A sensorless motor control method and system based on multi-algorithm fusion

Abstract

The present invention provides a multi-algorithm fusion sensorless motor control method and system, which relates to the field of sensorless motor control technology. The method includes determining first massage path information based on multiple first sore and swollen point information input by the user; generating a first motor control scheme; obtaining multiple second sore and swollen point information input by the user; determining multiple self-perceived key massage point information based on the multiple first sore and swollen point information input by the user, the multiple second sore and swollen point information input by the user, and the first massage path information; determining multiple electromyography analysis key massage point information based on the electromyography data of each first sore and swollen point and the first massage path information; constructing a massage map; processing the massage map based on a neural network model to determine the final massage path information; generating a second motor control scheme, and controlling the motor operation based on the second motor control scheme. The method can accurately control the sensorless motor to provide a massage scheme that best fits the user's condition.

Description

Technical Field

[0001] The present invention relates to the technical field of sensorless motor control, and in particular to a multi-algorithm fusion sensorless motor control method and system. Background Art

[0002] In the massage equipment industry, traditional massage solutions primarily rely on preset programs or simple manual settings, lacking effective response to individual user differences and real-time feedback. In existing technologies, most massage devices use fixed massage paths and intensity settings, failing to dynamically adjust based on the user's actual sensations. This approach results in poor massage effectiveness and makes it difficult to accurately relieve soreness and discomfort in specific areas of the user. Existing technologies are unable to fully capture changes in the user's physical condition. Most current systems rely on a single data source, such as subjective user feedback or simple electromyographic signals, lacking intelligent analysis methods that combine subjective user experiences with objective physiological indicators. Existing massage path planning algorithms often use simple linear interpolation or fixed patterns, failing to generate an optimal massage trajectory based on the spatial distribution of soreness and intensity variations. Massage devices commonly suffer from unnatural interactions and delayed feedback. Furthermore, they lack a scientific evaluation mechanism for massage effectiveness, making it impossible to quantify and analyze the actual effectiveness of different massage strategies.

[0003] Therefore, how to accurately control the sensorless motor to provide a massage solution that best suits the user's condition is an urgent problem that needs to be solved. Summary of the Invention

[0004] The main technical problem solved by the present invention is how to accurately control the sensorless motor to provide a massage solution that best suits the user's condition.

[0005] According to a first aspect, the present invention provides a multi-algorithm fusion sensorless motor control method, comprising: obtaining a plurality of first sore and swollen point information input by a user; determining a first massage path information based on the plurality of first sore and swollen point information input by the user; generating a first motor control scheme based on the first massage path information, and controlling the motor operation based on the first motor control scheme, synchronously obtaining electromyography data of each first sore and swollen point during operation; obtaining a plurality of second sore and swollen point information input by the user when the first motor control scheme is completed; determining a plurality of self-perceived key massage point information based on the plurality of first sore and swollen point information input by the user, the plurality of second sore and swollen point information input by the user, and the first massage path information. ; Based on the electromyogram data of each first sore and swollen point and the first massage path information, multiple electromyogram analysis key massage point information is determined; a massage map is constructed, the massage map includes multiple self-perception key massage nodes and multiple electromyogram analysis key massage nodes, the node characteristics of the self-perception key massage nodes are the self-perception key massage point information, the node characteristics of the electromyogram analysis key massage nodes are the electromyogram analysis key massage point information, and the edges between nodes are the distances between nodes; the massage map is processed based on a neural network model to determine the final massage path information; a second motor control scheme is generated based on the final massage path information, and the motor operation is controlled based on the second motor control scheme.

[0006] In one possible implementation, the processing of the massage map based on the neural network model to determine the final massage path information includes: processing the massage map based on the graph neural network to determine the final massage path; and determining the final massage path information based on the multiple second sore and swollen point information input by the user and the final massage path using a path information output model.

[0007] In a possible implementation, the input of the graph neural network is a massage graph, and the output of the graph neural network is a final massage path.

[0008] In a possible implementation, the path information output model is a deep neural network model.

[0009] According to the second aspect, the present invention provides a multi-algorithm fusion sensorless motor control system, including: an acquisition module for acquiring multiple first sore and swollen point information input by a user; a first path determination module for determining first massage path information based on the multiple first sore and swollen point information input by the user; a first control module for generating a first motor control scheme based on the first massage path information, and controlling the motor operation based on the first motor control scheme, and synchronously acquiring electromyography data of each first sore and swollen point during operation; a second acquisition module for acquiring multiple second sore and swollen point information input by the user after the first motor control scheme is completed; a self-feeling key point determination module for determining multiple self-feeling key presses based on the multiple first sore and swollen point information input by the user, the multiple second sore and swollen point information input by the user, and the first massage path information. Massage point information; an electromyogram key point determination module, used to determine multiple electromyogram analysis key massage point information based on the electromyogram data of each first sore and swollen point and the first massage path information; a map construction module, used to construct a massage map, the massage map includes multiple self-perception key massage nodes and multiple electromyogram analysis key massage nodes, the node characteristics of the self-perception key massage nodes are the self-perception key massage point information, the node characteristics of the electromyogram analysis key massage nodes are the electromyogram analysis key massage point information, and the edges between nodes are the distances between nodes; a final path determination module, used to process the massage map based on a neural network model to determine the final massage path information; a second control module, used to generate a second motor control scheme based on the final massage path information, and control the motor operation based on the second motor control scheme.

[0010] In a possible implementation, the final path determination module is further used to: process the massage map based on a graph neural network to determine the final massage path; and determine the final massage path information based on the multiple second sore and swollen point information input by the user and the final massage path using a path information output model.

[0011] In a possible implementation, the input of the graph neural network is a massage graph, and the output of the graph neural network is a final massage path.

[0012] In a possible implementation, the path information output model is a deep neural network model.

[0013] According to a third aspect, an embodiment of the present invention provides an electronic device, comprising: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method comprising: obtaining a plurality of first sore and swollen point information input by a user; determining a first massage path information based on the plurality of first sore and swollen point information input by the user; generating a first motor control scheme based on the first massage path information, and controlling the motor operation based on the first motor control scheme, synchronously obtaining electromyography data of each first sore and swollen point during operation; obtaining a plurality of second sore and swollen point information input by the user when the first motor control scheme is completed; and determining a first massage path information based on the plurality of first sore and swollen point information input by the user and the plurality of second sore and swollen point information input by the user. The swelling point information and the first massage path information determine multiple self-perceived key massage point information; based on the electromyogram data of each first sore and swollen point and the first massage path information, multiple electromyogram analysis key massage point information are determined; a massage map is constructed, and the massage map includes multiple self-perceived key massage nodes and multiple electromyogram analysis key massage nodes, the node characteristics of the self-perceived key massage nodes are the self-perceived key massage point information, the node characteristics of the electromyogram analysis key massage nodes are the electromyogram analysis key massage point information, and the edges between the nodes are the distances between the nodes; the massage map is processed based on a neural network model to determine the final massage path information; a second motor control scheme is generated based on the final massage path information, and the motor operation is controlled based on the second motor control scheme.

[0014] According to the fourth aspect, the present embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned multi-algorithm fusion sensorless motor control method, the method comprising: obtaining a plurality of first sore and swollen point information input by a user; determining a first massage path information based on the plurality of first sore and swollen point information input by the user; generating a first motor control scheme based on the first massage path information, and controlling the motor operation based on the first motor control scheme, synchronously obtaining electromyography data of each first sore and swollen point during operation; obtaining a plurality of second sore and swollen point information input by the user when the first motor control scheme is completed; and determining a first massage path information based on the plurality of first sore and swollen point information input by the user, the plurality of second sore and swollen point information input by the user, and the first massage path information. A massage path information is used to determine multiple self-perceived key massage point information; based on the electromyogram data of each first sore and swollen point and the first massage path information, multiple electromyogram analysis key massage point information is determined; a massage map is constructed, the massage map including multiple self-perceived key massage nodes and multiple electromyogram analysis key massage nodes, the node characteristics of the self-perceived key massage nodes are the self-perceived key massage point information, the node characteristics of the electromyogram analysis key massage nodes are the electromyogram analysis key massage point information, and the edges between the nodes are the distances between the nodes; the massage map is processed based on a neural network model to determine the final massage path information; a second motor control scheme is generated based on the final massage path information, and the motor operation is controlled based on the second motor control scheme.

[0015] The present invention provides a multi-algorithm fusion sensorless motor control method and system, which includes obtaining a plurality of first sore and swollen point information input by a user; determining a first massage path information based on the plurality of first sore and swollen point information input by the user; generating a first motor control scheme based on the first massage path information, and controlling the motor operation based on the first motor control scheme, and synchronously obtaining electromyography data of each first sore and swollen point during operation; obtaining a plurality of second sore and swollen point information input by the user when the first motor control scheme is completed; determining a plurality of self-perceived key massage point information based on the plurality of first sore and swollen point information input by the user, the plurality of second sore and swollen point information input by the user, and the first massage path information; and generating a first motor control scheme based on the first massage path information. The method comprises the following steps: first, determining a plurality of electromyography analysis key massage point information based on the graph data and the first massage path information; constructing a massage atlas, wherein the massage atlas includes a plurality of self-perception key massage nodes and a plurality of electromyography analysis key massage nodes, the node characteristics of the self-perception key massage nodes are the self-perception key massage point information, the node characteristics of the electromyography analysis key massage nodes are the electromyography analysis key massage point information, and the edges between the nodes are the distances between the nodes; processing the massage atlas based on the neural network model to determine the final massage path information; generating a second motor control scheme based on the final massage path information, and controlling the motor operation based on the second motor control scheme. The method can accurately control the sensorless motor to provide a massage scheme that best suits the user's condition. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of a multi-algorithm fusion sensorless motor control method provided by an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of an inductorless motor provided by an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of constructing a detection map provided by an embodiment of the present invention;

[0019] Figure 4 A schematic diagram of a process for determining final massage path information provided by an embodiment of the present invention;

[0020] Figure 5 Schematic diagram of a sensorless motor control system with multi-algorithm fusion provided by an embodiment of the present invention DETAILED DESCRIPTION

[0021] The present invention will be further described in detail below by means of specific embodiments in conjunction with the accompanying drawings. Similar elements in different embodiments are numbered with associated similar elements. In the following embodiments, many detailed descriptions are provided to enable the present invention to be better understood. However, those skilled in the art will readily appreciate that some of the features may be omitted under different circumstances, or may be replaced by other elements, materials, or methods. In some cases, some operations related to the present invention are not shown or described in the specification. This is to avoid the core of the present invention being overwhelmed by excessive descriptions, and for those skilled in the art, it is not necessary to describe these related operations in detail. They can fully understand the related operations based on the description in the specification and the general technical knowledge in the art.

[0022] In an embodiment of the present invention, there is provided Figure 1 A sensorless motor control method using multiple algorithms is shown, and the sensorless motor control method using multiple algorithms includes steps S1 to S9:

[0023] Step S1: Acquire multiple first sore swelling point information input by a user.

[0024] The multiple first soreness and swelling points are a collection of information about the locations of soreness and swelling on the body that the user has marked based on their own feelings through an interactive device including a touch screen and sensors. Each first soreness and swelling point information includes the specific location of the point and the soreness and swelling intensity level at that point.

[0025] The soreness and swelling intensity level can be expressed by a value of 1-5, with the larger the value, the higher the soreness and swelling degree.

[0026] Step S2: determining first massage path information based on the plurality of first sore and swollen point information input by the user.

[0027] In some embodiments, the first massage path information can be determined using a first information determination model based on the multiple first sore and swollen point information input by the user. The first information determination model is a deep neural network. The input of the first information determination model is the multiple first sore and swollen point information input by the user, and the output of the first information determination model is the first massage path information.

[0028] A deep neural network is an artificial neural network composed of multiple hidden layers. Through its hierarchical structure, a deep neural network can perform nonlinear transformations and feature extraction on input data. Deep neural networks are capable of learning highly nonlinear mappings between input features and complex output targets. During training, network weights are adjusted using a backpropagation algorithm, enabling the deep neural network to predict a sequence of massage trajectories that effectively relieve multiple initial soreness and swelling points based on the input information.

[0029] The first massage path information is massage execution information for user sensory experience output by the first information determination model. The first massage path information includes the execution path sequence of massage parts, massage time allocation, operation techniques, massage intensity and other information.

[0030] The time allocation for massage includes the total massage time and the start and end time of the operation on each part.

[0031] Manipulative techniques refer to the types of techniques applied during each time period, such as kneading, pressing, vibration, etc.

[0032] Massage intensity is the level of force that the user can perceive, such as medium kneading, deep pressing, etc.

[0033] The user inputs information about multiple first soreness and swelling points, including the location and intensity of the user's perceived soreness and swelling. This information directly reflects the target area requiring massage intervention and its urgency. The location information provides the spatial distribution of the massage path, while the intensity information suggests that different points may require different massage pressures or durations. A deep neural network model learns implicit, continuous path planning rules that align with ergonomics and massage principles from this discrete point information.

[0034] Through multiple layers of nonlinear processing units, a deep neural network performs in-depth feature extraction and pattern recognition on multiple primary pain and swelling points input by the user. The model learns the spatial correlations between pain and swelling points, such as how adjacent points may form a path. Furthermore, intensity differences influence path priority, such as prioritizing or focusing treatment on high-intensity points. The deep neural network effectively connects these points into a smooth, efficient massage trajectory sequence that covers key areas, ultimately outputting structured primary massage path information.

[0035] Step S3: generating a first motor control scheme based on the first massage path information, and controlling the motor to operate based on the first motor control scheme, and synchronously acquiring electromyography data of each first sore and swollen point during operation.

[0036] The motor is a sensorless motor. A sensorless motor does not require additional physical position sensors such as Hall sensors or encoders. Instead, it can estimate the rotor position and speed in real time using a built-in algorithm. Figure 2 A schematic diagram of an inductorless motor provided in an embodiment of the present invention.

[0037] In some embodiments, a first motor control scheme can be generated based on the first massage path information using a control scheme generation model, wherein the control scheme generation model is a generative adversarial network, the input of the control scheme generation model is the first massage path information, and the output of the control scheme generation model is the first motor control scheme.

[0038] A generative adversarial network (GAN) is a deep learning model consisting of two sub-networks: a generator and a discriminator. The generator receives input data and attempts to generate realistic output samples, while the discriminator distinguishes between the generator's data and real samples. During training, the generator and discriminator compete with each other, which in turn drives the generator to continuously improve its generation capabilities.

[0039] The first motor control scheme is a set of specific instructions generated by the control scheme generation model to directly drive the massage device's motor. This scheme includes the motor's operations at various time points, including the motor shaft angle control sequence, torque output values ​​(N·m) at each time point, vibration motor frequency parameters (Hz), the massage head's three-dimensional movement speed curve (mm / s), and acceleration thresholds.

[0040] The electromyographic (EMG) data for each first sore point is synchronously collected by the EMG sensor while the massage device's motor operates according to the first motor control scheme. It reflects the electrophysiological activity of the muscle groups corresponding to the multiple first sore points marked by the user. The EMG data for each first sore point is a time series signal that records the voltage fluctuations of the corresponding muscles over time under massage stimulation. The generator model of the generative adversarial network (GAN) learns the complex mapping relationship between the user's sensory massage scheme and the motor control instruction set, thereby mastering the inherent rules for translating abstract sensory needs into physical execution parameters. The generator receives the first massage path information as input and, based on the massage path's inherent sequence of massage points, manipulation type, subjective intensity level, and time allocation constraints, infers the precise set of motor control parameters required to achieve the sensory effect, such as the target position sequence, movement speed, acceleration, output torque, vibration frequency and amplitude of each motor axis. The discriminator model then evaluates the authenticity and feasibility of the candidate motor control schemes output by the generator, and determines whether they comply with real-world device response characteristics and safety specifications. Through continuous adversarial training and optimization of the generator and the discriminator, the final model can output the first motor control solution that can accurately achieve the sensory effect described by the first massage path information at the physical level, while also having device executability and safety guarantees.

[0041] Step S4: After the first motor control solution is completed, a plurality of second soreness and swelling point information input by the user is obtained.

[0042] The multiple second soreness and swelling point information input by the user is a collection of information about the soreness and swelling points on the body that the user marked again on the interactive device interface after the first motor control solution is executed. Each second soreness and swelling point information includes the location coordinates and soreness intensity level of each point.

[0043] By having the user input information about a plurality of second soreness and swelling points, the execution effect of the first motor control scheme can be evaluated, and changes in the user's soreness and swelling sensation, such as relief, transfer, or new appearance, can be captured.

[0044] Step S5: determining a plurality of self-perceived key massage point information based on the plurality of first sore and swollen point information input by the user, the plurality of second sore and swollen point information input by the user, and the first massage path information.

[0045] In some embodiments, determining a plurality of self-perceived key massage point information based on the plurality of first sore and swollen point information input by the user, the plurality of second sore and swollen point information input by the user, and the first massage path information includes steps S51-S53:

[0046] Step S51 : determining multiple pieces of sore swelling area information, sore swelling migration rules, and potential sore swelling areas based on the multiple pieces of first sore swelling point information input by the user and the multiple pieces of second sore swelling point information input by the user.

[0047] In some embodiments, a deep neural network may be used to determine multiple sore swelling area information, sore swelling migration rules, and potential sore swelling areas based on multiple first sore swelling point information and multiple second sore swelling point information input by the user.

[0048] The sour swelling area information includes the center position coordinates, area, average sour swelling intensity level and area boundary outline of each sour swelling area.

[0049] The migration law of acid swelling includes the intensity change trend of the original acid swelling point, the spatial correlation between the new acid swelling point and the original acid swelling point, and the transfer path of the acid swelling relief area.

[0050] Potential soreness and swelling regions are areas of the body that may develop soreness and swelling in the future, predicted using a deep neural network. These regions include the predicted coordinates of the region's location, the estimated level of potential soreness and swelling, and the probability of the region forming.

[0051] The multiple first soreness and swelling points input by the user provide the spatial distribution and severity benchmark of the initial swelling points, where the location distribution and intensity gradient of the soreness and swelling points can identify the core discomfort area. The multiple second soreness and swelling points input by the user can reflect the change in soreness and swelling after massage intervention, and can deduce the intensity change and transfer direction of the soreness and swelling. By extracting the spatial clustering characteristics and temporal change patterns of these soreness and swelling points, the deep neural network can aggregate discrete points into continuous soreness and swelling area information, explore the soreness and swelling migration patterns from intensity fluctuations, and predict potential soreness and swelling areas based on the correlation between adjacent areas.

[0052] Step S52: determining an effect information sequence of each massage point in the first massage path based on the information of the multiple sore and swollen areas, the sore and swollen migration rule, and the first massage path information.

[0053] In some embodiments, a deep neural network may be used to determine an effect information sequence for each massage point in the first massage path based on the multiple sore and swollen area information, the sore and swollen migration rules, and the first massage path information.

[0054] The effect information sequence for each massage point in the first massage path is a time series information set that comprehensively evaluates the intervention effect of each manipulation point recorded in the first massage path information. The effect information sequence for each massage point in the first massage path includes the starting time of each point, the duration of the intervention, the type of manipulation performed, the change in intensity of the soreness and swelling area associated with the point, and the effect score.

[0055] Deep neural networks can use attention mechanisms to focus on massage locations that are critical for relieving soreness and swelling, and capture the cumulative effects of massage through time series modeling. The model can learn the effectiveness of massage points on sore areas, generating a quantified sequence of effect information for each massage point.

[0056] Step S53: determining a plurality of self-perceived key massage point information based on the plurality of sore and swollen area information, the sore and swollen migration pattern, the potential sore and swollen areas, and the effect information sequence of each massage point in the first massage path.

[0057] In some embodiments, a deep neural network can be used to determine multiple key massage point information based on the multiple sore and swollen area information, the sore and swollen migration rules, the potential sore and swollen areas, and the effect information sequence of each massage point in the first massage path.

[0058] Multiple self-perceived key massage points are a collection of massage locations that contribute significantly to relieving soreness and swelling, identified through deep neural networks. This information includes key point coordinates, criticality scores, and associated soreness and swelling area identifiers.

[0059] By establishing an evidence fusion layer, the deep neural network performs a preliminary criticality screening of the effect information sequence of each massage point in the first massage path. It then weights the data based on the regional severity and migration trends of multiple soreness and swelling areas. For points located within potential soreness and swelling areas, the deep neural network increases their criticality score using a preventative gain factor. Finally, a gating mechanism filters out low-contribution points and outputs multiple, self-perceived, key massage point information with precise location and quantitative criticality.

[0060] Step S6: determining a plurality of electromyography analysis key massage point information based on the electromyography data of each first sore and swollen point and the first massage path information.

[0061] In some embodiments, a key point analysis model can be used to determine multiple electromyography analysis key massage point information based on the electromyography data of each first sore and swollen point and the first massage path information. The key point analysis model is a deep neural network. The input of the key point analysis model is the electromyography data of each first sore and swollen point and the first massage path information. The output of the key point analysis model is multiple electromyography analysis key massage point information.

[0062] Multiple EMG analysis key massage point information is obtained by analyzing the EMG data of each first soreness point and the first massage path information using a key point analysis model. The information then identifies, based on objective physiological data, massage locations within the first massage path that have a significant positive impact on muscle condition. This EMG analysis key massage point information includes the location of the key points, the amplitude of the EMG signal change, the degree of muscle relaxation, and the importance of the key points.

[0063] The EMG data from the first sore point records the dynamic changes in muscle electrical activity during the massage. Combining this data with the timing of the movements in the first massage path information forms a time series correlation from massage action to EMG response, providing objective physiological indicator features for the deep neural network. By processing the timing features of the EMG data from the first sore point and the movement parameters of the massage path in the first massage path information, the deep neural network can identify key massage points that significantly improve the EMG signal.

[0064] Step S7, constructing a massage map, wherein the massage map includes multiple self-perception key massage nodes and multiple electromyography analysis key massage nodes, the node characteristics of the self-perception key massage nodes are the self-perception key massage point information, the node characteristics of the electromyography analysis key massage nodes are the electromyography analysis key massage point information, and the edges between the nodes are the distances between the nodes.

[0065] A massage graph is a data structure used to aggregate and represent key massage points and their relationships. The massage graph contains multiple self-perceived key massage nodes and multiple electromyography (EMG) analysis key massage nodes. Each self-perceived key massage node is characterized by information about the self-perceived key massage point. Each EMG analysis key massage node is characterized by information about the EMG analysis key massage point. The edge connecting any two nodes in the massage graph represents the spatial distance between the two nodes. Figure 3 A schematic diagram of constructing a detection spectrum provided by an embodiment of the present invention. Figure 3 As shown, Figure 3It includes self-perception key massage node A, self-perception key massage node B, electromyography analysis key massage node C, and electromyography analysis key massage node D, where the edge between each two nodes is the distance between the two key massage nodes.

[0066] Step S8: Processing the massage map based on a neural network model to determine final massage path information.

[0067] In some embodiments, Figure 4 A schematic diagram of a process for determining final massage path information provided by an embodiment of the present invention, wherein determining the final massage path information includes steps S81 to S82:

[0068] Step S81, processing the massage atlas based on the graph neural network to determine the final massage path;

[0069] Graph neural networks (GNNs) are a type of deep learning model that can process graph-structured data. Through a message-passing mechanism, GNNs enable nodes in graph data to aggregate information about their neighboring nodes. Nodes then update their feature representations by aggregating neighbor information. After multiple iterations, each node incorporates both structural information from the graph and neighbor feature information within a certain range. This enables GNNs to effectively capture the complex topological relationships, dependencies, and global structural features between nodes in graph data, enabling them to perform operations such as node classification, link prediction, and level prediction on graph data. The input of the GNN is a massage graph, and the output is the final massage path.

[0070] The final massage path is the optimal sequence of massage points determined by the graph neural network on the massage graph. The final massage path includes the order of key points that the massage head needs to visit.

[0071] In the massage graph, each node represents a key massage point in the first massage path. Its node features are derived from the user's subjective feedback and objective physiological data, respectively. Together, they identify the points that play a key role in relieving soreness and swelling, and their relative contribution weights. Edges between nodes represent the spatial distance relationships between key massage points. This distance information reflects the anatomical proximity of these massage points. Constructing such a massage graph integrates and structures the key points and their spatial relationships derived from both subjective and objective analyses. Processing this massage graph through a graph neural network optimizes the final massage path. This helps plan an efficient and anatomically appropriate massage route, thereby improving overall massage effectiveness. Through its graph convolution operations, the graph neural network aggregates node features and information from neighboring nodes, generating comprehensive node embeddings that represent both global spatial relationships and local importance distributions. These embeddings encompass all key points in the graph and the spatial dependencies between them. They provide the structured information foundation for the graph neural network to generate a natural, coherent, and focused massage path sequence.

[0072] Step S82: determining final massage path information based on the information of the plurality of second sore and swollen points input by the user and the final massage path using a path information output model.

[0073] The path information output model is a deep neural network model. The input of the path information output model is the multiple second sore and swollen point information input by the user and the final massage path usage path information. The output of the path information output model is the final massage path information.

[0074] The final massage path information is refined massage execution information generated by the path information output model based on the final massage path. The final massage path information includes the execution path sequence of massage parts, massage time allocation, operation techniques, massage intensity and other information.

[0075] The multiple second soreness and swelling points information input by the user is the latest soreness and swelling status feedback after the execution of the first massage path, providing a basis for the refinement of the path information, such as indicating which key points need to increase the intensity of massage, prolong the action time, or supplement transitional techniques between key points for residual soreness and swelling. The final massage path provides a basic key point sequence framework, clarifying the core nodes and sequence of massage. Through multi-layer feature extraction, the deep neural network can fuse the intensity characteristics and spatial distribution of the second soreness and swelling points with the node sequence and spatial association of the final massage path, learn the matching relationship between key points and residual soreness and swelling, and the rules of manipulation parameters corresponding to soreness and swelling of different intensities, and then transform the fixed path framework and dynamic soreness and swelling feedback into structured final massage path information with specific time allocation, manipulation type and intensity level.

[0076] Step S9: generating a second motor control scheme based on the final massage path information, and controlling the motor operation based on the second motor control scheme.

[0077] In some embodiments, a second motor control scheme can be generated based on the final massage path information using a second scheme generation model, wherein the second scheme generation model is a generative adversarial network, the input of the second scheme generation model is the final massage path information, and the output of the second scheme generation model is a second motor control scheme.

[0078] The second motor control scheme is a set of specific instructions for directly driving the massage device motor, generated by the second scheme generation model to achieve the sensory effect described by the final massage path information. The second motor control scheme includes the operations that the motor needs to perform at various time points.

[0079] The generator of the generative adversarial network, through its encoder and decoder structure, decodes the time allocation parameters in the final massage path information into motor timing instructions, encodes the force level into a torque output curve, and maps the manipulation type into motion pattern parameters. The discriminator verifies the feasibility of the solution by combining it with the motor physical constraint library. Through adversarial training, it optimizes the output to ensure that the second motor control solution accurately achieves the sensory effect required by the final massage path information.

[0080] Based on the same inventive concept, Figure 5 A schematic diagram of a multi-algorithm fusion sensorless motor control system provided by an embodiment of the present invention, wherein the multi-algorithm fusion sensorless motor control system includes:

[0081] An acquisition module 91 is configured to acquire a plurality of first soreness and swelling point information input by a user;

[0082] A first path determining module 92 is configured to determine first massage path information based on the plurality of first sore and swollen point information input by the user;

[0083] A first control module 93 is configured to generate a first motor control scheme based on the first massage path information, control the motor operation based on the first motor control scheme, and synchronously obtain electromyography data of each first sore and swollen point during operation;

[0084] A second acquisition module 94 is configured to acquire a plurality of second soreness and swelling point information input by a user after the first motor control scheme is completed;

[0085] a self-perceived key point determination module 95 for determining a plurality of self-perceived key massage point information based on the plurality of first sore and swollen point information input by the user, the plurality of second sore and swollen point information input by the user, and the first massage path information;

[0086] an electromyography key point determination module 96 for determining a plurality of electromyography key massage point information based on the electromyography data of each first sore and swollen point and the first massage path information;

[0087] a map construction module 97 for constructing a massage map, wherein the massage map includes a plurality of self-perception key massage nodes and a plurality of electromyography analysis key massage nodes, wherein the node features of the self-perception key massage nodes are the self-perception key massage point information, the node features of the electromyography analysis key massage nodes are the electromyography analysis key massage point information, and the edges between the nodes are the distances between the nodes;

[0088] a final path determination module 98, configured to process the massage atlas based on a neural network model to determine final massage path information;

[0089] The second control module 99 is configured to generate a second motor control scheme based on the final massage path information, and control the motor operation based on the second motor control scheme.

[0090] It should be noted that, in order to simplify the presentation of this specification and facilitate understanding of one or more embodiments of the invention, the foregoing descriptions of the embodiments of this specification sometimes combine multiple features into a single embodiment, figure, or description thereof. However, this disclosure method does not imply that the subject matter of this specification requires more features than those recited in the claims. In fact, an embodiment may have fewer features than all of the features of a single disclosed embodiment.

[0091] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. A multi-algorithm fusion sensorless motor control method, characterized in that: include: Acquire multiple first soreness and swelling point information input by the user; determining first massage path information based on the plurality of first sore and swollen point information input by the user; generating a first motor control scheme based on the first massage path information, and controlling the motor to operate based on the first motor control scheme, while synchronously acquiring electromyography data of each first sore and swollen point during the operation; After the first motor control scheme is completed, a plurality of second soreness and swelling point information input by the user is obtained; Determining a plurality of self-perceived key massage point information based on the plurality of first sore and swollen point information input by the user, the plurality of second sore and swollen point information input by the user, and the first massage path information; Determine multiple electromyography analysis key massage point information based on the electromyography data of each first sore and swollen point and the first massage path information; Constructing a massage atlas, the massage atlas including a plurality of self-perception key massage nodes and a plurality of electromyography analysis key massage nodes, wherein the node features of the self-perception key massage nodes are self-perception key massage point information, the node features of the electromyography analysis key massage nodes are electromyography analysis key massage point information, and the edges between the nodes are the distances between the nodes; Processing the massage atlas based on a neural network model to determine final massage path information; A second motor control scheme is generated based on the final massage path information, and the motor operation is controlled based on the second motor control scheme.

2. The multi-algorithm fusion sensorless motor control method according to claim 1, characterized in that: The processing of the massage atlas based on the neural network model to determine the final massage path information includes: The massage map is processed based on the graph neural network to determine the final massage path; Based on the multiple second sore and swollen point information input by the user, the final massage path information is determined using a path information output model. The final massage path information includes the execution path sequence of the massage parts, the time allocation of the massage, the operation method, and the massage intensity.

3. The multi-algorithm fusion sensorless motor control method according to claim 2, characterized in that: The input of the graph neural network is the massage atlas, and the output of the graph neural network is the final massage path.

4. The multi-algorithm fusion sensorless motor control method according to claim 2, characterized in that: The path information output model is a deep neural network model.

5. A sensorless motor control system with multi-algorithm fusion, characterized in that: include: An acquisition module, configured to acquire a plurality of first sore swelling point information input by a user; a first path determining module, configured to determine first massage path information based on the plurality of first sore and swollen point information input by the user; a first control module, configured to generate a first motor control scheme based on the first massage path information, and control the motor operation based on the first motor control scheme, and synchronously obtain electromyography data of each first sore and swollen point during operation; A second acquisition module is used to acquire a plurality of second soreness and swelling point information input by a user after the first motor control scheme is completed; a self-perceived key point determination module, configured to determine a plurality of self-perceived key massage point information based on the plurality of first sore and swollen point information input by the user, the plurality of second sore and swollen point information input by the user, and the first massage path information; an electromyography key point determination module, configured to determine a plurality of electromyography analysis key massage point information based on the electromyography data of each first sore and swollen point and the first massage path information; a map construction module for constructing a massage map, wherein the massage map includes a plurality of self-perception key massage nodes and a plurality of electromyography analysis key massage nodes, wherein the node features of the self-perception key massage nodes are the self-perception key massage point information, the node features of the electromyography analysis key massage nodes are the electromyography analysis key massage point information, and the edges between the nodes are the distances between the nodes; a final path determination module, configured to process the massage atlas based on a neural network model to determine final massage path information, wherein the final massage path information includes the execution path sequence of the massage parts, the time allocation of the massage, the operation technique, and the massage intensity; The second control module is configured to generate a second motor control scheme based on the final massage path information, and control the operation of the motor based on the second motor control scheme.

6. The multi-algorithm fusion sensorless motor control system according to claim 5, characterized in that: The final path determination module is further configured to: The massage map is processed based on the graph neural network to determine the final massage path; Based on the multiple second sore and swollen point information input by the user, the final massage path information is determined using a path information output model. The final massage path information includes the execution path sequence of the massage parts, the time allocation of the massage, the operation method, and the massage intensity.

7. The multi-algorithm fusion sensorless motor control system according to claim 6, characterized in that: The input of the graph neural network is the massage atlas, and the output of the graph neural network is the final massage path.

8. The multi-algorithm fusion sensorless motor control system according to claim 6, characterized in that: The path information output model is a deep neural network model.

9. An electronic device, characterized in that: include: processor; Memory; And a computer program; wherein, the computer program is stored in the memory and is configured to be executed by the processor to implement the multi-algorithm fusion sensorless motor control method according to any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the sensorless motor control method with multi-algorithm fusion as claimed in any one of claims 1 to 4 is implemented.

Images