes and two community hospitals. The experiment was conducted in a spacious indoor space, and the participants were arranged to sit on chairs with their hands naturally placed on both sides of the body, and completed specific upper limb motor tasks as required. To minimize the risk of potential side effects, each virtual reality exposure time was controlled to no more than 20 minutes. The evaluation covered two interaction methods: one was to interact by touching virtual objects, and the other was to simulate grabbing actions by pressing buttons. In addition, the study also explored the impact of virtual environment style on participants' preferences, providing a variety of environmental options such as cities, hospitals, natural scenery, and fitness scenes, such as Figure 3 shown.

[0067] A23. Dynamic difficulty adjustment of upper limb movement tasks.

[0068] In upper limb movement tasks, dynamic difficulty adjustment is regarded as a dual-objective optimization problem, the core of which is to balance the two objective functions of intervention intensity and intervention performance. To this end, the present invention proposes a dynamic difficulty adjustment algorithm based on Pareto optimality theory, such as Figure 4 As shown. The algorithm generates a set of Pareto optimal solutions by modeling the difficulty of the upper limb movement task (i.e., intervention intensity) and the participant's motor performance (i.e., intervention performance). Then, by analyzing the trade-off characteristics of each system in the generated solution set, the most suitable solution system is selected so that the task parameters or difficulty can be adjusted accordingly. First, the two decision objective functions are defined as intervention intensity and intervention performance, respectively. The intervention intensity reflects the difficulty level of the task, while the intervention performance represents the participant's motor performance in the task. The intervention intensity formula is:

[0069]

[0070] Among them, I iis the standardized value of the i-th intervention intensity, v xi 、v yi and v zi are the speeds of the ball in the x, y and z axes at the i-th time, min(I) and max(I) are the minimum and maximum intervention intensities, I i The larger the value, the greater the difficulty.

[0071] The intervention performance formula is:

[0072] P i =1-[λ×(1-Accuracy i )+(1-λ)×Time i ]

[0073] Among them, P i is the intervention performance at the ith time, Accuracy i and Time i are the accuracy and standardized time of the participants in the upper limb task at the i-th time, P i The larger the value, the better the participants’ performance. In this experiment, λ is taken as 0.3.

[0074] Secondly, a set of candidate solutions is generated in the search space by using a random algorithm. Each candidate solution corresponds to a point in the search space and is composed of the values ​​of several decision variables. Specifically, the speed of the ball in the x, y and z axes is used as the decision variable, and v x 、v y 、v z For each axis, preset its speed range. Taking the x-axis as an example, its speed range is v x ={1,2,...,8},v y and v z The value range of v x Then, according to their respective speed ranges, a certain number of feasible solutions are randomly generated, and the values ​​of these solutions are respectively obtained from the corresponding v x 、v y 、v z Randomly selected from the speed range.

[0075] Next, for each candidate solution, its mapping value in the objective function definition is calculated. In particular, in the practice phase before each intervention, the intervention intensity generated by each candidate solution is applied to the participant's practice to generate the target value of the intervention performance. Subsequently, the dual objective function under each candidate solution is encapsulated as a two-dimensional objective function vector F i =[I i ,P i], and the previous intervention information (two-dimensional objective function vectors corresponding to different candidate solutions) will also be included to construct a new objective function vector F = [F1, F2, ..., F i ,...] to ensure that the algorithm gradually approaches the optimal solution or Pareto front when searching the solution space.

[0076] Then, in order to ensure the effectiveness of sorting solutions within the Pareto front and reduce computational complexity, an efficient non-dominated sorting is introduced, in which all solutions in the feasible solution set are sorted in ascending order according to the first objective value (intervention intensity). If the first objective values ​​of two or more solutions are the same, they are sorted in ascending order according to the second objective value (intervention performance). If all objective values ​​of two or more solutions are the same, their sorting can be arbitrary. Through this sorting process, efficient non-dominated sorting can gradually assign solutions to the Pareto front, thereby optimizing the decision-making process.

[0077] Finally, a method to measure the trade-off information of a pair of non-dominated solutions is introduced. The trade-off information of each solution on the frontier is calculated to select one of the solutions as the new game level. The formula is:

[0078]

[0079] Among them, F m (v i ) and F m (v j ) are respectively the solution system v i and v j The value of the solution of the mth objective function, and is the maximum and minimum value of the mth objective function value in the candidate solution system.

[0080] In addition, any solution v i The trade-off value can be given by the following formula:

[0081]

[0082] Among them, v i and v j are mutually non-dominant solutions in the non-dominant solution set S, μ(v i ,S) indicates that v i Replace any solution v in the solution set S j The minimum improvement resulting from a decrease in an indicator.

[0083] A3 is a virtual cognitive task determined based on needs. The design process of this task covers cognitive domain task design, interaction mode and style layout evaluation, and dynamic difficulty adjustment of virtual cognitive tasks.

[0084] The specific steps of step A3 are as follows:

[0085] A31. Task design in virtual cognitive domain.

[0086] In order to better meet the needs of the elderly in cognitive training, three virtual cognitive tasks were identified, namely the box block test (for training executive function), the working memory test (for training memory), and the depth perception test (for training depth perception).

[0087] Box-block test: Participants are required to move a specified number of virtual blocks from one compartment to another. Their hands must pass through the middle partition. When they successfully pick up a block, the controller will vibrate to provide tactile feedback. In addition, two extended versions of the test were designed: the first extended version requires participants to place the blocks in the corresponding areas according to color matching, with red blocks placed in the red area, yellow blocks in the yellow area, and so on. This version is designed to strengthen participants' spatial visual perception ability and test their ability to extract and retain classified information. The second extended version requires participants to place the grabbed blocks in the white area row by row in order. If the first white area is empty, the blocks must not be placed in the second white area. It aims to improve participants' ability to precisely control their movements.

[0088] Working memory: Participants first need to remember the positions of a series of objects, then all the objects will temporarily disappear, and the system will randomly present one of the objects and ask the participants to put it back to the original position. This process will be repeated until all the objects are correctly restored. The correctness of each position is displayed through visual feedback. If the error rate of restoring objects exceeds 2 / 3 in three consecutive tasks, the test is terminated. This task is designed to assess the participants' cognitive ability to process, manipulate and store information in a short period of time, especially the decision-making ability when performing multiple tasks at the same time.

[0089] Depth perception test: In a virtual environment, multiple white spheres are randomly arranged, and participants are asked to click on the white spheres as quickly as possible in order from far to near or from near to far. During the task, the system will record the time interval between each click on the sphere in real time and calculate the reaction time. This task is mainly used to evaluate the spatial perception ability, reaction speed and executive control of the participants. Through a comprehensive analysis of the reaction time and the accuracy of the click sequence, the cognitive strategies and efficiency of the participants in spatial positioning and target selection are further explored.

[0090] A32. Interaction style and layout evaluation.

[0091] To explore the interactive experience of the elderly when using a virtual reality controller for grasping operations, an interactive experiment was designed and implemented. During the experiment, participants interacted with a variety of virtual objects (such as cubes, balls, cars, and vases) using the Oculus Quest 2 controller. The experiment was divided into two stages: in the first 3 minutes, participants completed the grasping operation by pressing the trigger button of the controller, and then switched to the side grip button to perform the same task. After the experiment, unstructured interviews were used to collect participants' subjective feedback on the two operation methods, covering aspects such as comfort, ease of use, and personal preferences.

[0092] In addition, in order to further explore the impact of different interactive layouts on the operating experience, the experiment also introduced an extended version of the box test, using three different target area arrangements of 4×4, 3×3 and 2×2 to examine the impact of these arrangements on the participants' operating perception. Figure 5 shown.

[0093] A33. Dynamic difficulty adjustment of virtual cognitive tasks.

[0094] During the intervention of virtual cognitive tasks, the dynamic adjustment of task difficulty is similar to that of upper limb movement tasks, and both need to be optimized in real time according to the individual performance of the participants. To this end, the study introduced a dynamic difficulty adjustment algorithm based on Pareto optimality theory to improve the efficiency of difficulty adjustment of cognitive tasks. It is worth noting that the algorithm only has different objective functions in each cognitive task, and the rest of the operation process is consistent with the upper limb movement task. The objective function settings of the three cognitive tasks are introduced below.

[0095] Objective function of the virtual box test:

[0096] In this task, participants are required to move a specified number of blocks from one side of the compartment to the other as efficiently as possible without a time limit. The number of blocks in the task will be dynamically adjusted based on the participant's completion time in the previous task to ensure that the difficulty matches the participant's ability level. Therefore, the objective function is defined as:

[0097]

[0098] in, is the number of blocks dragged and dropped for the ith time, ranging from [30,31,...,80], and are the start and end time of the i-th drag and drop.

[0099] Objective function for working memory tests

[0100] In this task, intervention intensity is defined as a composite measure of the number of items and the correlation between items. At the same time, intervention performance is evaluated based on two key performance indicators: the accuracy of item recall and the time to complete the task. Therefore, the objective function is defined as:

[0101]

[0102] in, is the number of items in the i-th time, and its value range is [3,4,...,8], is the correlation between items at the i-th time, N i To correctly recall the number of items, is the standardized time for recalling the i-th item. In this experiment, λ1 and λ2 are set to 0.8 and 0.6.

[0103] Objective function for depth perception testing

[0104] In this task, the adjustment of ball density and depth distance will directly affect the difficulty level of participants searching and selecting balls in the virtual environment. At the same time, the evaluation of intervention performance is based on accuracy and completion time. Therefore, the objective function is defined as:

[0105]

[0106] in, is the number of balls in the ith time, ranging from [3,4,...,8], is the depth distance of the ith time, and its value range is [10,11,...,60], To click the ball accuracy, In order to standardize the time of clicking the ball, λ3 and λ4 were taken as 0.6 and 0.6 in this experiment.

[0107] Field study: 30 elderly participants were recruited from various channels such as community hospitals, clinics, and long-term care institutions, and were randomly divided into an intervention group and a control group. The intervention group will receive virtual reality-based cognitive and motor training during the 12-week intervention period, while the control group will not receive any intervention and maintain their daily living habits. The study conducted baseline assessments on both groups of participants before the intervention (such as cognitive ability screening tools, mini-mental state examinations, and Montreal Cognitive Assessment Scales), and measured them again after 12 weeks to conduct a comparative analysis of the effects between the groups.

[0108] Figure 6 The box plots show the participants’ performance on different tasks during the 12-week intervention, with the top numbers indicating the average scores for each week. Figure 6(a)), the average score of the participants in the first week was only 0.457, which was a poor performance. It is speculated that the randomly assigned training intensity was too high. Table 1 shows that the average intensity in the first week was 0.617, and the average intensity in the subsequent weeks dropped to 0.503, which supports this inference. Figure 6 (b) shows that the performance of the upper limb raising exercise improved slightly from week 2 to week 12, reflecting the participants' gradual adaptation to the high-intensity training. Although outliers indicate that some participants performed poorly, these cases were usually affected by external factors unrelated to the task. Figure 6 (c)) The average score in the first week was the best, reaching 0.547, but then the average score stabilized at a lower level. The randomly selected lower intensity may have led to excellent results in the first week, but also caused boredom in the subsequent period. Therefore, the subsequent intervention adopted a dynamic difficulty adjustment strategy based on Pareto optimality theory, which increased the optimal average intensity to 0.731 (see Table 1). Although the overall performance decreased slightly, the performance of each week maintained a certain consistency.

[0109] also, Figure 6 (d) to (f) show performance on three virtual cognitive tasks. On the virtual box block test, participants’ mean completion time improved significantly, from 114 seconds in week 1 to 60 to 80 seconds. Table 1 shows that the optimal intensity for this task was 45 seconds, with a randomly selected mean of 60 seconds designed to balance task complexity with cognitive load to promote stable and efficient performance. For the working memory test ( Figure 6 (e)), although the performance did not change significantly before and after the optimal intensity was selected, the outliers in the first week indicated that the intervention intensity was low, resulting in abnormally good performance. Table 1 further verifies this, showing that the average intensity in the first week was 0.742, while the optimal intensity in the subsequent weeks increased to 0.814. Finally, Figure 6 The analysis of (f) shows that after difficulty adjustment, the performance in subsequent weeks significantly improved compared to the first week, from 0.756 to 0.678, emphasizing the effectiveness and adaptability of the difficulty adjustment strategy in the task execution process.

[0110] Table 1 Average intensity of different tasks during the 12-week intervention

[0111]

[0112]

[0113] Table 2 shows the statistical analysis results of the intervention group and the control group on the cognitive assessment scale at baseline and posttest. The data of the control group showed that the scores of each test in the posttest were significantly improved compared with the baseline level. This trend was also observed in the intervention group, and the increase in the scores of the intervention group was more significant. At the baseline measurement, the scores of the intervention group and the control group on each test were basically the same, while in the posttest, the scores of the intervention group were significantly higher than those of the control group. In addition, the independent sample t-test was used to compare the scores of the cognitive assessment scales at baseline and posttest, and it was found that all cognitive test scores of the control group did not show statistically significant differences (P>0.05), indicating that the cognitive state of this group remained stable during the study period. In contrast, in the comparison before and after the intervention, the scores of the intervention group showed significant improvement in the scores of the cognitive function tests except for the long-term memory, orientation ability, abstract thinking and judgment tests (P<0.05), indicating that the implemented intervention system had a positive impact on specific cognitive areas.

[0114] Table 2 Results of cognitive assessment scales in the intervention group and the control group at baseline and post-test

[0115]

[0116] To evaluate the significance of the intervention effect, the present invention conducted a covariance analysis on the continuous variables of the two groups before and after the intervention, and the results are shown in Table 3. Compared with the control group, the scores of the participants in the intervention group on the cognitive ability screening tool after the intervention were significantly improved (F = 201.571, p < 0.001, η2 = 0.780), indicating the effectiveness of the intervention system in enhancing overall cognitive function. Except for long-term memory, abstract thinking and judgment (P> 0.05), most cognitive domains showed significant improvement after the intervention (P < 0.05), suggesting that the intervention system had a positive impact on most cognitive domains, although there were some differences in the effects between different domains. In addition, the intervention group was also significantly better than the control group in the Mini-Mental State Examination score (F = 140.810, P < 0.001, η2 = 0.712) and the Montreal Cognitive Assessment score (F = 94.834, p < 0.001, η2 = 0.625), further confirming the effectiveness of the intervention system in a wide range of cognitive domains. Overall, η2 values ​​for the cognitive ability screening tool, Mini-Mental State Examination, and Montreal Cognitive Assessment scores ranged from 0.625 to 0.780, reflecting relatively high effect sizes, indicating that participants in the intervention group had significant improvements in cognitive function after 12 weeks of intervention training.

[0117] Table 3 Effect size of intervention system on improvement of cognitive function

[0118]

[0119] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. An MCI intervention system based on dynamic adjustment of virtual reality cognitive motor training difficulty, characterized in that: It includes upper limb movement task prototype module, upper limb movement task dynamic difficulty adjustment module, virtual cognitive task prototype module and virtual cognitive task dynamic difficulty adjustment; Upper limb movement task prototype module: Determine the upper limb movement task based on the requirements. The design process of this task covers the upper limb movement action design, interaction mode and virtual environment evaluation; The upper limb movement task dynamic difficulty adjustment module adopts a dynamic difficulty adjustment algorithm based on the Pareto optimal theory. It generates a set of Pareto optimal solutions by modeling the difficulty of the upper limb movement task, i.e., the intervention intensity, and the movement performance of the participants, i.e., the intervention performance. Then, by analyzing the trade-off characteristics of each system in the generated solution set, the most suitable solution is selected so that the task parameters or difficulty can be adjusted accordingly. Virtual cognitive task prototype module: Determine the virtual cognitive task based on the requirements. The design process of this task covers the cognitive domain task design, interaction mode and style layout evaluation; The dynamic difficulty adjustment of virtual cognitive tasks adopts a dynamic difficulty adjustment algorithm based on Pareto optimal theory to improve the efficiency of difficulty adjustment of cognitive tasks. This algorithm only has different objective functions in each cognitive task, and the rest of the operation process is consistent with the upper limb movement task.

2. The system according to claim 1, characterized in that The upper limb movement task prototype: the upper limb movement task includes elbow flexion and extension movement, upper limb raising movement and shoulder circle movement; elbow flexion and extension movement is performed in a four-beat cycle, while upper limb raising and shoulder circle are designed with an eight-beat cycle.

3. The system according to claim 1, characterized in that The dynamic difficulty adjustment of the upper limb movement task described above: defines two decision objective functions: intervention intensity and intervention performance, where the intervention intensity is related to the difficulty of the task, and the intervention performance is closely related to the participant's movement performance. By optimizing these two objectives, effective adaptation and improvement of the participant's ability can be achieved; the intervention intensity formula is: Among them, I i is the standardized value of the i-th intervention intensity, v xi 、v yi and v zi are the speeds of the ball in the x, y and z axes at the i-th time, min(I) and max(I) are the minimum and maximum intervention intensities, I i The larger the value, the greater the difficulty; The intervention performance formula is: P i =1-[λ×(1-Accuracy i )+(1-λ)×Time i ] Among them, P i is the intervention performance at the ith time, Accuracy i and Time i are the accuracy and standardized time of the participants in the upper limb task at the i-th time, P i The larger the value, the better the participant’s performance; Secondly, a set of candidate solutions is generated in the search space by using a random algorithm. Each candidate solution is a point in the search space and consists of a set of values ​​of decision variables. Next, for each candidate solution, its mapping value in the objective function definition is calculated, and at the same time, in each practice phase before the intervention, the intervention intensity generated by each candidate solution is applied to the participant's practice to generate the target value of the intervention performance; Then, in order to ensure the effectiveness of sorting solutions within the Pareto front and reduce computational complexity, an efficient non-dominated sorting method is introduced, in which all solutions in the feasible solution set are sorted in ascending order according to the intervention intensity of the first objective value. If the first objective values ​​of two or more solutions are the same, they are sorted in ascending order according to the intervention performance of the second objective value. If all objective values ​​of two or more solutions are the same, their sorting can be arbitrary. Through this sorting process, the efficient non-dominated sorting can gradually allocate solutions to the Pareto front, thereby optimizing the decision-making process. Finally, a method to measure the trade-off information of a pair of non-dominated solutions is introduced. The trade-off information of each solution on the frontier is calculated to select one of the solutions as the new game level. The formula is: Among them, F m (v i ) and F m (v j ) are respectively the solution system v i and v j The value of the solution of the mth objective function, and is the maximum and minimum value of the mth objective function value in the candidate solution system; In addition, any solution v i The trade-off value can be given by the following formula: Among them, v i and v j are mutually non-dominant solutions in the non-dominant solution set S, μ(v i ,S) indicates that v i Replace any solution v in the solution set S j The minimum improvement resulting from a decrease in an indicator.

4. The system according to claim 1, characterized in that The virtual cognitive tasks include three virtual cognitive tasks, namely, a box block test, which is used to train executive functions; a working memory test, which is used to improve memory ability; a depth perception test, which is used to train spatial positioning ability; Box-block test: Participants are required to move a specified number of virtual blocks from one compartment to another, requiring their hands to pass through the middle partition; two extended versions of the test are designed: the first requires participants to place the grabbed virtual blocks in the corresponding area according to color matching; the second requires participants to place the grabbed blocks in the white area in order, row by row, to improve their task organization and execution ability; Working memory test: This task is designed to assess the participant's ability to process, manipulate and store information in a short period of time. The participant must first remember the location of a series of objects, which will then disappear. The system will randomly display one of the objects, and the participant must put it back to its original location accurately. This process will be repeated until all objects are correctly placed. If the participant's error rate exceeds 2 / 3 for three consecutive times in the test, the test will be terminated. Depth perception test: This test uses multiple white balls in a virtual environment to stimulate the subject's spatial positioning and execution ability; the participant needs to click on the balls in order of far and near or near and far, and the system records the order of clicks and reaction time. The reaction time is composed of the interval between each click of the ball, which is ultimately used to evaluate the subject's cognitive performance.

5. The system according to claim 1, characterized in that The dynamic difficulty adjustment of the virtual cognitive task: During the intervention process of the virtual cognitive task, the dynamic adjustment of the task difficulty is similar to that of the upper limb movement task, and both need to be optimized in real time according to the individual performance of the participants; a dynamic difficulty adjustment algorithm based on the Pareto optimal theory is adopted to improve the efficiency of the difficulty adjustment of the cognitive task. The algorithm only has different objective functions in each cognitive task, and the rest of the operation process is consistent with the upper limb movement task.

6. The system according to claim 4, characterized in that The objective function settings for the three cognitive tasks include the objective function of the virtual box block test, the objective function of the working memory test, and the objective function of the depth perception test.

7. The system according to claim 6, characterized in that The objective function of the virtual box test is: In this task, participants are required to move a specified number of blocks from one side of the compartment to the other as efficiently as possible without a time limit; the number of blocks in the task will be dynamically adjusted based on the completion time of the participant in the previous task to ensure that the difficulty matches the ability level of the participant; therefore, the objective function is defined as: in, is the number of blocks dragged and dropped for the ith time, ranging from [30,31,...,80], and are the initial and end times of the i-th drag and drop.

8. The system according to claim 6, characterized in that The objective function of the working memory test is: In this task, intervention intensity is defined as a composite measure of the number of items and the correlation between items. At the same time, intervention performance is evaluated based on two key performance indicators: the accuracy of item recall and the time to complete the task. The objective function is defined as: in, is the number of items in the i-th time, and its value range is [3,4,...,8], is the correlation between items at the i-th time, N i To correctly recall the number of items, is the normalized time for item recall at the ith time.

9. The system according to claim 6, characterized in that The objective function of the depth perception test is: In this task, the adjustment of ball density and depth distance will directly affect the difficulty level of participants searching and selecting balls in the virtual environment; at the same time, the evaluation of intervention performance is based on accuracy and completion time; therefore, the objective function is defined as: in, is the number of balls in the ith time, ranging from [3,4,...,8], is the depth distance of the ith time, and its value range is [10,11,...,60], To click the ball accuracy, Normalize the time for clicking the ball.

Images