Intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics

Abstract

The invention relates to the field of combination of smart home and health medical technologies, in particular to an intelligent spine adjusting health chair based on gravity physics and human body autonomous balance characteristics. According to the method, gravity distribution semantic feature extraction and user sitting posture semantic feature extraction are respectively performed on a user gravity distribution image and a user sitting posture image collected by cameras of a gravity sensor and a posture sensor by adopting an image processing and analysis technology based on deep learning, and feature enhancement is performed; and automatically recommending an angle value according to the normalized semantic features of the sitting posture of the user under the guidance of the gravity distribution semantic features serving as auxiliary condition information. Therefore, the optimal spine adjusting scheme can be calculated according to different conditions of each user, so that the spine is effectively supported and corrected, and personalized comfortable experience is improved and provided.

Description

Technical Field

[0001] The present application relates to the field of integration of smart home and health care technology, and specifically to an intelligent spine adjustment health chair based on gravity physics and the autonomous balance characteristics of the human body. Background Art

[0002] Modern people's pursuit of a healthy lifestyle is constantly increasing, which has prompted spinal health care and traditional Chinese medicine health care methods to become the focus of more and more people. However, existing spinal adjustment methods often ignore the importance of individual differences and self-regulation. Specifically, since each person's spinal condition and health status are unique, traditional spinal adjustment methods may be too uniform and fail to take into account the differences in individual gravity and sitting posture, resulting in limited adjustment effects.

[0003] Therefore, an optimized spine adjustment health chair is desired to meet the personalized needs of different users with different spinal conditions and sitting postures. Summary of the invention

[0004] This application is made in consideration of the above problems. One purpose of this application is to provide an intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics.

[0005] The embodiment of the present application provides an intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics, which includes:

[0006] Seat body;

[0007] The sensor module includes a gravity sensor and a posture sensor, wherein the gravity sensor is used to collect a user's gravity distribution image, and the posture sensor is used to collect a user's sitting posture image using a camera in the posture sensor;

[0008] A control and driving module, configured to analyze and process the user's gravity distribution image and the user's sitting posture image to obtain a spine adjustment plan, and drive a spine adjustment module to perform adjustment based on the spine adjustment plan;

[0009] The spinal adjustment module includes a plurality of adjustable support blocks, wherein the adjustable support blocks are used to make adjustments based on the spinal adjustment scheme;

[0010] The intelligent control module is used to control the spine adjustment module through a terminal device.

[0011] For example, according to the embodiment of the present application, the intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics, wherein the control and drive module includes:

[0012] A gravity distribution semantic feature extraction unit, used for inputting the user's gravity distribution image into a gravity distribution feature extractor to obtain a gravity distribution semantic feature vector;

[0013] A user sitting posture semantic feature extraction unit, used for inputting the user sitting posture image into a user sitting posture semantic feature extractor to obtain a user sitting posture semantic feature map;

[0014] A user sitting posture semantic feature enhancement unit, used for inputting the user sitting posture semantic feature map into a feature attention selection enhancer under a compression suppression structure to obtain an enhanced user sitting posture semantic feature map;

[0015] A gravity feature assisted sitting posture semantic normalization unit, used to use the gravity distribution semantic feature vector as auxiliary condition information, and perform conditional normalization on the enhanced user sitting posture semantic feature map based on the auxiliary condition information to obtain a gravity distribution condition guided user sitting posture normalized semantic feature map as a gravity distribution condition guided user sitting posture normalized semantic feature;

[0016] The angle recommendation unit is used to guide the user to normalize the semantic features of the sitting posture based on the gravity distribution conditions to obtain the angle recommendation value as the spinal adjustment plan.

[0017] For example, according to the embodiment of the present application, the intelligent spine adjustment health chair based on gravity physics and human body's autonomous balance characteristics, wherein the gravity distribution semantic feature extraction unit is used to: input the user's gravity distribution image into the gravity distribution feature extractor based on the void convolutional neural network model to obtain the gravity distribution semantic feature vector.

[0018] For example, according to the embodiment of the present application, the intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics, wherein the user sitting posture semantic feature extraction unit is used to: input the user sitting posture image into the user sitting posture semantic feature extractor based on the AlexNet model to obtain the user sitting posture semantic feature map.

[0019] For example, according to the intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics of the embodiment of the present application, the user sitting posture semantic feature enhancement unit includes:

[0020] A global mean processing subunit, used for calculating the global mean of each feature matrix along the channel dimension of the user sitting posture semantic feature map to obtain a user sitting posture semantic feature compression information representation vector;

[0021] A one-dimensional convolutional coding subunit, used for performing one-dimensional convolutional coding on the user sitting posture semantic feature compression information representation vector to obtain a correlation representation feature vector between the user sitting posture semantic feature compression information;

[0022] a cascade subunit, configured to cascade the user sitting posture semantic feature compression information representation vector and the user sitting posture semantic feature compression information association representation feature vector to obtain a user sitting posture semantic feature compression information multi-scale representation vector;

[0023] An information compression subunit, used for inputting the multi-scale representation vector of the user sitting posture semantic feature compression information into a compression information feature extraction module to obtain a multi-scale associated feature vector of the user sitting posture semantic feature compression information, wherein the compression information feature extraction module is a multi-layer perceptron including two fully connected layers and a SiLU activation function;

[0024] A normalization subunit, used for performing a normalization operation on the multi-scale associated feature vector of the user sitting posture semantic feature compression information using a Sigmoid function to obtain a user sitting posture semantic weight feature vector;

[0025] The amplification and suppression subunit is used to perform feature amplification and suppression operations on the user sitting posture semantic feature map based on the user sitting posture semantic weight feature vector to obtain the enhanced user sitting posture semantic feature map.

[0026] For example, according to the embodiment of the present application, the intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics, wherein the gravity feature assisted sitting posture semantic normalization unit includes:

[0027] A mean and standard deviation calculation subunit, used for calculating the mean and standard deviation between the feature values ​​of each position in the enhanced user sitting posture semantic feature map to obtain the enhanced user sitting posture semantic feature mean and the enhanced user sitting posture semantic feature standard deviation;

[0028] The enhanced user sitting posture semantic feature value batch normalization processing subunit is used to perform batch normalization processing on the feature value of each position in the enhanced user sitting posture semantic feature map based on the enhanced user sitting posture semantic feature mean and the enhanced user sitting posture semantic feature standard deviation to obtain a normalized enhanced user sitting posture semantic feature map composed of multiple enhanced user sitting posture semantic batch normalized feature values;

[0029] A multi-layer perception subunit, used for inputting the gravity distribution semantic feature vector as an auxiliary modal condition into two multi-layer perceptrons with different network parameters to obtain a gravity distribution semantic auxiliary scaling factor and a gravity distribution semantic auxiliary translation factor;

[0030] A fusion subunit, used for multiplying the feature value of each position in the normalized enhanced user sitting posture semantic feature map by the gravity distribution semantic auxiliary scaling factor and then adding the feature value by the gravity distribution semantic auxiliary translation factor to obtain the gravity distribution condition guided user sitting posture normalized semantic feature map composed of multiple gravity distribution assisted enhanced user sitting posture semantic feature values;

[0031] The enhanced user sitting posture semantic feature value batch normalization processing subunit is used to:

[0032] Subtracting the feature value of each position in the enhanced user sitting posture semantic feature map from the mean value of the enhanced user sitting posture semantic feature to obtain an enhanced user sitting posture semantic offset value;

[0033] Calculating the square root of the sum of the standard deviation of the enhanced user sitting posture semantic feature and the adjustment parameter to obtain the enhanced user sitting posture semantic adjustment value;

[0034] The enhanced user sitting posture semantics offset value is divided by the enhanced user sitting posture semantics adjustment value to obtain the enhanced user sitting posture semantics batch normalized feature value.

[0035] For example, according to the embodiment of the present application, an intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics, wherein the angle recommendation unit is used to: input the gravity distribution condition to guide the user's sitting posture normalized semantic feature map into the decoder-based support part angle recommender to obtain the angle recommendation decoding value.

[0036] For example, according to the embodiment of the present application, the intelligent spine adjustment health chair based on gravity physics and the human body's autonomous balance characteristics also includes a training module for training the gravity distribution feature extractor based on the void convolutional neural network model, the user sitting posture semantic feature extractor based on the AlexNet model, the feature attention selection enhancer under the compression suppression structure, and the support part angle recommender based on the decoder.

[0037] For example, according to the intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics of the embodiment of the present application, the training module includes:

[0038] A training data acquisition unit, used to acquire training data, wherein the training data includes a training user's gravity distribution image and a training user's sitting posture image;

[0039] A training gravity distribution semantic feature extraction unit, used for inputting the training user gravity distribution image into the gravity distribution feature extractor based on the dilated convolutional neural network model to obtain a training gravity distribution semantic feature vector;

[0040] A training user sitting posture semantic feature extraction unit, used for inputting the training user sitting posture posture image into the user sitting posture semantic feature extractor based on the AlexNet model to obtain a training user sitting posture semantic feature map;

[0041] A training user sitting posture semantic feature enhancement unit, used for inputting the training user sitting posture semantic feature map into the feature attention selection enhancer under the compression suppression structure to obtain a training enhanced user sitting posture semantic feature map;

[0042] A training gravity feature assisted sitting posture semantic normalization unit is used to use the training gravity distribution semantic feature vector as auxiliary condition information, and perform conditional normalization based on the auxiliary condition information on the training enhanced user sitting posture semantic feature map to obtain a training gravity distribution condition guided user sitting posture normalized semantic feature map;

[0043] A decoding loss calculation unit, used for inputting the normalized semantic feature map of the user's sitting posture guided by the training gravity distribution condition into the decoder-based support part angle recommender to obtain a decoding loss function value;

[0044] A semantic loss calculation unit, used to calculate the gravity distribution condition-guided user sitting posture normalized semantic loss function value of the training gravity distribution condition-guided user sitting posture normalized semantic feature map;

[0045] A loss training unit is used to guide the weighted sum of the normalized semantic loss function values ​​of the user sitting posture based on the decoding loss function value and the gravity distribution condition as the loss function value and train the gravity distribution feature extractor based on the void convolutional neural network model, the user sitting posture semantic feature extractor based on the AlexNet model, the feature attention selection enhancer under the compression suppression structure and the decoder-based support part angle recommender through back propagation of gradient descent.

[0046] According to the embodiment of the present application, the intelligent spine adjustment health chair based on gravity physics and the human body's autonomous balance characteristics can calculate the best spine adjustment plan according to the different conditions of each user, so as to effectively support and correct the spine and enhance the personalized comfort experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0048] Figure 1 The structure diagram of the intelligent spine adjustment health chair based on gravity physics and human body's autonomous balance characteristics in the embodiment of the present application is shown;

[0049] Figure 2 A schematic diagram showing the structure of the control and drive module of the intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics in the embodiment of the present application; and

[0050] Figure 3 A flow chart of an intelligent spine adjustment method based on gravity physics and the autonomous balance characteristics of the human body in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0052] The terms used in this specification are those common terms currently widely used in the art in consideration of the functions of the present application, but these terms may vary according to the intentions of those skilled in the art, precedents, or new technologies in the art. In addition, specific terms may be selected, and in this case, their detailed meanings will be described in the detailed description of the present application. Therefore, the terms used in the specification should not be understood as simple names, but rather as a general description based on the meaning of the terms and the present application.

[0053] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0054] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0055] In addition, the application architecture diagram in the embodiment of the present application is intended to more clearly illustrate the technical solution in the embodiment of the present application, and does not constitute a limitation on the technical solution provided in the embodiment of the present application. Of course, for other application architectures and business applications, the technical solution provided in the embodiment of the present application is also applicable to similar problems.

[0056] The following is a non-restrictive description of the intelligent spine adjustment health chair based on gravity physics and the human body's autonomous balance characteristics provided according to at least one embodiment of the present application through several examples or embodiments. As described below, different features in these specific examples or embodiments can be combined with each other without conflicting with each other to obtain new examples or embodiments, and these new examples or embodiments also fall within the scope of protection of the present application.

[0057] Based on this, the present application proposes an intelligent spine adjustment health chair based on gravity physics and the autonomous balance characteristics of the human body, such as Figure 1 As shown, it includes: a seat body 100; a sensor module 200, including a gravity sensor 210 and a posture sensor 220, wherein the gravity sensor 210 is used to collect a user's gravity distribution image, and the posture sensor 220 is used to use a camera in the posture sensor 220 to collect a user's sitting posture image; a control and drive module 300, used to analyze and process the user's gravity distribution image and the user's sitting posture image to obtain a spinal adjustment plan, and drive the spinal adjustment module 400 to adjust based on the spinal adjustment plan; the spinal adjustment module 400, including a plurality of adjustable support blocks, and the adjustable support blocks are used to adjust based on the spinal adjustment plan; an intelligent control module 500, used to control the spinal adjustment module 400 through a terminal device.

[0058] Among them, the design of the seat body 100 is ergonomic, using soft and well-supportive materials, aiming to provide users with the best sitting experience. The seat not only provides good support and comfort, but also has excellent durability and breathability, ensuring that it remains comfortable after long-term use. The spine adjustment module 400 includes a plurality of adjustable support blocks; the sensor module 200 includes a gravity sensor 210 and a posture sensor 220 that can detect the user's gravity distribution and sitting posture in real time, and based on these data, calculate the best spine adjustment plan through an intelligent algorithm, and drive the corresponding (motor and pneumatic system) actuator to automatically adjust. This design ensures that no matter how the user's spinal condition changes, the seat can quickly adapt and provide the best support and correction effect.

[0059] Correspondingly, in the above-mentioned control and drive module 300, it uses deep learning-based image processing and analysis technology to extract gravity distribution semantic features and user sitting posture semantic features and feature enhancement for the user gravity distribution image and user sitting posture image collected by the camera of the gravity sensor 210 and the posture sensor 220, so as to automatically recommend angle values ​​based on the gravity distribution semantic features as the normalized semantic features of the user sitting posture guided by auxiliary condition information. In this way, through the intelligent algorithm based on deep learning, the best spinal adjustment plan can be calculated for each user's different situation to effectively support and correct the spine and provide a personalized comfort experience.

[0060] Accordingly, if Figure 2As shown, the control and drive module 300 includes: a gravity distribution semantic feature extraction unit 310, which is used to input the user gravity distribution image into the gravity distribution feature extractor to obtain a gravity distribution semantic feature vector; a user sitting posture semantic feature extraction unit 320, which is used to input the user sitting posture image into the user sitting posture semantic feature extractor to obtain a user sitting posture semantic feature map; a user sitting posture semantic feature enhancement unit 330, which is used to input the user sitting posture semantic feature map into the feature attention selection enhancer under the compression suppression structure to obtain an enhanced user sitting posture semantic feature map; a gravity feature auxiliary sitting posture semantic normalization unit 340, which is used to use the gravity distribution semantic feature vector as auxiliary condition information, and perform conditional normalization on the enhanced user sitting posture semantic feature map based on the auxiliary condition information to obtain a gravity distribution condition-guided user sitting posture normalized semantic feature map as a gravity distribution condition-guided user sitting posture normalized semantic feature; an angle recommendation unit 350, which is used to obtain an angle recommendation value as the spinal adjustment plan based on the gravity distribution condition-guided user sitting posture normalized semantic feature.

[0061] Specifically, in the technical solution of the present application, firstly, considering that the user's gravity distribution image contains semantic feature information and structures of different scales about gravity distribution, and the gravity distribution feature extractor can extract and mine feature information about gravity distribution from the user's gravity distribution image, but has a weak perception ability for semantic feature information of gravity distribution of different scales. Based on this, in the technical solution of the present application, the user's gravity distribution image is input into a gravity distribution feature extractor based on a dilated convolutional neural network model to capture and mine speech feature information of gravity distribution of different scales, thereby obtaining a gravity distribution semantic feature vector.

[0062] Correspondingly, the gravity distribution semantic feature extraction unit 310 is used to: input the user gravity distribution image into a gravity distribution feature extractor based on a dilated convolutional neural network model to obtain the gravity distribution semantic feature vector.

[0063] Next, considering that the user sitting posture image expresses rich voice feature information and complex semantic relationships in the user posture information, and the user sitting posture semantic feature extractor has a weak ability to capture the complex visual features in the user sitting posture image. Therefore, in the technical solution of the present application, the user sitting posture image is input into the user sitting posture semantic feature extractor based on the AlexNet model to obtain a user sitting posture semantic feature map. That is, the AlexNet model has a powerful feature extraction capability and can learn rich visual features from the user sitting posture image, which are crucial for understanding the user's sitting posture and spinal state.

[0064] Correspondingly, the user sitting posture semantic feature extraction unit 320 is used to: input the user sitting posture image into a user sitting posture semantic feature extractor based on the AlexNet model to obtain the user sitting posture semantic feature map.

[0065] Then, considering that the user sitting posture semantic feature map hides subtle semantic features that are extremely critical to the user sitting posture angle adjustment, and is accompanied by some non-critical background information. And strengthening the recognition ability of these key subtle user sitting posture semantic features plays a decisive role in improving the accuracy of angle recommendation. Therefore, in the technical solution of the present application, the user sitting posture semantic feature map is input into the feature attention selection enhancer under the compression suppression structure to obtain an enhanced user sitting posture semantic feature map. Specifically, firstly, the user sitting posture semantic feature map is feature compressed by global average pooling processing, and the key user sitting posture semantic information is extracted and retained to form a user sitting posture semantic feature compression information representation vector. Then, the one-dimensional convolutional coding technology is applied to deeply analyze the compressed semantic feature vector, explore its inherent semantic relevance, and enhance the perception ability of local semantic features through cascade operations to achieve multi-scale feature fusion. Then, a multi-layer perceptron is used to further extract and refine the multi-scale semantic features of the secondary user sitting posture to form a multi-scale correlation feature vector. Afterwards, the multi-scale correlation feature vector is normalized by the Sigmoid function to generate the weight value of the feature channel to guide the model to focus on the key user posture semantic features. Finally, the user sitting posture semantic feature map is amplified and suppressed based on the weight value to enhance the recognition ability of key user posture semantic features, suppress irrelevant background features, and finally obtain an enhanced user sitting posture semantic feature map.

[0066] Correspondingly, the user sitting posture semantic feature enhancement unit 330 includes: a global mean processing subunit, used to calculate the global mean of each feature matrix along the channel dimension of the user sitting posture semantic feature map to obtain a user sitting posture semantic feature compression information representation vector; a one-dimensional convolution encoding subunit, used to perform one-dimensional convolution encoding on the user sitting posture semantic feature compression information representation vector to obtain a user sitting posture semantic feature compression information correlation representation feature vector; a cascade subunit, used to cascade the user sitting posture semantic feature compression information representation vector and the user sitting posture semantic feature compression information correlation representation feature vector to obtain a user sitting posture semantic feature compression information multi-scale representation vector; an information compression subunit; A unit is used to input the multi-scale representation vector of the user sitting posture semantic feature compression information into a compression information feature extraction module to obtain a multi-scale associated feature vector of the user sitting posture semantic feature compression information, wherein the compression information feature extraction module is a multi-layer perceptron including two fully connected layers and a SiLU activation function; a normalization subunit is used to use a Sigmoid function to normalize the multi-scale associated feature vector of the user sitting posture semantic feature compression information to obtain a user sitting posture semantic weight feature vector; an amplification and suppression subunit is used to perform feature amplification and suppression operations on the user sitting posture semantic feature map based on the user sitting posture semantic weight feature vector to obtain the enhanced user sitting posture semantic feature map.

[0067] In a specific example, the user sitting posture semantic feature enhancement unit 330 is used to: input the user sitting posture semantic feature map into the feature attention selection enhancer under the compression suppression structure, and process it with the following enhancement formula to obtain the enhanced user sitting posture semantic feature map; wherein the enhancement formula is:

[0068]

[0069]

[0070] Where F is the user's sitting posture semantic feature map, f x (i, j) represents the feature value of the point with coordinates (i, j) in the x-th channel of the user sitting posture semantic feature map F, H and W are the height and width of the user sitting posture semantic feature map, respectively, V s1 is the user's sitting posture semantic feature compressed information representation vector, Conv 1d (·) is a one-dimensional convolutional code, concat(·,·) is a cascade process of vectors, MLP is a multi-layer perceptron, sigmoid is a sigmoid function, represents multiplication by position, and F' is the enhanced user sitting posture semantic feature map.

[0071] Furthermore, considering that the gravity distribution semantic feature vector can assist in identifying subtle changes in sitting posture in the enhanced user sitting posture semantic feature map, these changes may have an important impact on spinal health. Based on this, by utilizing the gravity distribution semantic feature vector, the expression strength of the semantic information domain in the enhanced user sitting posture semantic feature map is enhanced, and at the same time, these feature maps are standardized to eliminate the offset of internal covariates. In the technical solution of the present application, the gravity distribution semantic feature vector is used as auxiliary condition information, and the enhanced user sitting posture semantic feature map is conditionally normalized based on the auxiliary condition information to obtain a gravity distribution condition-guided user sitting posture normalized semantic feature map. Specifically, on the basis of batch normalization, the gravity distribution semantic feature vector is further used as auxiliary condition information, so that the normalization process can integrate additional gravity distribution semantic content, and realize dynamic normalization adjustment based on the enhanced user sitting posture semantic characteristics, so as to more accurately map the normalized information of the user sitting posture based on the gravity distribution condition, and form a user sitting posture normalized semantic feature map that integrates the gravity distribution semantic information.

[0072] Accordingly, the gravity feature assisted sitting posture semantic normalization unit 340 includes: a mean and standard deviation calculation subunit, which is used to calculate the mean and standard deviation between the feature values ​​of each position in the enhanced user sitting posture semantic feature map to obtain the enhanced user sitting posture semantic feature mean and the enhanced user sitting posture semantic feature standard deviation; an enhanced user sitting posture semantic feature value batch normalization processing subunit, which is used to perform batch normalization processing on the feature value of each position in the enhanced user sitting posture semantic feature map based on the enhanced user sitting posture semantic feature mean and the enhanced user sitting posture semantic feature standard deviation to obtain a plurality of enhanced user sitting posture semantic batch normalized feature values. A normalized enhanced user sitting posture semantic feature map composed of the gravity distribution semantic feature vectors; a multi-layer perception subunit, used to input the gravity distribution semantic feature vectors as auxiliary modal conditions into two multi-layer perceptrons with different network parameters to obtain a gravity distribution semantic auxiliary scaling factor and a gravity distribution semantic auxiliary translation factor; a fusion subunit, used to multiply the feature value of each position in the normalized enhanced user sitting posture semantic feature map by the gravity distribution semantic auxiliary scaling factor and then add the feature value by the gravity distribution semantic auxiliary translation factor to obtain the gravity distribution condition-guided user sitting posture normalized semantic feature map composed of multiple gravity distribution-assisted enhanced user sitting posture semantic feature values.

[0073] Among them, the enhanced user sitting posture semantic feature value batch normalization processing subunit is used to: subtract the feature value of each position in the enhanced user sitting posture semantic feature map from the mean value of the enhanced user sitting posture semantic feature to obtain the enhanced user sitting posture semantic offset value; calculate the square root of the sum of the standard deviation of the enhanced user sitting posture semantic feature and the adjustment parameter to obtain the enhanced user sitting posture semantic adjustment value; divide the enhanced user sitting posture semantic offset value by the enhanced user sitting posture semantic adjustment value to obtain the enhanced user sitting posture semantic batch normalized feature value.

[0074] In a specific example, the gravity feature assisted sitting posture semantic normalization unit 340 is used to: use the gravity distribution semantic feature vector as auxiliary condition information, and perform conditional normalization based on the auxiliary condition information on the enhanced user sitting posture semantic feature map using the following normalization formula to obtain a gravity distribution condition guided user sitting posture normalized semantic feature map; wherein the normalization formula is:

[0075]

[0076] γ mlp =MLP γ (c)

[0077]

[0078]

[0079] Among them, F' is the enhanced user sitting posture semantic feature map, F' i represents the feature value of each position in the enhanced user sitting posture semantic feature map, μ F' and σ F' 2 are the mean and standard deviation of the enhanced user sitting posture semantic features, ∈ is the adjustment parameter, is the feature value of each position in the normalized enhanced user sitting posture semantic feature map, c is the auxiliary condition information, MLP γ and MLP β are two multilayer perceptrons with different network parameters, γ mlp and β mlp are the gravity distribution semantics-assisted scaling factor and the gravity distribution semantics-assisted translation factor, respectively, It is the feature value of each position in the normalized semantic feature map of the user's sitting posture guided by the gravity distribution condition.

[0080] Then, the gravity distribution condition guiding the user's sitting posture normalized semantic feature map is input into the support part angle recommender based on the decoder to obtain the angle recommendation decoding value. That is, the gravity distribution condition guiding the user's sitting posture normalized semantic feature obtained by conditional normalization using the gravity distribution semantic feature vector as auxiliary condition information and the enhanced user sitting posture semantic feature is decoded to automatically recommend the angle value. In this way, through the intelligent algorithm based on deep learning, the best spinal adjustment plan can be calculated for each user's different situation to effectively support and correct the spine and provide a personalized comfort experience.

[0081] Correspondingly, the angle recommendation unit 350 is used to: input the gravity distribution condition-guided user sitting posture normalized semantic feature map into the decoder-based support part angle recommender to obtain an angle recommendation decoding value.

[0082] Considering that the gravity distribution semantic feature vector and the enhanced user sitting posture semantic feature map respectively represent the image semantic features of the user gravity distribution image and the user sitting posture image based on different feature extraction dimensions, therefore, when the gravity distribution semantic feature vector is used as auxiliary condition information to perform conditional normalization on the enhanced user sitting posture semantic feature map based on the auxiliary condition information, the gravity distribution condition-guided user sitting posture normalized semantic feature map will also have decoding regression recognition difficulties attributed to the complex image semantic fusion representation due to the differences in the distribution of semantic features of heterogeneous images, thereby affecting the accuracy of the decoding results.

[0083] In a preferred embodiment of the present application, inputting the gravity distribution condition-guided user sitting posture normalized semantic feature map into a decoder-based support part angle recommender to obtain an angle recommendation decoded value includes:

[0084] Expanding the gravity distribution condition-guided user sitting posture normalized semantic feature map into a gravity distribution condition-guided user sitting posture normalized semantic feature vector;

[0085] Subtract the zero norm of the normalized semantic feature vector of the user's sitting posture guided by the gravity distribution condition from the length of the normalized semantic feature vector of the user's sitting posture guided by the gravity distribution condition to obtain a normalized semantic isolated representation value of the user's sitting posture guided by the gravity distribution condition;

[0086] Calculate the logarithm of the square of the normalized semantic isolated representation value of the user's sitting posture guided by the gravity distribution condition and the sum of the normalized semantic isolated representation value of the user's sitting posture guided by the gravity distribution condition to obtain the normalized semantic information order value of the user's sitting posture guided by the gravity distribution condition;

[0087] Calculate each eigenvalue of the normalized semantic feature vector of the user's sitting posture guided by the gravity distribution condition as the base, the difference between the normalized semantic isolated representation value of the user's sitting posture guided by the gravity distribution condition and one is the exponential power function, and perform point multiplication with the normalized semantic information order value of the user's sitting posture guided by the gravity distribution condition to obtain the normalized semantic leading vector of the user's sitting posture guided by the gravity distribution condition;

[0088] After performing a dot multiplication on the difference between the gravity distribution condition-guided user sitting posture normalized semantic feature vector and the gravity distribution condition-guided user sitting posture normalized semantic isolated representation value minus one, the dot multiplication is then performed on the inverse of the gravity distribution condition-guided user sitting posture normalized semantic isolated representation value to obtain the gravity distribution condition-guided user sitting posture normalized semantic field constraint vector;

[0089] Calculate an exponential function with a natural constant as the base and each eigenvalue of the gravity distribution condition guiding the user's sitting posture normalized semantic field constraint vector as the exponent to obtain a gravity distribution condition guiding the user's sitting posture normalized semantic field bias vector;

[0090] The gravity distribution condition guided user sitting posture normalized semantic leading vector and the gravity distribution condition guided user sitting posture normalized semantic field bias vector are added to obtain an optimized gravity distribution condition guided user sitting posture normalized semantic feature vector;

[0091] The optimized gravity distribution condition guides the user's sitting posture normalized semantic feature vector to be input into the decoder-based support part angle recommender to obtain the angle recommendation decoding value.

[0092] Specifically, the gravity distribution condition guides the user's sitting posture normalized semantic feature vector, for example, the optimization process denoted as V is expressed as:

[0093]

[0094] n=L-||V||0

[0095] Wherein, V represents the normalized semantic feature vector of the user's sitting posture guided by the gravity distribution condition, ||·||0 represents the zero norm of the vector, L represents the length of the normalized semantic feature vector of the user's sitting posture guided by the gravity distribution condition, n represents the normalized semantic isolated representation value of the user's sitting posture guided by the gravity distribution condition, ⊙ is the point multiplication by position, is added by position, (·) ⊙(n-1) A power function is represented by calculating the feature value of each position in the feature vector, with the difference value of the normalized semantic isolated representation value of the user's sitting posture guided by the gravity distribution condition minus one as the exponent, and V' represents the normalized semantic feature vector of the user's sitting posture guided by the optimized gravity distribution condition.

[0096] Therefore, the high-dimensional feature manifold of the normalized semantic feature vector guided by the gravity distribution condition, which is obtained by expanding the normalized semantic feature map of the user's sitting posture guided by the gravity distribution condition, is represented by the eigenvalue of the feature set as the vector field of the aggregation dimension, and the superposition value of the vector field of the normalized semantic feature vector guided by the gravity distribution condition at the isolated zero position is used as the order information to fix the local position of the eigenvalue of its feature set, and a bias for the reversibility of the feature regression distribution field of the normalized semantic feature vector guided by the gravity distribution condition is added as a reward to achieve the mapping target tracking of the regression distribution of the normalized semantic feature vector guided by the gravity distribution condition for the eigenvalue position, so that the feature set of the normalized semantic feature vector guided by the gravity distribution condition can perceive the mapping migration to the aggregate distribution, thereby improving the accuracy of the angle recommendation decoding value obtained by inputting the normalized semantic feature vector guided by the gravity distribution condition into the decoder-based support part angle recommender by improving the complex image semantic fusion regression comprehensibility of the normalized semantic feature vector guided by the gravity distribution condition.

[0097] Furthermore, in the technical solution of the present application, the intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics also includes a training module for training the gravity distribution feature extractor based on the void convolutional neural network model, the user sitting posture semantic feature extractor based on the AlexNet model, the feature attention selection enhancer under the compression suppression structure, and the support part angle recommender based on the decoder.

[0098] Among them, the training module includes: a training data acquisition unit, which is used to collect training data, and the training data includes a training user gravity distribution image and a training user sitting posture image; a training gravity distribution semantic feature extraction unit, which is used to input the training user gravity distribution image into the gravity distribution feature extractor based on the void convolutional neural network model to obtain a training gravity distribution semantic feature vector; a training user sitting posture semantic feature extraction unit, which is used to input the training user sitting posture image into the user sitting posture semantic feature extractor based on the AlexNet model to obtain a training user sitting posture semantic feature map; a training user sitting posture semantic feature enhancement unit, which is used to input the training user sitting posture semantic feature map into the feature attention selection enhancer under the compression suppression structure to obtain a training enhanced user sitting posture semantic feature map; a training gravity feature auxiliary sitting posture semantic normalization unit, which is used to use the training gravity distribution semantic feature vector as auxiliary condition information to enhance the training enhanced user sitting posture semantic feature map. The feature map is conditionally normalized based on auxiliary condition information to obtain a training gravity distribution condition-guided user sitting posture normalized semantic feature map; a decoding loss calculation unit is used to input the training gravity distribution condition-guided user sitting posture normalized semantic feature map into the decoder-based support part angle recommender to obtain a decoding loss function value; a semantic loss calculation unit is used to calculate the gravity distribution condition-guided user sitting posture normalized semantic loss function value of the training gravity distribution condition-guided user sitting posture normalized semantic feature map; a loss training unit is used to train the gravity distribution feature extractor based on the void convolutional neural network model, the user sitting posture semantic feature extractor based on the AlexNet model, the feature attention selection enhancer under the compression suppression structure, and the decoder-based support part angle recommender based on the weighted sum of the decoding loss function value and the gravity distribution condition-guided user sitting posture normalized semantic loss function value as the loss function value and through gradient descent back propagation.

[0099] Therefore, in a preferred embodiment of the present application, the semantic loss calculation unit is used to: expand the training gravity distribution condition-guided user sitting posture normalized semantic feature map into a training gravity distribution condition-guided user sitting posture normalized semantic feature vector; calculate the gravity distribution condition-guided user sitting posture normalized semantic sum matrix and the gravity distribution condition-guided user sitting posture normalized semantic difference matrix based on the training gravity distribution condition-guided user sitting posture normalized semantic feature vector, wherein the eigenvalue of the (i, j)th position of the gravity distribution condition-guided user sitting posture normalized semantic sum matrix is ​​the average of the i-th eigenvalue and the j-th eigenvalue of the training gravity distribution condition-guided user sitting posture normalized semantic feature vector, and the eigenvalue of the (i, j)th position of the gravity distribution condition-guided user sitting posture normalized semantic difference matrix is ​​half of the absolute value of the difference between the i-th eigenvalue and the j-th eigenvalue of the training gravity distribution condition-guided user sitting posture normalized semantic feature vector; and compare the training gravity distribution condition-guided user sitting posture normalized semantic feature vector with the gravity distribution condition-guided user sitting posture normalized semantic feature vector respectively. The query-based matrix multiplication of the semantic sum matrix and the gravity distribution condition-guided user sitting posture normalized semantic difference matrix is ​​performed to obtain the gravity distribution condition-guided user sitting posture normalized semantic query sum vector and the gravity distribution condition-guided user sitting posture normalized semantic query difference vector; the vector inner product of the training gravity distribution condition-guided user sitting posture normalized semantic feature vector and the gravity distribution condition-guided user sitting posture normalized semantic query difference vector is calculated to obtain the first gravity distribution condition-guided user sitting posture normalized semantic feature loss term; the gravity distribution condition-guided user sitting posture normalized semantic sum matrix and the gravity distribution condition-guided user sitting posture normalized semantic difference matrix are matrix multiplied, and the Frobenius norm of the product matrix is ​​calculated to obtain the second gravity distribution condition-guided user sitting posture normalized semantic feature loss term; the first gravity distribution condition-guided user sitting posture normalized semantic feature loss term is subtracted from the product of the predetermined weight hyperparameter and the second gravity distribution condition-guided user sitting posture normalized semantic feature loss term to obtain the gravity distribution condition-guided user sitting posture normalized semantic loss function value.

[0100] The calculation process of the gravity distribution condition guiding the user's sitting posture normalized semantic loss function value is specifically expressed by the following loss calculation formula:

[0101]

[0102] Among them, M μ (i,j)=v i +v j / 2, and M σ (i,j)=|v i -v j| / 2, V is the normalized semantic feature vector of the user's sitting posture guided by the training gravity distribution condition, v i and v j are the feature values ​​of the i-th and j-th positions in the normalized semantic feature vector of the user's sitting posture guided by the training gravity distribution condition, M μ (i,j) and M σ (i, j) are the eigenvalues ​​of the (i, j)th position in the normalized semantic sum matrix of the gravity distribution condition guiding the user's sitting posture and the normalized semantic difference matrix of the gravity distribution condition guiding the user's sitting posture, respectively. μ and M σ are the normalized semantic sum matrix of user sitting posture guided by the gravity distribution condition and the normalized semantic difference matrix of user sitting posture guided by the gravity distribution condition, α is the predetermined weight hyperparameter, ||·|| F is the norm of the matrix, T represents the transpose of the vector, is the matrix multiplication, and Loss is the normalized semantic loss function value of the user's sitting posture guided by the gravity distribution condition.

[0103] Here, considering that the training gravity distribution semantic feature vector and the training enhanced user sitting posture semantic feature map respectively represent the image semantic features of the training user gravity distribution image and the training user sitting posture image based on different feature extraction dimensions, therefore, when the training gravity distribution semantic feature vector is used as auxiliary condition information to perform conditional normalization based on the auxiliary condition information on the training enhanced user sitting posture semantic feature map, the training gravity distribution condition-guided user sitting posture normalized semantic feature map will also have decoding regression recognition difficulties attributed to the complex image semantic fusion representation due to the differences in the distribution of semantic features of heterogeneous images, thereby affecting the accuracy of the decoding results.

[0104] Therefore, the present application uses the structured feature representation of the short-range-long-range cross-scale detail links of the training gravity distribution condition-guided user sitting posture normalized semantic feature graph to perform the query composition of the detail inner product space in the training gravity distribution condition-guided user sitting posture normalized semantic feature graph, so as to approximate the low-rank independent observable composition of the link detail composition provided by the structured detail interaction of the training gravity distribution condition-guided user sitting posture normalized semantic feature graph. In this way, by training with the gravity distribution condition-guided user sitting posture normalized semantic loss function, the distributed detail group of the training gravity distribution condition-guided user sitting posture normalized semantic feature graph can be used to decompose the detail group based on the complexity of the details, so as to promote the decoding regression decomposition recognition of the training gravity distribution condition-guided user sitting posture normalized semantic feature graph based on complex feature representation, and improve the accuracy of the decoding result. In this way, through the intelligent algorithm based on deep learning, the best spinal adjustment plan can be calculated for each user's different situation to effectively support and correct the spine and provide a personalized comfort experience.

[0105] Furthermore, the intelligent spine adjustment health chair based on gravity physics and the human body's autonomous balance characteristics can also include a TCM five-tone transmission part: the seat has a built-in audio playback device that can play TCM five-tone (gong, shang, jiao, zheng, yu). These audios regulate the human body's qi and blood through vibration and resonance, which helps relieve stress, improve sleep, and enhance physical fitness. Users can accurately control the type of music played and the playback time through a mobile phone APP or remote control to meet personalized needs.

[0106] Furthermore, in another embodiment of the present application, there is also provided an intelligent spine adjustment health chair based on gravity physics and the autonomous balance characteristics of the human body, which includes: a data acquisition module, which is used to collect the gravity distribution image and sitting posture image of the user, and transmit the data wirelessly. A gravity distribution feature extraction module, which is used to extract gravity distribution semantic features from the user's gravity distribution image. A sitting posture semantic feature extraction module, which is used to extract sitting posture semantic features from the user's sitting posture image. A feature enhancement module, which is used to enhance the sitting posture semantic features, retain key features, and remove background interference. A normalization processing module, which is used to conditionally normalize the enhanced sitting posture semantic features based on the gravity distribution semantic features. A control and drive module, which is used to generate a spine adjustment plan based on the analysis results and drive the spine adjustment module to make adjustments. A spine adjustment module, which is used to adjust the user's spine through multiple adjustable support blocks. A user control module, which is used to control the spine adjustment module through a terminal device.

[0107] In another embodiment, an intelligent spine adjustment health chair based on gravity physics and the human body's autonomous balance characteristics uses the following algorithm. First, data preprocessing is performed, that is, the gravity distribution image and the sitting posture image are preprocessed. Then feature extraction is performed, and the feature extraction includes: using a convolutional neural network (CNN) to extract gravity distribution semantic features from the gravity distribution image. Use the AlexNet model to extract sitting posture semantic features from the sitting posture image. Then feature enhancement is performed, that is, the sitting posture semantic features are enhanced, and the key features are enhanced using the feature attention mechanism. Then, normalization processing is performed, that is, conditional normalization processing is performed on the enhanced sitting posture semantic features based on the gravity distribution semantic features, and auxiliary condition information is used to improve accuracy. Finally, control instructions are generated, that is, angle recommendation values ​​are generated through the decoder to formulate a spinal adjustment plan.

[0108] In another embodiment, an application scenario of an intelligent spine adjustment health chair based on gravity physics and the human body's autonomous balance characteristics is also provided, which includes: 1. User use, the user sits on the health chair, and the system automatically collects gravity distribution and sitting posture data. The user controls the adjustment of the chair through a terminal device (such as a mobile phone APP). 2. Spine adjustment process. The system analyzes the data in real time and generates a spine adjustment plan. The drive module adjusts the support block to support and correct the spine. 3. User feedback. The user can provide feedback through the APP, and the system further optimizes the adjustment plan based on the feedback.

[0109] Compared with the prior art, another embodiment of the present application has the following innovations: 1. Multi-dimensional feature fusion: integrating gravity distribution and sitting posture semantic features to provide accurate spinal adjustment solutions. 2. Intelligent algorithm: using deep learning algorithm for feature extraction and processing to improve the accuracy of adjustment. 3. Real-time adjustment: the system can collect and analyze data in real time to provide dynamic adjustment solutions. 4. Personalized solution: providing personalized spinal support and correction solutions based on the different situations of each user. In other words, compared with the prior art, this system can integrate multi-dimensional data to improve the accuracy of spinal adjustment and user experience.

[0110] Further, for another embodiment described above, a verification field and a proof of concept are also provided. Among them, verification field one is a home environment, and its goal is to verify the application effect of the intelligent spine adjustment health chair in an actual home environment. The following steps are provided for verification field one, and the steps include: 1. Installation and configuration: Install the intelligent health chair in the home, connect the gravity sensor, posture sensor and control module. Configure the data transmission channel to ensure that physiological parameters are transmitted via Bluetooth and environmental parameters are transmitted via Wi-Fi. 2. Data collection: Let the user sit on the health chair to monitor the gravity distribution and sitting posture in real time. 3. Data processing and analysis: The collected multidimensional data is reorganized in time series and analyzed for associated features. The convolutional neural network and AlexNet model are used to jointly encode the data to generate multimodal parameter time series joint graph encoding features. 4. System regulation: Based on the analysis results, control instructions are generated to adjust the support blocks and angles of the health chair. 5. Effect verification: Record user feedback and the actual regulation effect of the system to evaluate the accuracy of the system and user satisfaction.

[0111] Furthermore, a verification field two is also provided, and the verification field two is a laboratory environment, and its goal is to perform precise verification in a controlled environment and eliminate external interference factors. The following steps are provided for the verification field two, and the steps include: 1. Laboratory configuration: simulate the home environment in the laboratory and install the smart health chair. 2. Data collection and transmission: perform data collection and transmission in the same way as the home environment. 3. Simulation experiment: design a variety of experimental scenarios to simulate different user sitting postures and gravity distribution conditions. Record the system's control instructions and actual execution effects in different scenarios. 4. Data analysis: compare experimental data and system output to verify the system's analysis accuracy and response speed. 5. Result evaluation: evaluate according to the experimental results, adjust and optimize the system algorithm and configuration.

[0112] Furthermore, a verification field three is also provided, and the verification field three is a public place, and its goal is to verify the stability and adaptability of the system in a complex environment. The following steps are provided for verification field three, and the steps include: 1. Installation and configuration: Install the smart health chair in a public place (such as an office, clinic). 2. Data collection: Collect gravity distribution and sitting posture data of multiple users. 3. Data processing and analysis: Perform real-time data processing and analysis to generate control instructions. 4. System regulation: Adjust the support block and angle of the health chair to achieve spinal adjustment. 5. Effect verification: Record user feedback and evaluate the performance of the system in a complex environment.

[0113] At the same time, a proof of concept is also provided, the goal of which is to verify the overall function and actual application effect of the system, and provide a basis for large-scale application. The following steps are provided for the proof of concept, including: 1. POC design: Select a group of users, including individuals of different ages, genders and health conditions. Test in actual use environments, such as homes, offices and clinics. 2. System configuration: Install and configure the smart health chair to ensure that all sensors and modules are working properly. 3. Data collection and processing: Collect user data in real time, and perform multi-dimensional feature fusion and analysis. 4. System regulation and feedback: Generate a spinal adjustment plan based on the analysis results, and adjust the health chair in real time. Collect user feedback, record the adjustment effect and user experience. 5. Data analysis and optimization: Analyze the data in the POC process in detail, and summarize the system performance and user feedback. Optimize the system's algorithm and configuration based on the analysis results to improve the system's accuracy and user satisfaction.

[0114] Based on the above embodiments, see Figure 3 As shown in FIG. 1 , it is a flow chart of an intelligent spine adjustment method based on gravity physics and the autonomous balance characteristics of the human body in an embodiment of the present application. Figure 3 As shown, according to the embodiment of the present application, the intelligent spine adjustment method based on gravity physics and human body autonomous balance characteristics includes the following steps: S510, collecting the user's gravity distribution image through the gravity sensor, and using the camera in the posture sensor to collect the user's sitting posture image; S520, inputting the user's gravity distribution image into the gravity distribution feature extractor to obtain the gravity distribution semantic feature vector; S530, inputting the user's sitting posture image into the user's sitting posture semantic feature extractor to obtain the user's sitting posture semantic feature map; S540, inputting the user's sitting posture semantic feature map into the feature attention selection enhancer under the compression suppression structure to obtain the enhanced user sitting posture semantic feature map; S550, using the gravity distribution semantic feature vector as auxiliary condition information, conditionally normalizing the enhanced user sitting posture semantic feature map based on the auxiliary condition information to obtain the gravity distribution condition-guided user sitting posture normalized semantic feature map as the gravity distribution condition-guided user sitting posture normalized semantic feature; S560, based on the gravity distribution condition-guided user sitting posture normalized semantic feature, obtaining the angle recommendation value as the spine adjustment plan.

[0115] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned intelligent spine adjustment method based on gravity physics and human body autonomous balance characteristics have been described in the above reference. Figure 1 to Figure 2 It has been introduced in detail in the description of the intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics, and therefore, its repeated description will be omitted.

[0116] Furthermore, in an example of the present application, first, the user sits on the intelligent spine adjustment health chair, at which time the gravity sensor of the chair starts to work and detects the user's gravity distribution in real time. At the same time, the posture sensor also starts to detect the user's sitting posture, including the curvature and angle of the spine, etc. These data will be transmitted to the control and drive system in real time.

[0117] Then, the user adjusts the support blocks of the spine adjustment part according to his or her spine curve and sitting posture requirements through the mobile phone APP or remote control. After receiving the user's instructions, the control and drive system combines the data transmitted by the sensor system and analyzes and processes it through the built-in intelligent algorithm. The algorithm calculates the best spine adjustment plan based on the user's gravity distribution, sitting posture and personalized needs.

[0118] According to the calculated adjustment plan, the control and drive system drives the actuator to make corresponding adjustments. The actuator includes the support part and the angle adjustment part of the seat. By changing the shape and angle of the support part, the user's spine can be automatically adjusted and optimized. This process does not require manual operation by the user. The intelligent spine adjustment health chair can complete it automatically, ensuring that the user is provided with the best spine support and correction effect.

[0119] While adjusting the spine, users can also choose to play the five tones of traditional Chinese medicine. Send instructions through the mobile phone APP or remote control, and the audio player will start playing the selected five tones of traditional Chinese medicine after receiving the instructions. The vibration and resonance of the audio can regulate the human body's qi and blood, achieving the effect of relieving stress, improving sleep, and strengthening physical fitness.

[0120] Throughout the implementation process, this application combines gravity physics with the human body's autonomous balance characteristics and Chinese medicine's five-tone therapy to achieve automatic adjustment and optimization of the spine, while providing all-round physical and mental health support. This innovative design not only improves the user's comfort, but also promotes the user's physical and mental health.

[0121] Based on the above embodiments, another exemplary embodiment of an electronic device is also provided in the embodiments of the present application. In some possible implementations, the electronic device in the embodiments of the present application may include a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can implement the steps of the intelligent spine adjustment method based on gravity physics and the autonomous balance characteristics of the human body in the above embodiments when executing the program.

[0122] The embodiment of the present application also provides a computer-readable storage medium, on which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, the intelligent spine adjustment method based on gravity physics and human body autonomous balance characteristics according to the embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.

[0123] The embodiment of the present application also provides a computer program product or computer program, which includes computer executable instructions, and the computer executable instructions are stored in a computer readable storage medium. The processor of the computer device reads the computer executable instructions from the computer readable storage medium, and the processor executes the computer executable instructions, so that the computer device executes the intelligent spine adjustment method based on gravity physics and human body autonomous balance characteristics according to the embodiment of the present application.

[0124] Those skilled in the art will appreciate that the contents disclosed in this application may be subject to various variations and improvements. For example, the various devices or components described above may be implemented by hardware, or by software, firmware, or a combination of some or all of the three.

[0125] In addition, although the present application makes various references to certain units in the system according to embodiments of the present application, any number of different units can be used and run on the client and / or server. The units are only illustrative, and different aspects of the system and method can use different units.

[0126] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present application is not limited to any particular form of combination of hardware and software.

[0127] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0128] The above is an explanation of the present application and should not be considered as limiting thereof. Although several exemplary embodiments of the present application have been described, those skilled in the art will readily appreciate that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application.

Claims

1. An intelligent spine adjustment health chair based on gravity physics and human body's autonomous balance characteristics, characterized in that: include: Seat body; The sensor module includes a gravity sensor and a posture sensor, wherein the gravity sensor is used to collect a user's gravity distribution image, and the posture sensor is used to collect a user's sitting posture image using a camera in the posture sensor; A control and driving module, configured to analyze and process the user's gravity distribution image and the user's sitting posture image to obtain a spine adjustment plan, and drive a spine adjustment module to perform adjustment based on the spine adjustment plan; The spinal adjustment module includes a plurality of adjustable support blocks, wherein the adjustable support blocks are used to make adjustments based on the spinal adjustment scheme; The intelligent control module is used to control the spine adjustment module through a terminal device.

2. The intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics according to claim 1, characterized in that: The control and drive module comprises: A gravity distribution semantic feature extraction unit, used for inputting the user's gravity distribution image into a gravity distribution feature extractor to obtain a gravity distribution semantic feature vector; A user sitting posture semantic feature extraction unit, used for inputting the user sitting posture image into a user sitting posture semantic feature extractor to obtain a user sitting posture semantic feature map; A user sitting posture semantic feature enhancement unit, used for inputting the user sitting posture semantic feature map into a feature attention selection enhancer under a compression suppression structure to obtain an enhanced user sitting posture semantic feature map; A gravity feature assisted sitting posture semantic normalization unit, used to use the gravity distribution semantic feature vector as auxiliary condition information, and perform conditional normalization on the enhanced user sitting posture semantic feature map based on the auxiliary condition information to obtain a gravity distribution condition guided user sitting posture normalized semantic feature map as a gravity distribution condition guided user sitting posture normalized semantic feature; The angle recommendation unit is used to guide the user to normalize the semantic features of the sitting posture based on the gravity distribution conditions to obtain the angle recommendation value as the spinal adjustment plan.

3. The intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics according to claim 2, characterized in that: The gravity distribution semantic feature extraction unit is used to: input the user gravity distribution image into a gravity distribution feature extractor based on a hole convolutional neural network model to obtain the gravity distribution semantic feature vector.

4. The intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics according to claim 3 is characterized in that: The user sitting posture semantic feature extraction unit is used to: input the user sitting posture image into a user sitting posture semantic feature extractor based on the AlexNet model to obtain the user sitting posture semantic feature map.

5. The intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics according to claim 4, characterized in that: The user sitting posture semantic feature enhancement unit includes: A global mean processing subunit, used for calculating the global mean of each feature matrix along the channel dimension of the user sitting posture semantic feature map to obtain a user sitting posture semantic feature compression information representation vector; A one-dimensional convolutional coding subunit, used for performing one-dimensional convolutional coding on the user sitting posture semantic feature compression information representation vector to obtain a feature vector representing the association between the user sitting posture semantic feature compression information; a cascade subunit, configured to cascade the user sitting posture semantic feature compression information representation vector and the user sitting posture semantic feature compression information association representation feature vector to obtain a user sitting posture semantic feature compression information multi-scale representation vector; An information compression subunit, used for inputting the multi-scale representation vector of the user sitting posture semantic feature compression information into a compression information feature extraction module to obtain a multi-scale associated feature vector of the user sitting posture semantic feature compression information, wherein the compression information feature extraction module is a multi-layer perceptron including two fully connected layers and a SiLU activation function; A normalization subunit, used for performing a normalization operation on the multi-scale associated feature vector of the user sitting posture semantic feature compression information using a Sigmoid function to obtain a user sitting posture semantic weight feature vector; The amplification and suppression subunit is used to perform feature amplification and suppression operations on the user sitting posture semantic feature map based on the user sitting posture semantic weight feature vector to obtain the enhanced user sitting posture semantic feature map.

6. The intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics according to claim 5, characterized in that: The gravity feature assisted sitting posture semantic normalization unit includes: A mean and standard deviation calculation subunit, used for calculating the mean and standard deviation between the feature values ​​of each position in the enhanced user sitting posture semantic feature map to obtain the enhanced user sitting posture semantic feature mean and the enhanced user sitting posture semantic feature standard deviation; The enhanced user sitting posture semantic feature value batch normalization processing subunit is used to perform batch normalization processing on the feature value of each position in the enhanced user sitting posture semantic feature map based on the enhanced user sitting posture semantic feature mean and the enhanced user sitting posture semantic feature standard deviation to obtain a normalized enhanced user sitting posture semantic feature map composed of multiple enhanced user sitting posture semantic batch normalized feature values; A multi-layer perception subunit, used for inputting the gravity distribution semantic feature vector as an auxiliary modal condition into two multi-layer perceptrons with different network parameters to obtain a gravity distribution semantic auxiliary scaling factor and a gravity distribution semantic auxiliary translation factor; A fusion subunit, used for multiplying the feature value of each position in the normalized enhanced user sitting posture semantic feature map by the gravity distribution semantic auxiliary scaling factor and then adding the feature value by the gravity distribution semantic auxiliary translation factor to obtain the gravity distribution condition guided user sitting posture normalized semantic feature map composed of multiple gravity distribution assisted enhanced user sitting posture semantic feature values; The enhanced user sitting posture semantic feature value batch normalization processing subunit is used to: Subtracting the feature value of each position in the enhanced user sitting posture semantic feature map from the mean value of the enhanced user sitting posture semantic feature to obtain an enhanced user sitting posture semantic offset value; Calculating the square root of the sum of the standard deviation of the enhanced user sitting posture semantic feature and the adjustment parameter to obtain the enhanced user sitting posture semantic adjustment value; The enhanced user sitting posture semantics offset value is divided by the enhanced user sitting posture semantics adjustment value to obtain the enhanced user sitting posture semantics batch normalized feature value.

7. The intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics according to claim 6, characterized in that: The angle recommendation unit is used to: input the gravity distribution condition to guide the user's sitting posture normalized semantic feature map into the decoder-based support part angle recommender to obtain the angle recommendation decoding value.

8. The intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics according to claim 7, characterized in that: It also includes a training module for training the gravity distribution feature extractor based on the hole convolutional neural network model, the user sitting posture semantic feature extractor based on the AlexNet model, the feature attention selection enhancer under the compression suppression structure, and the decoder-based support part angle recommender.

9. The intelligent spine adjustment health chair based on gravity physics and human body autonomous balance characteristics according to claim 8, characterized in that: The training module comprises: A training data acquisition unit, used to acquire training data, wherein the training data includes a training user's gravity distribution image and a training user's sitting posture image; A training gravity distribution semantic feature extraction unit, used for inputting the training user gravity distribution image into the gravity distribution feature extractor based on the dilated convolutional neural network model to obtain a training gravity distribution semantic feature vector; A training user sitting posture semantic feature extraction unit, used for inputting the training user sitting posture posture image into the user sitting posture semantic feature extractor based on the AlexNet model to obtain a training user sitting posture semantic feature map; A training user sitting posture semantic feature enhancement unit, used for inputting the training user sitting posture semantic feature map into the feature attention selection enhancer under the compression suppression structure to obtain a training enhanced user sitting posture semantic feature map; A training gravity feature assisted sitting posture semantic normalization unit is used to use the training gravity distribution semantic feature vector as auxiliary condition information, and perform conditional normalization based on the auxiliary condition information on the training enhanced user sitting posture semantic feature map to obtain a training gravity distribution condition guided user sitting posture normalized semantic feature map; A decoding loss calculation unit, used for inputting the normalized semantic feature map of the user's sitting posture guided by the training gravity distribution condition into the decoder-based support part angle recommender to obtain a decoding loss function value; A semantic loss calculation unit, used to calculate the gravity distribution condition-guided user sitting posture normalized semantic loss function value of the training gravity distribution condition-guided user sitting posture normalized semantic feature map; A loss training unit is used to guide the weighted sum of the normalized semantic loss function values ​​of the user sitting posture based on the decoding loss function value and the gravity distribution condition as the loss function value and train the gravity distribution feature extractor based on the void convolutional neural network model, the user sitting posture semantic feature extractor based on the AlexNet model, the feature attention selection enhancer under the compression suppression structure and the decoder-based support part angle recommender through back propagation of gradient descent.

Images