Intelligent office table height automatic adjustment and ergonomic optimization method

Abstract

The invention relates to the technical field of intelligent office, and discloses an intelligent office table height automatic adjustment and ergonomic optimization method, which comprises the following steps: S1, basic data information acquisition: acquiring body data of a user, S2, original data processing and analysis, and S3, intelligent office table height automatic adjustment and ergonomic optimization. The data is analyzed and processed through a preset algorithm, S3, mechanical action instruction conversion, S4, mechanical action execution, and S5, action information feedback adjustment. According to the method, the adjustment direction and amplitude of the height of the office table are determined through the established comparison algorithm of deep learning in combination with the adjustment decision algorithm based on fuzzy logic, so that the height of the office table is dynamically adjusted to better conform to the ergonomic standard, the comfort and health level of a user are improved, and the user experience is improved. And meanwhile, control can be realized through a switchable adjusting mode according to reality. According to actual working scenes and personal requirements of users, a personalized height adjusting scheme conforming to ergonomics is provided, and the flexibility and adaptability of the system are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent office, and specifically to a method for automatically adjusting the height of an intelligent desk and optimizing ergonomics. Background Art

[0002] In recent years, significant progress has been made in sensor technology, data analysis algorithms, and ergonomic research. High-precision pressure sensors, infrared sensors, and image recognition technology have been widely applied to various intelligent devices, providing strong technical support for the research and development of intelligent desks. At the same time, the continuously improved ergonomic theory provides a scientific basis for determining the optimal office postures and desktop heights for people with different body characteristics in various working scenarios.

[0003] However, in traditional office environments, the height of desks is usually fixed. This fixed-height design fails to fully consider the significant differences in height, body shape, etc. among different users. For example, when taller users work at a desk for a long time, they often need to bend over and hunch their backs excessively to accommodate the lower desktop height. This not only causes excessive pressure on the neck and waist, easily leading to muscle strain, but may even cause spinal diseases in the long run. Shorter users may face the problem of a too-high desktop and need to keep their arms raised for a long time to operate, which also causes fatigue in the shoulder and arm muscles. According to relevant medical research, more than 60% of office workers have varying degrees of physical discomfort symptoms due to long-term use of fixed-height desks. In view of this, we propose a method for automatically adjusting the height of an intelligent desk and optimizing ergonomics. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a method for automatically adjusting the height of an intelligent desk and optimizing ergonomics, which solves the problem that in traditional office environments, the height of desks is usually fixed, and this fixed-height design fails to fully consider the significant differences in height, body shape, etc. among different users.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for automatically adjusting the height of an intelligent desk and optimizing ergonomics, including the following steps:

[0006] S1; Collection of basic data information

[0007] Use the data collection module set on the desk to collect the body data of the user, and the body data at least includes weight, height, and body posture information;

[0008] S2; Processing and analysis of raw data

[0009] Transmit the collected body data to the data analysis and processing module, and analyze and process the data through a preset multiple linear regression model and a bone key point detection algorithm;

[0010] S3; Mechanical action instruction conversion

[0011] Based on the analysis results, the data analysis and processing module combines an adjustment decision algorithm based on fuzzy logic to generate an instruction for the height adjustment execution module. This algorithm takes factors such as the degree of body posture deviation and working duration as fuzzy input quantities, and determines the adjustment direction and amplitude of the desk height through the processes of fuzzification, fuzzy inference, and defuzzification;

[0012] S4; Mechanical action execution

[0013] The electric lifting device in the height adjustment execution module adopts a high-precision lead screw and nut transmission structure. It receives the instruction sent by the data analysis and processing module through the motor controller, and drives the electric lifting device to accurately adjust the desk height;

[0014] S5; Action information feedback adjustment

[0015] Based on the ergonomics database, using a comparison algorithm based on deep learning, compare the real-time data of the user with the standards in the database to dynamically adjust the desk height.

[0016] Preferably, in the S1 basic data information collection, the data collection module includes a pressure sensor, an infrared sensor, and a camera. The pressure sensor is used to detect the pressure exerted by the user on the seat, the infrared sensor is used to measure the distance between the user and the desk, and the camera is used to capture the body posture information of the user.

[0017] Preferably, the preset algorithm in the S2 raw data processing and analysis calculates the user's weight based on the pressure detected by the pressure sensor, combines the distance measured by the infrared sensor to assist in judging the user's height, and at the same time analyzes the user's body posture according to the image captured by the camera, including the position and angle information of the head, shoulders, and waist. Among them, the bone key point detection algorithm uses a bone key point detection algorithm based on a graph neural network to improve the accuracy of posture analysis.

[0018] Preferably, in the S4 mechanical action, the electric lifting device in the height adjustment execution module adopts a high-precision lead screw and nut transmission structure. It receives the instruction sent by the data analysis and processing module through the motor controller, and drives the electric lifting device to accurately adjust the desk height.

[0019] Preferably, in the S5 action information feedback adjustment, the established ergonomics database stores the optimal desk height parameters for people of different heights and weights in different working scenarios, as well as the corresponding body posture standards.

[0020] Preferably, in the conversion of the S3 mechanical action instruction, the data analysis and processing module compares the real-time data of the user in the S5 action information feedback adjustment with the database standard to determine the adjustment plan for the desk height. Among them, the comparison algorithm based on deep learning optimizes the comparison accuracy by training the network model.

[0021] Preferably, in the S1 basic data information collection, when the user uses it for the first time, the data collection module collects the basic body information of the user. The data analysis and processing module calculates the initial desk height suitable for the user according to this information combined with the ergonomic database, and controls the height adjustment execution module to adjust the desk to this height.

[0022] Preferably, in the S1 basic data information collection, after the data collection module obtains the current user information through the pressure sensor, it continuously and real-time collects the body data of the current user and transmits it to the data analysis and processing module. When the data analysis and processing module determines that the user is in a bad posture or needs to adjust the working state, it sends a corresponding height adjustment instruction to the height adjustment execution module.

[0023] Preferably, in the S2 raw data processing and analysis, when the data analysis and processing module analyzes the body posture data, when it detects that the user continuously bows his head for more than the preset time and the angle between the head and the shoulder is less than the preset angle, it determines that the user is in a bad posture state and triggers the height adjustment execution module to adjust the desk height.

[0024] Preferably, in the S5 action information feedback adjustment, a data self-check function is synchronously established. When it detects that the desk height adjustment is abnormal or the user's body data is abnormal, the data self-check function will be automatically triggered to comprehensively check the data collection module, the data analysis and processing module, and the height adjustment execution module to ensure the stability and accuracy of the system operation.

[0025] The present invention provides an intelligent desk height automatic adjustment and ergonomic optimization method, which has the following beneficial effects:

[0026] 1. Through the established comparison algorithm based on deep learning and combined with the adjustment decision algorithm based on fuzzy logic, the present invention determines the adjustment direction and amplitude of the desk height, realizes the dynamic adjustment of the desk height, so as to better meet the ergonomic standards, improve the comfort and health level of the user. At the same time, through the controllable adjustment mode switch, it can provide a personalized and ergonomic height adjustment plan according to the actual working scenario and the personal needs of the user, enhancing the flexibility and adaptability of the system.

[0027] 2. The present invention establishes a bone key point detection algorithm. Compared with the traditional CNN-based bone key point detection algorithm, the GNN-based algorithm can better capture the spatial relationship and context information between key points, thereby improving the detection accuracy and robustness. Especially when dealing with complex postures and occlusion situations, GNN can use the information of the graph structure for reasoning and completion, showing better performance.

[0028] 3. The present invention establishes a collaborative computing verification mechanism. The statistical anomaly detection algorithm is good at discovering anomalies that deviate greatly from the overall statistical characteristics of the data, while the machine learning-based algorithm can learn complex patterns and feature combinations in the data and has a better detection effect on those situations that are difficult to detect from a simple statistical perspective but show anomalies in the feature space. The combination of the two can cover a wider range of anomaly types and can form a mutual verification mechanism to improve the comprehensiveness and accuracy of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a flowchart of the method for automatically adjusting the height of the intelligent desk and optimizing human engineering;

[0030] Figure 2 is a flowchart of the comparison algorithm based on deep learning in the adjustment process of the present invention;

[0031] Figure 3 is a flowchart of the bone key point detection algorithm based on graph neural network of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the specification of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0033] Embodiment:

[0034] Please refer to the attached Figure 1 - attached Figure 3 , the embodiment of the present invention provides a method for automatically adjusting the height of an intelligent desk and optimizing human engineering, including the following steps:

[0035] S1; Basic data information collection

[0036] Use the data collection module set on the desk to collect the body data of the user, and the body data at least includes weight, height, and body posture information;

[0037] S2; Original data processing and analysis

[0038] Transmit the collected body data to the data analysis and processing module, and analyze and process the data through a preset multiple linear regression model and a bone key point detection algorithm;

[0039] S3; Mechanical action instruction conversion

[0040] Based on the analysis results, the data analysis and processing module generates an instruction for the height adjustment execution module in combination with an adjustment decision algorithm based on fuzzy logic. This algorithm takes factors such as the degree of body posture deviation and working duration as fuzzy input quantities, and determines the adjustment direction and amplitude of the desk height through the processes of fuzzification, fuzzy inference, and defuzzification;

[0041] S4; Mechanical action execution

[0042] The electric lifting device in the height adjustment execution module adopts a high-precision screw-nut transmission structure, receives the instruction sent by the data analysis and processing module through the motor controller, and drives the electric lifting device to accurately adjust the desk height;

[0043] S5; Action information feedback adjustment

[0044] Based on the ergonomics database, use a comparison algorithm based on deep learning to compare the real-time data of the user with the standards in the database, and dynamically adjust the desk height.

[0045] In the S1 basic data information collection, the data collection module includes a pressure sensor, an infrared sensor, and a camera. The pressure sensor is used to detect the pressure exerted by the user on the seat, the infrared sensor is used to measure the distance between the user and the desk, and the camera is used to capture the body posture information of the user.

[0046] The preset algorithm in the S2 raw data processing and analysis estimates the user's weight based on the pressure detected by the pressure sensor, combines the distance measured by the infrared sensor to assist in judging the user's height, and at the same time analyzes the user's body posture based on the image captured by the camera, including the position and angle information of the head, shoulders, and waist. Among them, the bone key point detection algorithm is a bone key point detection algorithm based on a graph neural network, which improves the accuracy of posture analysis. In the detection task, each bone key point can be regarded as a node in the graph, and the connection relationship between key points (such as the connection between joints) constitutes the edge of the graph. In common human pose estimation tasks, key points such as the head, shoulders, elbows, wrists, hips, knees, and ankles are selected as nodes. Each node can be represented by its corresponding feature vector, and these features can be local features extracted from the image, such as the convolutional features of the image patches around the key points, and the following algorithm is established;

[0047] Define edges according to the physiological structure and kinematic relationship of the human skeleton. For example, the edge connecting the left and right shoulders, the edge between the shoulder and the elbow, the edge between the elbow and the wrist, etc. The existence of an edge indicates that there is a certain association between two key points, and this association can be represented by an adjacency matrix. The adjacency matrix A is an N×N matrix, where N is the number of nodes. If there is an edge between node i and node j, then A ij = 1; otherwise, A ij = 0. To consider the information of the nodes themselves, self-loops are usually added to the adjacency matrix, that is, A ij = 1, to obtain where I is the identity matrix

[0048] Graph neural network layer design:

[0049] The GCN layer updates the feature representation of nodes by aggregating the features of nodes and their neighbor nodes. For the l-th layer of the GCN layer, the update formula for node features is as follows:

[0050]

[0051] where, H (l) is the node feature matrix of the l-th layer, with size N×C (l) , N is the number of nodes, and C (l) is the feature dimension of the l-th layer; W (l) is a learnable weight matrix, with size C (l) ×C (l+1) ; is 's degree matrix, which is a diagonal matrix, and its diagonal element σ is the activation function, and the commonly used one is the ReLU function;

[0052] This step is to normalize the adjacency matrix, aiming to make the feature update of each node take into account the balance of the number of neighbor nodes and the feature contribution. H (1) W (l) is a linear transformation of the node features, mapping the features from the C (l) dimension to the C (l+1) dimension. Finally, non-linear factors are introduced through the activation function to enhance the expressive power of the model;

[0053] Graph Attention Network (GAT) layer

[0054] When the GAT layer aggregates the features of neighbor nodes, it introduces an attention mechanism that can adaptively assign different weights to different neighbor nodes. For node i, its attention coefficient α ij represents the degree of attention of node i to neighbor node j, and the calculation formula is as follows:

[0055]

[0056] Among them, h i and h j are the feature vectors of node i and node j respectively, W is a learnable weight matrix, and a is a learnable attention vector is the set of neighbor nodes of node i, and ∥ represents the vector concatenation operation

[0057] operation. The updated feature h′ of node i i is as follows:

[0058] The attention coefficient α ij reflects the importance of node i to neighbor node j. Through the attention mechanism, the model can pay more attention to the neighbor nodes that are more closely related to the current node, so as to better capture the complex relationships between nodes;

[0059] Skeleton keypoint detection model architecture based on GNN

[0060] Input layer: The input is a human body image or a feature map extracted from the image. Usually, a convolutional neural network (CNN) is used to preprocess the image and extract the local features around each keypoint as the initial features of the graph nodes;

[0061] GNN layer: It is stacked by multiple GCN layers or GAT layers. Through these layers, the features of the nodes are continuously updated, gradually integrating more global information and context information;

[0062] Output layer: The node features output by the last GNN layer are passed through a fully connected layer to output the coordinates of each keypoint. For example, if K keypoints are to be detected and each keypoint has two coordinate values x and y, the dimension of the output layer is 2K;

[0063] Training process

[0064] Collect a large number of image datasets containing human postures and annotate the skeleton keypoints in each image. Input the image data into the CNN to extract the initial node features, and at the same time construct the adjacency matrix of the graph according to the connection relationship of the keypoints;

[0065] Loss function: Establish a loss function to measure the difference between the predicted keypoint coordinates and the true coordinates. The formula is as follows:

[0066]

[0067] Among them, K is the number of keypoints, and are the x and y coordinates of the predicted i-th keypoint, and are the x and y coordinates of the real i-th key point;

[0068] Optimizer: Use optimizers (such as Stochastic Gradient Descent SGD, Adam, etc.) to minimize the loss function. During training, continuously adjust the weight parameters of the GNN layer and the fully connected layer so that the predicted key point coordinates gradually approach the real coordinates.

[0069] In the S4 mechanical action, the electric lifting device in the height adjustment execution module adopts a high-precision screw-nut transmission structure. It receives the instruction sent by the data analysis and processing module through the motor controller and drives the electric lifting device to precisely adjust the height of the desk.

[0070] In the S5 action information feedback adjustment, the established ergonomic database stores the optimal desk height parameters for people of different heights and weights in different working scenarios, as well as the corresponding body posture standards.

[0071] In the S3 mechanical action instruction conversion, the data analysis and processing module compares the real-time data of the user in the S5 action information feedback adjustment with the database standard, determines the adjustment plan for the desk height, and can switch the real-time adjustment mode or the custom adjustment mode of the intelligent desk according to the actual use requirements. The adjustment process is based on a deep learning-based comparison algorithm to train the network model and continuously optimize the accuracy of the comparison. The following algorithm is established during the process:

[0072] I. Model Architecture

[0073] Mainly adopt a convolutional neural network architecture, and its structure design is specifically for the feature extraction and comparison tasks of ergonomics-related data. The network contains multiple convolutional layers, pooling layers, and fully connected layers, and each layer cooperates with each other to gradually extract high-level features from the original data.

[0074] 1. Convolutional layer: Perform convolution operations by sliding different-sized convolutional kernels on the input data. Taking two-dimensional convolution as an example, for the input feature map I and the convolutional kernel K, the calculation method of the output feature map O is:

[0075]

[0076] where (i,j) are the coordinates of the output feature map, and M and N are the height and width of the convolutional kernel respectively;

[0077] For example, a 3×3 or 5×5 convolutional kernel is used to perform convolution on the images or feature vectors corresponding to the user's real-time body data and the standard data in the ergonomics database. These convolutional kernels can learn local features in the data, such as specific angles of body postures, relative positional relationships between joints, etc. Each convolutional layer generates multiple feature maps, and each feature map represents the response of the input data under a specific convolutional kernel. In practical applications, to increase the nonlinear representation ability of the model, an activation function, such as the ReLU function: f(x) = max(0, x), is added after the convolution operation.

[0078] 2. Pooling layer: The pooling layer follows the convolutional layer immediately and adopts max pooling or average pooling operations. Taking max pooling as an example, it selects the maximum value within a fixed-size window as the output. For the input feature map I, under a 2×2 pooling window, the calculation method of the output feature map O is as follows:

[0079] O(i,j) = max{I(2i,2j),I(2i,2j + 1),I(2i + 1,2j),I(2i + 1,2j + 1)}

[0080] where (i,j) are the coordinates of the output feature map. The role of the pooling layer is to reduce the dimension of the data, reduce the computational amount, and at the same time retain the main features of the data, making the model more robust to changes such as translation and rotation of the data. Average pooling calculates the average value within the pooling window as the output, and the calculation formula is:

[0081]

[0082] 3. Fully connected layer: After passing through multiple convolutional layers and pooling layers, the data is transformed into a one-dimensional vector and input into the fully connected layer. The neurons in the fully connected layer are connected to all neurons in the previous layer. A linear transformation is performed on the input vector x through the weight matrix W, and the calculation formula for the output vector y is

[0083] y = Wx + b

[0084] where b is the bias vector. Then, an activation function (such as the ReLU function: f(x) = max(0, x)) is used to introduce non-linearity, further combine and abstract the features, and finally output the feature vector for comparison. In the fully connected layer, the number of parameters of the weight matrix W is large, and a large amount of data is required for training to determine its optimal value;

[0085] II. Training process

[0086] 1. Data preparation:

[0087] Collect a large amount of standard body posture data of people with different heights and weights in various working scenarios and the corresponding optimal desk height parameters from the ergonomics database. These data include human posture images captured by cameras, data collected by pressure sensors and infrared sensors, etc., and organize them into training sets and test sets. To increase the diversity and generalization of the data, data augmentation operations are usually performed on the data, such as randomly rotating, flipping, cropping, and adjusting the brightness of the image data. For example, for image I, the transformation formula for randomly rotating by an angle θ is:

[0088] I ′ (x ′ ,y ′ ) = I(xcosθ - ysinθ, xsinθ + ycosθ)

[0089] where (x, y) are the original image coordinates, and (x ′ ,y ′ ) are the coordinates of the rotated image;

[0090] For the data collected in real time by the user, preprocessing is also performed to make its format and dimension consistent with the data in the database for subsequent comparison. This includes normalizing the sensor data and mapping it to the range of [0, 1] or [-1, 1] to accelerate the convergence speed of the model. For example, for the sensor data s, the normalization formula is:

[0091]

[0092] 2. Selection of loss function and optimizer:

[0093] Use the cross-entropy loss function or the mean squared error loss function to measure the difference between the model prediction result and the true label. For binary classification problems, the calculation formula of the cross-entropy loss function is:

[0094]

[0095] where N is the number of samples, y i is the true label (0 or 1), and p i is the probability value predicted by the model;

[0096] For regression problems, the calculation formula of the mean squared error loss function is:

[0097]

[0098] where y i is the true value, is the value predicted by the model.

[0099] Select appropriate optimizers, including Stochastic Gradient Descent (SGD), Adagrad, Adadelta, and Adam. Taking the Adam optimizer as an example, it combines the advantages of the Adagrad and RMSProp algorithms and can adaptively adjust the learning rate. When updating the parameters using the Adam optimizer, first calculate the first moment estimate m t and the second moment estimate v t :

[0100] m t = β1m t-1 + (1 - β1)g t

[0101]

[0102] where g t is the gradient at the current moment, and β1 and β2 are decay coefficients, usually set to 0.9 and 0.999. Then, perform bias correction on the first moment estimate and the second moment estimate:

[0103]

[0104] Finally, update the parameter θ t :

[0105]

[0106] where α is the learning rate, and ∈ is a very small constant to prevent the denominator from being zero;

[0107] 3. Training Iterations:

[0108] Input the training data into the CNN model in batches. Each time a batch of data is input, it is called a mini - batch. On each mini - batch, the model performs forward propagation to calculate the prediction results, and then calculates the error between the prediction results and the true labels through the loss function. During the forward propagation process, the data passes through the convolutional layer, pooling layer, and fully - connected layer in sequence, and calculations are performed according to the calculation formulas of each layer above.

[0109] Based on the error, calculate the gradient of each parameter through the backpropagation algorithm, and the optimizer updates the weight parameters of the model according to the gradient. The backpropagation algorithm uses the chain rule, starting from the loss function, calculates the gradient layer by layer and propagates it backward to the previous layers. For example, for the convolutional layer, it is necessary to calculate the gradient of the convolutional kernel and the gradient of the input of the previous layer to update the parameters of the convolutional kernel. After a large number of training iterations, continuously adjust the parameters of the model to gradually reduce the value of the loss function and continuously improve the prediction ability of the model.

[0110] During the training process, the model is regularly evaluated using the test set, and metrics such as the accuracy, recall, and F1 value of the model are monitored to prevent overfitting or underfitting of the model.

[0111] III. Comparison Process

[0112] 1. Feature Extraction: When the user's real-time data is input into the trained model, the model first extracts features from the data through the convolutional layer and the pooling layer, transforming the original body posture image or sensor data into a high-dimensional feature vector. The key information of the user's body posture includes the tilt angle of the head, the position of the shoulders, and the degree of curvature of the waist. During the feature extraction process, the model performs feature mapping on the input data through the learned convolutional kernels and weight parameters, transforming the low-level original data features into high-level abstract features;

[0113] 2. Similarity Calculation: At the same time, the feature vectors of the corresponding standard data are extracted from the ergonomics database. Then, the matching degree between the real-time data feature vector and the standard data feature vector is judged by calculating the similarity between them. Commonly used similarity calculation methods include Euclidean distance, cosine similarity, etc. Euclidean distance: The formula for calculating the Euclidean distance between two feature vectors A and B is:

[0114]

[0115] where n is the dimension of the feature vector A i and B i are the i-th elements of vectors A and B respectively. The smaller the Euclidean distance, the more similar the two vectors are. To further optimize the similarity calculation, the Mahalanobis distance can also be introduced, which takes into account the covariance information of the data, and the calculation formula is:

[0116]

[0117] where ∑ is the covariance matrix of the data;

[0118] The similarity formula is:

[0119]

[0120] where A·B is the dot product of vectors A and B, and ||A|| and ||B|| are the norms of vectors A and B respectively;

[0121] The closer the value of the cosine similarity is to 1, the more similar the directions of the two vectors are, that is, the more matching the data features are. In addition, the Pearson correlation coefficient can also be used to calculate the similarity, which measures the linear correlation degree between two variables. For feature vectors A and B, the calculation formula of the Pearson correlation coefficient is:

[0122]

[0123] wherein and are the means of vectors A and B respectively.

[0124] 3. Decision-making and adjustment: According to the similarity calculation result, set a threshold. If the similarity exceeds the threshold, it is considered that the user's real-time posture and the current desk height match the ergonomic standard; if the similarity is lower than the threshold, trigger the desk height adjustment mechanism. According to the degree of deviation of the similarity from the threshold and other relevant factors (such as working hours, body posture change trend, etc.), combined with the adjustment decision algorithm based on fuzzy logic, determine the adjustment direction and amplitude of the desk height, and realize the dynamic adjustment of the desk height to better meet the ergonomic standard and improve the user's comfort and health level. In the fuzzy logic adjustment decision algorithm, factors such as the degree of deviation of the similarity and working hours are used as fuzzy input quantities, which are transformed into membership degrees in the fuzzy set through the membership function, then reasoning is carried out according to the preset fuzzy rules, and finally the specific adjustment direction and amplitude are obtained through the defuzzification method.

[0125] In the S1 basic data information collection, when the user uses it for the first time, the data collection module collects the user's basic body information, and the data analysis and processing module calculates the initial desk height suitable for the user according to this information combined with the ergonomic database, and controls the height adjustment execution module to adjust the desk to this height.

[0126] In the S1 basic data information collection, after the pressure sensor obtains the current user information, the data collection module continuously and real-time collects the user's body data and transmits it to the data analysis and processing module. When the data analysis and processing module determines that the user is in a bad posture or needs to adjust the working state, it sends a corresponding height adjustment instruction to the height adjustment execution module.

[0127] In the S2 raw data processing and analysis, when the data analysis and processing module analyzes the body posture data, when it detects that the user continuously bows the head for more than the preset time and the angle between the head and the shoulder is less than the preset angle, it determines that the user is in a bad posture state and triggers the height adjustment execution module to adjust the desk height.

[0128] In the S5 action information feedback adjustment, a data self-check function is synchronously established. When it detects that the desk height adjustment is abnormal or the user's body data is abnormal, the data self-check function will be automatically triggered to comprehensively check the data collection module, the data analysis and processing module, and the height adjustment execution module to ensure the stability and accuracy of the system operation. For the abnormal detection process, two groups of algorithms are established to perform collaborative calculations at the same time, including the following algorithms:

[0129] Mean and standard deviation method: For a series of collected desk height data h1, h2, …, h n or user body data d1, d2, …, d n , first calculate their mean μ and standard deviation σ

[0130] Mean formula:

[0131]

[0132] where x i represents the height data h i or body data d i

[0133] Standard deviation formula:

[0134]

[0135] If a data point x j satisfies |x j - μ| > kσ, then this data point can be considered an outlier, where k is an empirical threshold, usually taken as 2 or 3.

[0136] Anomaly detection algorithm based on machine learning

[0137] Isolation Forest algorithm: This algorithm is based on a binary tree structure to isolate data points. For a given dataset D, construct multiple isolation trees T1, T2, …, T m For a data point x to be detected, calculate its path length h(x) in each tree, and then calculate the average path length E(h(x)). The formula for the anomaly score s(x) is where c(n) is a constant related to the sample size n. If s(x) is closer to 1, then x is more likely to be an outlier; if s(x) is closer to 0, then x is more normal.

[0138] Collaborative decision-making

[0139] Fusion result: Combine the anomaly detection results based on statistics and those based on machine learning. A simple method is to adopt a voting mechanism. For example, if both methods determine a data point as an outlier, then it is determined as an outlier; if only one method determines it as an outlier, then different weights are assigned according to the reliability of the two methods for comprehensive judgment;

[0140] Dynamic weight adjustment: The weights can be dynamically adjusted according to the performance of the two algorithms on historical data. For example, calculate the true positives TP, false positives FP, and false negatives FN detected by each algorithm in the past period of time, and then calculate the accuracy and recall rate Evaluation metrics such as these are used to dynamically adjust the weights, so that algorithms with better performance have a greater say in the decision-making process.

[0141] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for automatic height adjustment and ergonomic optimization of a smart office desk, characterized in that: The following steps are involved: S1; Basic data information collection Using a data collection module disposed on the desk to collect the user's physical data, the physical data at least includes weight, height and body posture information; S2: Raw data processing and analysis The collected body data is transmitted to the data analysis and processing module, and the data is analyzed and processed by a preset multivariate linear regression model and a bone key point detection algorithm; S3; Mechanical action command conversion The data analysis and processing module generates instructions for the height adjustment execution module based on the analysis results and the adjustment decision algorithm based on fuzzy logic. The algorithm takes factors such as the degree of body posture deviation and working hours as fuzzy inputs, and determines the adjustment direction and amplitude of the desk height through the process of fuzzification, fuzzy reasoning and defuzzification. S4; Mechanical action execution The electric lifting device in the height adjustment execution module adopts a high-precision screw nut transmission structure. The motor controller receives the instructions sent by the data analysis and processing module to drive the electric lifting device to accurately adjust the height of the office desk. S5; Action information feedback adjustment Based on the ergonomic database, a deep learning-based comparison algorithm is used to compare the user's real-time data with the standards in the database and dynamically adjust the height of the desk.

2. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 1, characterized in that: The data acquisition module in the S1 basic data information acquisition includes a pressure sensor, an infrared sensor and a camera. The pressure sensor is used to detect the pressure applied by the user on the seat, the infrared sensor is used to measure the distance between the user and the desk, and the camera is used to capture the user's body posture information.

3. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 1, characterized in that: The preset algorithm in the S2 raw data processing and analysis calculates the user's weight based on the pressure detected by the pressure sensor, and uses the distance measured by the infrared sensor to assist in determining the user's height. At the same time, it analyzes the user's body posture based on the image taken by the camera, including the position and angle information of the head, shoulders and waist. The skeleton key point detection algorithm uses the skeleton key point detection algorithm based on the graph neural network to improve the accuracy of posture analysis.

4. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 1, characterized in that: The electric lifting device in the height adjustment execution module in the S4 mechanical action adopts a high-precision screw nut transmission structure, and receives instructions sent by the data analysis and processing module through the motor controller to drive the electric lifting device to accurately adjust the height of the office desk.

5. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 1, characterized in that: The ergonomic database established in the S5 action information feedback adjustment stores the optimal desk height parameters for people of different heights and weights in different working scenarios, as well as the corresponding body posture standards.

6. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 1, characterized in that: The data analysis and processing module in the S3 mechanical action instruction conversion compares the user's real-time data in the S5 action information feedback adjustment with the database standard to determine the adjustment plan for the height of the desk. The comparison algorithm based on deep learning continuously optimizes the accuracy of the comparison by training the network model.

7. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 2, characterized in that: The S1 basic data information is collected when the user uses the desk for the first time. The data collection module collects the user's basic physical information. The data analysis and processing module calculates the initial desk height suitable for the user based on the information and the ergonomic database, and controls the height adjustment execution module to adjust the desk to this height.

8. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 1, characterized in that: In the S1 basic data information collection, after the pressure sensor obtains the current user information, the data collection module continuously collects the current user's body data in real time and transmits it to the data analysis and processing module. When the data analysis and processing module determines that the user is in a bad posture or needs to adjust the working state, it sends the corresponding height adjustment instruction to the height adjustment execution module.

9. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 3, characterized in that: When the data analysis and processing module in the S2 raw data processing and analysis analyzes the body posture data, if it is detected that the user has lowered his head continuously for more than a preset time and the angle between the head and the shoulder is less than a preset angle, it is determined that the user is in a bad posture state, and the height adjustment execution module is triggered to adjust the height of the desk.

10. The method for automatic height adjustment and ergonomic optimization of a smart office desk according to claim 5, characterized in that: A data self-check function is synchronously established in the S5 action information feedback adjustment. When an abnormality in the height adjustment of the desk or an abnormality in the user's body data is detected, the data self-check function will be automatically triggered to conduct a comprehensive check on the data acquisition module, the data analysis and processing module, and the height adjustment execution module to ensure the stability and accuracy of the system operation.

Images