Intelligent wheelchair static comfort evaluation method based on random forest

Abstract

The invention discloses an intelligent wheelchair static comfort evaluation method based on a random forest, and particularly relates to the technical field of intelligent rehabilitation assistive devices, and the method comprises the following steps: 1, configuring a subjective evaluation information table: building a systematic evaluation index system; 2, obtaining a subjective weight based on AHP; 3, collecting objective body pressure data; 4, calculating a comprehensive subjective comfort level; and 5, subjective and objective data integration and feature optimization. 6, calculating objective comfort through a random forest model: training the model to realize prediction from body pressure data to comfort; and 7, verifying subjective and objective consistency: ensuring that a model prediction result is consistent with the real feeling of a person. According to the method, an intelligent evaluation system with deep integration of subjective and objective data is constructed, and an accurate mapping relation between measurable objective physiological data (body pressure distribution) and subjective comfort feeling difficult to quantify is established through a machine learning model (random forest), so that the objective data is used for predicting the subjective feeling, and the accuracy of the subjective comfort feeling is improved. The purpose of objectively and quantitatively evaluating the comfort level is achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent rehabilitation aids technology, and more specifically, to a method for evaluating the static comfort of intelligent wheelchairs based on random forest. Background Technology

[0002] As society's demands for accessible travel and humanized rehabilitation aids continue to rise, wheelchair design and development are undergoing a profound evolution from a single-function orientation to one that integrates comfort, health, and intelligence. Traditional designs primarily focus on structural strength and basic mobility, neglecting long-term comfort, physiological adaptability, and psychological well-being. Therefore, modern research has expanded the core of comfort evaluation to include deeper ergonomic issues such as body pressure distribution, muscle fatigue, and postural dynamics. The design philosophy of future wheelchairs must fundamentally shift from "meeting basic mobility needs" to "providing a comprehensive comfort experience," integrating scientific and quantitative comfort assessment and optimization into the entire design process, becoming crucial for improving product quality and the tester experience.

[0003] In the research and development and adaptation evaluation of smart wheelchairs, current methods generally rely on users completing subjective questionnaires to assess comfort. This method typically requires testers to qualitatively describe or quantitatively rate the comfort of various parts of the body, such as the shoulders, back, waist, hips, and legs, after experiencing the product. The final judgment on the product's comfort is then made through statistical analysis of the questionnaire data.

[0004] However, the fundamental flaw of this mainstream method is that its evaluation results rely entirely on the individual's subjective feelings and are easily affected by many irrelevant variables such as the tester's instantaneous emotions, personal tolerance, and differences in expression. This results in insufficient objectivity and poor stability of the evaluation conclusions, and different testers or the same tester may give significantly different feedback in different situations.

[0005] To address the aforementioned core issues, this application proposes a static comfort evaluation method for intelligent wheelchairs based on random forests, using machine learning models to establish a reliable correlation between objective physiological data and comfort levels. This method serves as a further improvement, aiming to achieve a more stable and quantifiable comfort evaluation. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for evaluating the static comfort of intelligent wheelchairs based on random forests, in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for evaluating the static comfort of an intelligent wheelchair based on random forest, comprising the following steps: S1: Establish a systematic evaluation index system for subjective comfort; obtain subjective comfort evaluation parameters for testers riding in wheelchairs; S2: Based on the evaluation index system established in S1, the subjective weights of each evaluation index are obtained using the analytic hierarchy process (AHP). S3: Collect objective body pressure distribution data when the tester is in a wheelchair; S4: Based on the subjective weights obtained in S2, the testers calculate the overall subjective comfort level by combining the comfort scores of each part collected in S1. S5: Integrate and optimize the objective body pressure distribution data collected in S3, and combine it with the comprehensive subjective comfort obtained in S4 to construct a model training dataset; S6: Using the feature data in the model training dataset constructed in S5 as input and the corresponding comprehensive subjective comfort quantification value as the output target value, train the random forest regression model to obtain an intelligent model that can predict comfort based on body pressure characteristics; use the random forest algorithm to train the prediction model to obtain a trained model that can calculate objective comfort based on body pressure distribution data. S7: Compare and analyze the objective comfort level output by the trained model described in S6 with the comprehensive subjective comfort level calculated in S4 to verify the consistency between the objective and subjective evaluation results and complete the effectiveness evaluation of the model.

[0008] Furthermore, in S1, S11: Establish a systematic evaluation index system for subjective comfort: Establish a static wheelchair subjective comfort evaluation table, which includes: the wheelchair support scheme to be evaluated and the subjective comfort evaluation corresponding to the wheelchair support scheme; the wheelchair support scheme clearly defines each support part and its adjustable angle parameters; the subjective comfort evaluation is used to quantify the subjective comfort evaluation of the tester under each wheelchair support scheme.

[0009] S12: Obtain subjective comfort evaluation parameters for testers riding in wheelchairs: Testers experience riding in wheelchairs with different angle combinations and score each option according to the evaluation table in S11; the system records all scoring results and initially constructs the subjective evaluation dataset D1 for wheelchair comfort.

[0010] Furthermore, in S11, the wheelchair support solution includes: A seat cushion designed to support the buttocks; A backrest, used to support the shoulders, back, and waist, with an angle α between the backrest and the horizontal plane of the seat cushion, the angle ranging from 10° to 75°; and, The leg plate includes a thigh plate for supporting the thigh and a calf plate for supporting the lower leg. The angle between the thigh plate and the horizontal plane of the seat cushion is set to β, and the range of β is set to 0°-16°. The thigh plate is rotatably connected to the calf plate by a locking member, wherein the locking member is operable to switch between a locked state and a released state. When in the locked state, the relative rotation between the thigh plate and the calf plate is fixed; when in the released state, the two can rotate relative to each other.

[0011] Furthermore, in step S2, the method for obtaining the subjective weights of the evaluation indicators includes the following steps: S21: Generate a test questionnaire. Test participants compare and assign values ​​to the relative importance of indicators at the same level using a 1-9 scale. Construct a judgment matrix for each test subject using the indicator values ​​in the test questionnaire. S22: Based on the judgment matrix of each participant in the test, obtain the corresponding indicator weight, eigenvector, and maximum eigenvalue; S23: Perform a consistency check on the judgment matrix of each tester and calculate the consistency ratio: If the judgment matrix fails the consistency test, it indicates that there is a contradiction in the tester's judgment results. The tester should be notified to conduct the questionnaire survey again; return to S21. If the judgment matrix passes the consistency test, then the group weights are calculated for the judgment matrices that meet the consistency test.

[0012] Furthermore, in S3, the process of collecting objective body pressure distribution data includes the following steps: S31: The pressure distribution test pad is evenly and flatly laid on the surface of the wheelchair's seat cushion, backrest and legboard, and connected to the data acquisition system; S32: Guide the tester to sit in the wheelchair and adjust it to the preset angle; after the sitting posture and body pressure distribution are stable, the system automatically records the body pressure distribution data at this moment, and the tester completes the evaluation questionnaire based on the immediate subjective feeling; S33: Adjust the wheelchair to other preset angle combinations in sequence, and repeat step S32; S34: Associate the collected body pressure data with its corresponding questionnaire scores and work parameters to construct an objective body pressure distribution dataset D3.

[0013] Furthermore, in step S5, constructing the model training dataset specifically includes the following steps: S51: Using the subjective weights determined by the analytic hierarchy process, the scores of each item in the subjective evaluation dataset D1 are weighted and summed to calculate the comprehensive subjective comfort score, and the subjective evaluation dataset D2 is constructed. S52: Associate and align the subjective evaluation dataset D2 with the objective body pressure distribution dataset D3 to construct a combined subjective and objective dataset D; S53: Normalize the dataset D and assign comfort level labels; then divide the dataset D into training and testing sets; S54: Partition the body pressure distribution data matrix into three sub-matrices: calf region, buttocks and thigh region, and back region; extract the same pressure distribution statistical features from each sub-matrix to form a feature vector.

[0014] Furthermore, in S6, the trained random forest regression model uses the feature vector formed by the statistical features of the pressure distribution extracted from each partition submatrix obtained in S54 as input, and uses the comprehensive subjective comfort score of the corresponding working condition calculated in S51 as training label. The trained model can receive new body pressure distribution data, process it according to the same partitioning and feature extraction rules, and directly output an objective comfort prediction score corresponding to the current working condition.

[0015] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for evaluating the static comfort of an intelligent wheelchair based on random forest.

[0016] The technical effects and advantages of this invention are as follows: Compared to existing evaluation methods that rely solely on subjective user questionnaires, this invention adopts a comfort evaluation approach that moves from subjective qualitative assessment to a fusion of subjective and objective prediction. Its advantage lies not in the simple application of a single technology, but in constructing a complete, data-driven closed-loop system. This effectively overcomes the problems of large biases and instability in evaluation results caused by traditional methods that rely entirely on subjective questionnaires. By introducing objective body pressure distribution data and using a random forest regression algorithm to build a predictive model, it can automatically learn the intrinsic mapping relationship between complex body pressure characteristics and comfort levels. This ensures that the evaluation results are based on measurable physical data rather than instantaneous emotions, thus significantly improving the objectivity and repeatability of the output. Through subjective-objective consistency verification, it is ensured that the model's predictive results are statistically highly consistent with actual human sensations, proving the reliability of the method.

[0017] In guiding design optimization, this invention represents a shift from fuzzy, experience-based judgment to precise, data-driven approaches. Traditional questionnaires can only indicate "discomfort" but fail to pinpoint areas for improvement. This invention, however, refines the body pressure matrix by partitioning it into smaller sections, directly linking body pressure data for the seat, backrest, and legrest areas to perceived comfort levels in specific body parts (such as the lower back and hips). When the model identifies low comfort in a particular area, designers can precisely trace back the corresponding body pressure characteristics to pinpoint optimization goals, such as strengthening lumbar support or adjusting surface shapes, significantly improving the targeting and efficiency of design optimization.

[0018] Furthermore, this invention utilizes the Analytic Hierarchy Process (AHP) to scientifically determine the contribution weight of each body part's comfort to the overall experience, transforming previous rough estimates based on experience into quantifiable indicators. Through pairwise comparisons and consistency checks, AHP assigns precise weights to each body part, clearly defining its proportion in the overall score. This not only makes the calculation of overall comfort more reasonable but also enables designers to prioritize optimizing key parts that have a greater impact on overall comfort based on weighted data, thereby achieving optimal resource allocation and maximizing design benefits. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the process of the present invention.

[0020] Figure 2 This is a schematic diagram of adjusting the wheelchair to a preset angle according to the present invention.

[0021] Figure 3 This is a partition diagram of the body pressure data matrix of the present invention.

[0022] The attached diagram is labeled as follows: 1. Seat cushion; 2. Backrest; 3. Legrest; 31. Thigh rest; 32. Lower leg rest. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0024] As attached Figure 1 Appendix Figure 2 and attached Figure 3 The method for evaluating the static comfort of a smart wheelchair based on random forest, as shown, includes the following steps: S1: Establish a systematic evaluation index system for subjective comfort; obtain subjective comfort evaluation parameters for testers riding in wheelchairs; For example: S11: Establish a systematic evaluation index system for subjective comfort: Establish a static wheelchair subjective comfort evaluation table, which includes: the wheelchair support scheme to be evaluated and the subjective comfort evaluation corresponding to the wheelchair support scheme; the wheelchair support scheme clearly defines each support part and its adjustable angle parameters; the subjective comfort evaluation is used to quantify the subjective comfort evaluation of the tester under each wheelchair support scheme.

[0025] The wheelchair support system includes: a seat cushion 1 for supporting the buttocks; a backrest 2 for supporting the shoulders, back, and waist, with the angle between the backrest 2 and the horizontal plane of the seat cushion 1 set as α, ranging from 10° to 75°; and a leg rest 3, comprising a thigh plate 31 for supporting the thighs and a calf plate 32 for supporting the calves, with the angle between the thigh plate 31 and the horizontal plane of the seat cushion 1 set as β, ranging from 0° to 16°. The thigh plate 31 is rotatably connected to the calf plate 32 via a locking member, wherein the locking member is operable to switch between a locked state and a released state; when in the locked state, the relative rotation between the thigh plate 31 and the calf plate 32 is fixed; when in the released state, they can rotate relative to each other.

[0026] S12: Obtain subjective comfort evaluation parameters for testers riding in wheelchairs: Testers experience riding in wheelchairs with different angle combinations and score each option according to the evaluation table in S11; the system records all scoring results and initially constructs the subjective evaluation dataset D1 for wheelchair comfort.

[0027] The following table is used to establish a static wheelchair subjective comfort evaluation form:

[0028] Before using the wheelchair, the angles of various parts of the wheelchair were adjusted according to a predetermined wheelchair support scheme. After adjustment, the tester rode in the wheelchair and rated the wheelchair support scheme on a subjective comfort evaluation form. All wheelchair support schemes were recorded in the subjective comfort evaluation form, thus initially constructing the wheelchair comfort subjective evaluation dataset D1. The wheelchair support schemes were then configured according to the attached... Figure 2 Adjust the angle as shown.

[0029] The backrest 2 angle adjustment range is 10°–75°, with 14 settings in 5° increments (10°, 15°, ..., 75°); the thigh plate 31 angle adjustment range is 0°–16°, with 5 settings in 4° increments (0°, 4°, 8°, 12°, 16°). During data collection, the thigh plate 31 angle is initially fixed at 0°. Then, the backrest 2 angle is adjusted sequentially from 10° to 75°, recording 14 sets of data. Afterward, the thigh plate 31 angle is adjusted to 4°, and the backrest 2 angle adjustment process is repeated, and so on, to complete the data collection for all combinations. The above static wheelchair subjective comfort evaluation form can be manually filled in by the testers, and the data can be summarized and processed. The scoring criteria are as follows: each item is scored from 0 to 100. For example, 0-20 is considered very uncomfortable, 21-40 is considered uncomfortable, 41-60 is considered normal, 61-80 is considered comfortable, and 81-100 is considered very comfortable.

[0030] S2: Based on the evaluation index system established in S1, the subjective weights of each evaluation index are obtained using the analytic hierarchy process (AHP). Specifically, the six subdivided subjective perception characteristics of shoulder support, back support, waist support, hip support, thigh support, and calf support are defined using the AHP to assign weights based on the characteristics of the intelligent wheelchair. The specific methods for obtaining the subjective weights of evaluation indicators include the following steps: S21: Generate a test questionnaire. Test participants compare and assign values ​​to the relative importance of indicators at the same level using a 1-9 scale. Construct a judgment matrix for each test subject using the indicator values ​​in the test questionnaire. For example, as shown in the table below.

[0031] The target layer includes wheelchair comfort; the criteria layer includes upper limb comfort and lower limb comfort; the solution layer includes shoulder support, back support, lumbar support, hip support, thigh support, and calf support. Therefore, pairwise comparisons are made between different elements in each of the target, criteria, and solution layers, and the importance of each element is judged by a predefined judgment standard to construct a judgment matrix.

[0032] For example, when comparing the importance of the i-th element and the j-th element relative to a factor in the previous layer, a quantified relative importance 'a' can be used. ij To express.

[0033] Assuming there are n elements involved in the comparison, the judgment matrix is ​​represented as follows:

[0034] Matrix A is a positive-reverse judgment matrix; S22: Based on the judgment matrix of each participant in the test, obtain the corresponding indicator weight, eigenvector, and maximum eigenvalue; among them, the normalized eigenvector is the relative weight of each factor in the current layer.

[0035] S23: Perform a consistency check on the judgment matrix of each tester and calculate the consistency ratio:

[0036] First, calculate the consistency index (CI). Let CI be the largest eigenvalue of matrix A, and n be the order of the judgment matrix. CI = 0 indicates complete agreement, while the larger CI is, the less agreement there is.

[0037] The average random consistency index RI is obtained by taking the average of the eigenvalues ​​of the random judgment matrix after repeated calculations.

[0038] If the judgment matrix fails the consistency test, it means that the tester's judgment results are contradictory, and the tester is notified to conduct the questionnaire survey again; return to S21; If the judgment matrix passes the consistency test, then the group weights are calculated for the judgment matrices that meet the consistency test.

[0039] According to the formula:

[0040] If the CR is outside the set threshold range, it is determined that the matrix has failed the consistency test, which means that the tester's judgment is contradictory, and the tester is notified to conduct the questionnaire survey again.

[0041] The within-group weights are then calculated for the judgment matrix that meets the consistency test. First, the judgment matrix is ​​normalized column-wise. The specific calculation formula is as follows:

[0042] in, This represents the normalized value of the element in the i-th row and j-th column of the judgment matrix. is the value at the corresponding position in the original judgment matrix, and n is the order of the matrix, i.e., the total number of indicators.

[0043] Next, the normalized judgment matrix is ​​summed row by row, using the following formula:

[0044] In the formula, This represents the sum of all elements in the i-th row of the normalized matrix.

[0045] Then, for the above rows and vectors Normalization is performed to obtain the relative weight of each indicator within its respective criterion layer. The calculation formula is as follows:

[0046] Here, This refers to the local weight of the i-th indicator within the group.

[0047] Finally, the intra-group weights of all indicators are constructed into eigenvectors, denoted as Wintra-group = (W1, W2, W3, ..., Wn)T, which represents the intra-group weights of each indicator. Combined with the inter-group weight vector Winter-group determined by the importance comparison of the upper-level criteria, the subjective weight WI = Wintra-group × Winter-group is obtained through weighted calculation.

[0048] These six subdivided subjective perception characteristics are defined using the Analytic Hierarchy Process (AHP) with weights B1, B2, B3, C1, C2, and C3. One typical setting is B1 = 0.1002, B2 = 0.1998, B3 = 0.3000, C1 = 0.2284, C2 = 0.1144, C3 = 0.0572, and B1 + B2 + B3 + C1 + C2 + C3 = 1.

[0049] S3: Collect objective body pressure distribution data when the test subject is in a wheelchair; the specific process for collecting objective body pressure distribution data includes the following steps: S31: Place the wheelchair to be tested in the laboratory environment for a period of time; adjust the wheelchair to the predetermined position as required; evenly and flatly lay the pressure distribution test pad on the surface of the wheelchair's seat cushion 1, back panel 2 and leg panel 3, and connect it to the data acquisition system; start and debug the host computer software; S32: Guide the tester to sit in the wheelchair and adjust it to the preset angle; that is, adjust the wheelchair backrest 2, seat 1, and legrest 3 to the preset angle; and fill in the questionnaire according to the tester's own subjective feelings; after the sitting posture and body pressure distribution are stable, the system automatically records the body pressure distribution data at this moment, and the tester completes the evaluation questionnaire based on the immediate subjective feelings. S33: Adjust the wheelchair to other preset angle combinations in sequence, and repeat step S32; that is, the tester stays in the wheelchair, adjusts the wheelchair to different preset angles, and records the results in the evaluation form. S34: Associate the collected body pressure data with its corresponding questionnaire scores and work parameters to construct an objective body pressure distribution dataset D3.

[0050] S4: Based on the subjective weights obtained in S2, the testers calculate the overall subjective comfort level by combining the comfort scores of each part collected in S1. The specific calculation formula is as follows:

[0051] Wherein, B1 = 0.1002, B2 = 0.1998, B3 = 0.3000, C1 = 0.2284, C2 = 0.1144, C3 = 0.0572; F1 is the shoulder comfort score, F2 is the back comfort score, F3 is the lower back comfort score, F4 is the hip comfort score, F5 is the thigh comfort score, and F6 is the calf comfort score.

[0052] S5: Integrate and optimize the objective body pressure distribution data collected in S3, and combine it with the comprehensive subjective comfort obtained in S4 to construct a model training dataset; The specific steps for building the model training dataset are as follows: S51: Using the subjective weights determined by the analytic hierarchy process, the scores of each item in the subjective evaluation dataset D1 are weighted and summed to calculate the comprehensive subjective comfort score, thus constructing the subjective evaluation dataset D2; that is, the scores of each indicator in the subjective evaluation dataset D1 of the static comfort of the intelligent wheelchair are weighted and summed to calculate the comprehensive subjective comfort score, thus forming the final subjective evaluation dataset D2.

[0053] S52: Associate and align the subjective evaluation dataset D2 with the objective body pressure distribution dataset D3 to construct a subjective-objective fusion dataset D; that is, construct a subjective-objective experimental dataset D for wheelchair comfort for model training and validation.

[0054] S53: Normalize dataset D and assign comfort level labels; then divide dataset D into training and testing sets. For example, normalize dataset D to eliminate the influence of dimensions, and assign comfort level labels L based on the comprehensive comfort score, defined as: L = {1: very uncomfortable, 2: uncomfortable, 3: average, 4: comfortable, 5: very comfortable}. Divide the processed dataset into training and testing sets in an 8:2 ratio, respectively for training and performance testing of the subsequent random forest-based comfort evaluation model.

[0055] S54: The body pressure distribution data matrix is ​​partitioned into three sub-matrices: the calf region, the buttocks and thigh region, and the back region. The same pressure distribution statistical features are extracted from each sub-matrix to form feature vectors. Specifically, to improve the precision of feature extraction and avoid extracting features too coarsely from the entire body pressure matrix while ignoring local pressure distribution details, the body pressure data matrix is ​​partitioned. Based on the body part division scheme in the subjective questionnaire, the body pressure distribution data for seat cushion 1, backrest 2, and legrest 3 are divided into three sub-regions. (See attached image for example.) Figure 3 As shown, the partitioning scheme is as follows: Partition 1 corresponds to the entire lower leg area, Partition 2 corresponds to the seat cushion contact area formed by the buttocks and thighs, and Partition 3 corresponds to the back area. The same features are extracted from each sub-matrix for comparative analysis, thereby more precisely characterizing the impact of wheelchair parameter adjustments on the user's local and overall comfort. S6: Using the feature data in the model training dataset built in S5 as input and the corresponding comprehensive subjective comfort quantification value as the output target value, train the random forest regression model to obtain an intelligent model that can predict comfort based on body pressure characteristics. The random forest algorithm was used to train the prediction model, resulting in a trained model that can calculate objective comfort based on body pressure distribution data. The trained random forest regression model uses the feature vector formed by the statistical features of the stress distribution extracted from each partition submatrix obtained in S54 as input, and the comprehensive subjective comfort score of the corresponding working condition calculated in S51 as training label. The trained model can receive new body pressure distribution data, process it according to the same partitioning and feature extraction rules, and directly output an objective comfort prediction score corresponding to the current working condition.

[0056] For example, based on features extracted from body pressure distribution data, a random forest regression model was built using Python to predict objective comfort scores for six body parts, with a total of k1*k2 samples. This model uses the CART algorithm as the base learner to construct the decision regression tree and generates a random forest containing multiple decision trees through the Bagging ensemble learning framework. The core process is as follows: multiple training subsets are generated by sampling with replacement from the original training set, and multiple decision trees are trained separately. Finally, the prediction results are integrated using the averaging method. In terms of hyperparameter settings, the partitioning criterion is the mean squared error, and due to the limited number of features, each tree is constructed using all features. After training and evaluation testing, a trained adjustable intelligent wheelchair comfort evaluation model is obtained. S7: Compare and analyze the objective comfort level output by the trained model described in S6 with the comprehensive subjective comfort level calculated in S4 to verify the consistency between the objective and subjective evaluation results and complete the model's effectiveness assessment. Specifically, compare the consistency between the random forest calculation results and the wheelchair's subjective comfort level. By verifying the consistency between the intelligent wheelchair's subjective comfort level and objective evaluation indicators, the accuracy of the model's calculation results is verified.

[0057] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for evaluating the static comfort of an intelligent wheelchair based on random forest.

[0058] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the static comfort of intelligent wheelchairs based on random forest, characterized in that, Includes the following steps: S1: Establish a systematic evaluation index system for subjective comfort; obtain subjective comfort evaluation parameters for testers riding in wheelchairs; S2: Based on the evaluation index system established in S1, the subjective weights of each evaluation index are obtained using the analytic hierarchy process (AHP). S3: Collect objective body pressure distribution data when the tester is in a wheelchair; S4: Based on the subjective weights obtained in S2, the testers calculate the overall subjective comfort level by combining the comfort scores of each part collected in S1. S5: Integrate and optimize the objective body pressure distribution data collected in S3, and combine it with the comprehensive subjective comfort obtained in S4 to construct a model training dataset; S6: Using the feature data in the model training dataset constructed in S5 as input and the corresponding comprehensive subjective comfort quantification value as the output target value, train the random forest regression model to obtain an intelligent model that can predict comfort based on body pressure characteristics; use the random forest algorithm to train the prediction model to obtain a trained model that can calculate objective comfort based on body pressure distribution data. S7: Compare and analyze the objective comfort level output by the trained model described in S6 with the comprehensive subjective comfort level calculated in S4 to verify the consistency between the objective and subjective evaluation results and complete the effectiveness evaluation of the model.

2. The method for evaluating the static comfort of an intelligent wheelchair based on random forest according to claim 1, characterized in that: In S1, S11: Establish a systematic evaluation index system for subjective comfort: Establish a static wheelchair subjective comfort evaluation table, which includes: the wheelchair support scheme to be evaluated and the subjective comfort evaluation corresponding to the wheelchair support scheme; the wheelchair support scheme clearly defines each support part and its adjustable angle parameters; the subjective comfort evaluation is used to quantify the subjective comfort evaluation of the tester under each wheelchair support scheme. S12: Obtain subjective comfort evaluation parameters for testers riding in wheelchairs: Testers experience riding in wheelchairs with different angle combinations and score each option according to the evaluation table in S11; the system records all scoring results and initially constructs the subjective evaluation dataset D1 for wheelchair comfort.

3. The method for evaluating the static comfort of an intelligent wheelchair based on random forest according to claim 2, characterized in that: In S11, the wheelchair support solution includes: A seat cushion designed to support the buttocks; A backrest, used to support the shoulders, back, and waist, with an angle α between the backrest and the horizontal plane of the seat cushion, the angle ranging from 10° to 75°; and, The leg plate includes a thigh plate for supporting the thigh and a calf plate for supporting the lower leg. The angle between the thigh plate and the horizontal plane of the seat cushion is set to β, and the range of β is set to 0°-16°. The thigh plate is rotatably connected to the calf plate by a locking member, wherein the locking member is operable to switch between a locked state and a released state. When in the locked state, the relative rotation between the thigh plate and the calf plate is fixed; when in the released state, the two can rotate relative to each other.

4. The method for evaluating the static comfort of an intelligent wheelchair based on random forest according to claim 3, characterized in that: In step S2, the method for obtaining the subjective weights of the evaluation indicators includes the following steps: S21: Generate a test questionnaire. Test participants compare and assign values ​​to the relative importance of indicators at the same level using a 1-9 scale. Construct a judgment matrix for each test subject using the indicator values ​​in the test questionnaire. S22: Based on the judgment matrix of each participant in the test, obtain the corresponding indicator weight, eigenvector, and maximum eigenvalue; S23: Perform a consistency check on the judgment matrix of each tester and calculate the consistency ratio: If the judgment matrix fails the consistency test, it indicates that there is a contradiction in the tester's judgment results. The tester should be notified to conduct the questionnaire survey again; return to S21. If the judgment matrix passes the consistency test, then the group weights are calculated for the judgment matrices that meet the consistency test.

5. The method for evaluating the static comfort of an intelligent wheelchair based on random forest according to claim 4, characterized in that: In step S3, the process of collecting objective body pressure distribution data... The steps include: S31: The pressure distribution test pad is evenly and flatly laid on the surface of the wheelchair's seat, backrest, and legboard, and then connected to the data acquisition system; S32: Guide the tester to sit in the wheelchair and adjust it to the preset angle; after the sitting posture and body pressure distribution are stable, the system automatically records the body pressure distribution data at this moment, and the tester completes the evaluation questionnaire based on the immediate subjective feeling; S33: Adjust the wheelchair to other preset angle combinations in sequence, and repeat step S32; S34: Associate the collected body pressure data with its corresponding questionnaire scores and work parameters to construct an objective body pressure distribution dataset D3.

6. The method for evaluating the static comfort of an intelligent wheelchair based on random forest according to claim 5, characterized in that: In step S5, constructing the model training dataset specifically includes the following steps: S51: Using the subjective weights determined by the analytic hierarchy process, the scores of each item in the subjective evaluation dataset D1 are weighted and summed to calculate the comprehensive subjective comfort score, and the subjective evaluation dataset D2 is constructed. S52: Associate and align the subjective evaluation dataset D2 with the objective body pressure distribution dataset D3 to construct a combined subjective and objective dataset D; S53: Normalize the dataset D and assign comfort level labels; then divide the dataset D into training and testing sets; S54: Partition the body pressure distribution data matrix into three sub-matrices: calf region, buttocks and thigh region, and back region; extract the same pressure distribution statistical features from each sub-matrix to form a feature vector.

7. The method for evaluating the static comfort of an intelligent wheelchair based on random forest according to claim 6, characterized in that: In S6, the trained random forest regression model takes the feature vector formed by the statistical features of the pressure distribution extracted from each partition submatrix obtained in S54 as input, and the comprehensive subjective comfort score of the corresponding working condition calculated in S51 as training label. The trained model can receive new body pressure distribution data, process it according to the same partitioning and feature extraction rules, and directly output an objective comfort prediction score corresponding to the current working condition.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images