Portable traditional Chinese medicine massage teaching and testing instrument based on artificial intelligence

Abstract

A traditional Chinese medicine massage teaching and testing instrument based on artificial intelligence comprises a similar 3D human skin simulation model, a massage table with a sliding mechanism provided with double cameras, a data processing module, a host, a display device and a network communication device. Sensors are evenly distributed on the similar 3D human skin simulation model, can sense external acting force and are connected with an input port of the data processing module, an output port of the data processing module is connected with an external equipment interface of the host, and the display device is connected with the host and used for displaying dynamic massage operation and acting force distribution data based on an intelligent operation system interface. The teaching and testing instrument can realize an intelligent algorithm and a digital twinning algorithm which are respectively used for constructing an artificial intelligence model and tracking, focusing and simulating dynamic demonstration. Therefore, the evaluation score of the action is obtained, and repeated watching and learning are facilitated. And a diversified, scientific and accurate massage teaching and examination system is established.

Description

Technical Field

[0001] The invention relates to a Chinese medicine massage teaching and examination instrument, in particular to a portable Chinese medicine massage teaching and examination instrument based on artificial intelligence, and belongs to the teaching application field of Chinese medicine massage skills. Background Art

[0002] Traditional Chinese medicine massage is the main external treatment method of traditional Chinese medicine. It is based on the theory of traditional Chinese medicine meridians and acupoints, combined with the anatomy and pathological diagnosis of Western medicine. It uses pushing, holding, lifting, pinching, kneading and other techniques to act on the characteristic parts of the human body surface to regulate the body's physiological and pathological conditions and achieve the purpose of physical therapy. The operator's control over the force position, strength, amplitude, frequency, rhythm and other related movements of the massage technique directly affects the therapeutic effect of the massage and the physical experience of the patient.

[0003] The teaching and practice of traditional Chinese medicine massage techniques mainly focus on theoretical explanation and action demonstration, combined with the operator's actual practice. Since the theoretical explanation is mainly in text, for example, the basic requirements of the one-finger Zen technique are "persistent, powerful, even, and gentle", although the teacher can clearly express the connotation of the text during the explanation, it is difficult for beginners to understand the text thoroughly and transform it into actual actions. Therefore, by accurately collecting the action data of the massage technique through sensor technology and simultaneously displaying the data through docking with an external data display system, the accuracy and intuitiveness of the massage technique teaching can be effectively improved, providing practitioners with a scientific and precise massage technique learning and practice environment, and also providing teachers with an objective basis for judging the effect of the massage technique learning.

[0004] Conventional massage technique data collection is mainly based on the average pressure of the region. The corresponding data collection and display system is used to display the relationship between the pressure, frequency and time of the massage technique. However, since the massage technique contains many categories and is closely related to the contact part and contact area, traditional massage requires hands-on teaching by a master. There is also a problem of succession, and not all students can learn the essence of the massage school. Moreover, in reality, the real inheritors of massage cannot teach the general public face to face. Therefore, how to judge whether an ordinary student has mastered the specialized skills of various schools and how to evaluate them has become an urgent problem to be solved in this field. At the same time, in combination with the requirements of the national education system reform plan, the cross-domain integration of TCM massage skill teaching with sensor technology, skin-like material technology and other scientific and technological means is the most effective means to realize the digitization, precision and standardization of TCM teaching, promote scientific and technological innovation of traditional TCM teaching methods, and shorten the time cycle for the cultivation of qualified TCM talents.

[0005] On the other hand, the teaching and testing instrument is a kind of massage evaluation instrument that is specially designed to allow students to repeat the techniques by recording the techniques demonstrated by the master. The existing technology uses a soft material that simulates the material of the human body to perform massage exercises and assessment operations. However, this material is generally made into a flat surface, which cannot be compared with the real 3D shape of the human body. Therefore, the massage data is incomplete, and students and candidates cannot master the feel. Although the plane is suitable for most techniques, such as kneading, rubbing, rolling, patting, tapping, and rubbing; however, some pinching, holding, and lifting operations that need to be in the vertical direction cannot be recognized, thus limiting the comprehensive learning and evaluation of the techniques of the massage school.

[0006] From the algorithmic point of view, the existing technology cannot recognize 3D scenes. Although 3D recognition models already exist, they are only applicable to complex facial situations. For 3D recognition objects with simple contours such as human torso and limbs, using 3D complex algorithms will only increase the computing load of the instrument. Therefore, algorithm improvement is also a problem.

[0007] Therefore, there is an urgent need in the prior art for a massage teaching and testing instrument that can use 3D dimensionality reduction to 2D and utilize 2D models to solve 3D problems. Summary of the invention

[0008] In view of the above problems of the prior art, the present invention considers two directions. First, a 3D human model is used to simulate the human body massage object, comprehensively obtain 3D data, and reduce the 3D dimension to 2D algorithm, and use the 2D model to solve the 3D problem. Second, a portable structure is used to collect various manipulation data, and the artificial intelligence algorithm is used to find out the rules of each manipulation, and form the standard of massage manipulation for learners to refer to, so as to further facilitate the needs of teaching and examination.

[0009] Based on the above-mentioned direction, the present invention proposes a portable Chinese medicine massage teaching and examination instrument based on artificial intelligence, including a 3D simulated human skin simulation model, a massage table, a data processing module, a host, and a display device, wherein the 3D simulated human skin simulation model is a 3D human skin simulation model with the front or back facing up, and is divided into a simulated skin layer, a simulated fat layer and a simulated muscle layer from top to bottom. The 3D simulated human skin simulation model is fixed on the massage table, and sensors are distributed on the simulated fat layer, which can sense external forces and are connected to the input port of the data processing module, while the output port is connected to the external device interface of the host. The display device is connected to the host and is used to display the dynamics of the massage operation based on the intelligent operating system interface, and the force is divided into a simulated skin layer, a simulated fat layer and a simulated muscle layer. The host has a memory in which an intelligent algorithm application is installed. A camera is arranged on or around the massage table to capture the massage action. The captured massage action is virtually imaged on the interface through the digital twin algorithm of the intelligent algorithm application to realize the dynamic state. The intelligent algorithm application also performs 3D dimension reduction on the perception data received by the host from the sensor into a 2D force distribution map, and builds an artificial intelligence model based on the 2D force distribution map to identify the difference between the real-time massage action and the preset technique standard, thereby obtaining the evaluation result of the action.

[0010] Optionally, each layer of the quasi-3D human skin simulation model is made of silicone and can be combined or split in a detachable structure. The sensor is buried between the layers through the combination, and then detachably fixed to the rigid support structure on the massage table to form a simulated three-dimensional surface of the front and back of the human body.

[0011] Preferably, the sensor is buried corresponding to the acupuncture points of the human body, or buried in a manner of setting at least one of the corresponding acupuncture points of the human body and within a range of 2-3 cm around the corresponding acupuncture points of the human body.

[0012] It is easy to understand that by setting sensors at acupoints, it is possible to identify whether students or test takers have the basic skills to locate acupoints during the learning or examination process.

[0013] Optionally, one side of the massage table has a sliding mechanism, and there are two cameras installed on the sliding mechanism. The host controls the sliding mechanism according to the sensing data, so that the two cameras move synchronously or independently to one side of the massaged part to shoot, so as to capture the massage action.

[0014] The sliding mechanism has a slider and a rotary encoder, and the method in which the host controls the sliding mechanism according to the sensed data comprises the following steps: S1 numbers each sensor, sets the sensing data to data with number information, and outputs the digital quantity increment of the rotary encoder that should be generated when the slider on the sliding mechanism moves to the side of the massage position. With number Build the mapping: , is the mapping function; After receiving the data with number information, the S2 host calculates the , and control the slider to move until the digital output increment of the rotary encoder is .

[0015] Optionally, the force distribution data includes the magnitude of the force and the acupoint name converted by the host according to the number.

[0016] Optionally, the intelligent algorithm application further reduces the 3D dimension of the sensor data received by the host into a 2D force distribution map, and constructs an artificial intelligence model based on the 2D force distribution map, specifically including the following steps: Q1 establishes an XOY two-dimensional rectangular coordinate system, where the X direction is from the head to the feet of the human body, and the Y direction is the left and right direction. A blank model of the force distribution diagram is constructed in the two-dimensional rectangular coordinate system. The blank model is composed of multiple units corresponding to the positions of acupoints, and an array is formed according to the X and Y directions. It is stipulated that the units corresponding to the acupoints are projected on the two-dimensional rectangular coordinate system to obtain the array position.

[0017] This forms a dimensionality reduction, which makes it easy for teachers or examiners to identify. Even if it is not 3D, they can still imagine where it is in space.

[0018] It should be noted that if the X positions of multiple acupuncture points at different Y positions and / or the 2-3 cm range around the acupuncture points are staggered, the unit positions of the array are also staggered accordingly, that is, the array is not a matrix or a square array at this time. In this way, the spatial position distribution of the sensors set in the real acupuncture points and ranges is simulated as much as possible.

[0019] Q2 performs pseudo-color calibration on the sensed data and assigns the corresponding pseudo-color value to the corresponding unit to obtain a real-time 2D force distribution map; Q3: Let masters (teachers or examiners) of different schools and students of high, medium and elementary levels perform massage operations on each human body part on two groups of quasi-3D human skin simulation models, and collect the sensed data in the process of timed triggering. According to step Q2, multiple 2D force distribution maps of each school and each level of students are formed, and the multiple 2D force distribution maps are divided into training sets and verification sets. According to the three timed triggering time periods of the front section, middle section and back section of the completion process of a manual massage action on each human body part, the corresponding training sets and verification sets are respectively classified into the front section class, the middle section class and the back section class; Q4 builds a convolutional neural network group, which is divided into multiple groups. The output of each network group is fully connected and input into softmax for 12 classifications of time periods and grades, that is, it is divided into three time periods, each time period has a master and four levels: high, medium and elementary. The verification set is used for verification until the recognition accuracy is stable, and all groups are trained and a technique artificial intelligence model is obtained.

[0020] Q5 then builds other artificial intelligence models based on Q3-Q4.

[0021] Optionally, masters from different schools and advanced, intermediate and elementary level students are asked to perform massage operations on each part of the human body for a specified duration on two groups of 3D human skin simulation models, re-time the triggering of the perception data of the initial, intermediate and final segments of the collection time, and build an artificial intelligence model based on Q3-Q5 to determine whether the weight of the massage in segments of different lengths meets the corresponding level qualification standards.

[0022] This will further examine the students and candidates' understanding of the law of force changes in each massage technique. In this way, the difference between the massage action and the preset technique standard can be identified from the two dimensions of level and severity based on the real-time captured massage action.

[0023] Preferably, the convolutional neural network is a convolutional neural network with a residual mechanism, and one or more convolutional neural networks with a residual mechanism are trained for each time period and / or each duration segment.

[0024] Optionally, the force distribution data includes a real-time 2D force distribution diagram, and the current pressure converted from the perception data of each sensor, and the preset parameters include any one of the following parameters or a combination thereof: Number of manipulations: record the cumulative number of times generated during the entire practice process, and each force is recorded as one force area: record the area generated by the real-time force of the manipulation, based on the pressure applied each time; Manipulation time: record the cumulative practice time of the manipulation practice from 0 to the end; Manipulation rate: rate = distance of manipulation movement / time used; practice techniques, time of manipulation practice, reference to the master, start / end practice, and practice countdown.

[0025] The content displayed on the intelligent operating system interface also includes the essentials of the technique: describing the operational requirements of the selected technique, providing students with detailed text descriptions of the practical methods when practicing; technique stability: comparing the recorded techniques of expert teachers, and comparing the frequency, pressure, speed, rhythm and other aspects to obtain stability parameters. The calculation method combines three data for calculation; force statistics: recording the maximum force value, minimum force value, average pressure value, and pressure stability during the operation. And record the distribution of all pressures during the practice, according to the pressure characteristics of each technique; and technique video / force distribution: capture real-time techniques through the camera, transform them into digitized hand shapes, and display the recorded technique operation style in real time.

[0026] Optionally, the evaluation score calculation method is: T1, within the specified time, record the number of manipulations performed on the 3D human skin simulation model, the area of ​​force generated, the manipulation time, and the manipulation speed, and compare them with the corresponding data of the masters of each school, and use the percentage value as the score of this item. T2, within the prescribed time, collect the perception data of the front, middle, back, initial, intermediate and final sections of each technique, form a 2D force distribution map, and substitute it into the trained technique artificial intelligence model and weight artificial intelligence model respectively, to determine the corresponding level of the technique and the law of weight change to determine whether it is qualified, and score based on the level and whether it is qualified or not. T3 outputs the average of each score in T1 and the corresponding level and severity of the pass or fail judgment in T2 as the final score.

[0027] Optionally, the teaching and examination instrument further includes a network communication device for communicating with a cloud server, and the virtual image and / or 2D force distribution diagram is also uploaded to the cloud server through the network communication device for downloading and viewing by the teacher or examiner for manual evaluation. Therefore, instead of meeting face to face, the teacher and the examiner can participate in the evaluation of the massage operation of the trainee or examinee remotely online, thereby more objectively evaluating the teaching results through the combination with the instrument.

[0028] Beneficial Effects 1. The use of 3D human skin simulation models to replace flat materials can better perceive the massage feel in three-dimensional space, which is closer to real operations and improves the actual effectiveness of learning and evaluation.

[0029] 2. Use 3D dimension reduction to 2D distribution to make intelligent predictions on the levels of different martial arts techniques. At the same time, examine the rules of severity and weight, and combine with manual evaluation to establish a diversified, scientific and precise massage teaching and examination system.

[0030] 3. The digital twin algorithm captures the massage action into a virtual image on the interface to achieve a dynamic simulation teaching demonstration.

[0031] 4. By using dual cameras on one side of the massage table instead of the traditional video recording with selected camera positions, it is possible to better track and focus on the massaged parts, conduct simulated dynamic demonstrations of detailed techniques, and show the essentials of the movements more clearly. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 (a) is a schematic diagram of a top view showing a portable Chinese medicine massage teaching and examination instrument based on artificial intelligence in a fully opened state according to a specific embodiment 1 of the present invention; Figure 1 (b) is an overall stereoscopic view showing a half-opened box of a portable Chinese medicine massage teaching and examination instrument based on artificial intelligence according to a specific embodiment 1 of the present invention, Figure 2 To illustrate the anatomical diagram of a 3D human skin simulation model of a portable Chinese medicine massage teaching and examination instrument based on artificial intelligence according to a specific embodiment 1 of the present invention, Figure 3 A is a schematic diagram of the back side, in which two half pieces are assembled through a detachable structure to embed some exemplary pressure sensors; 3B is a schematic diagram of the process of fixing the two half pieces to the rigid support structure of the massage table after the pieces are assembled; 3C is a schematic diagram of the state in which 3B is fixed after the fixing is completed; Figure 4 To illustrate the general flow chart of the artificial intelligence model building process of the portable Chinese medicine massage teaching and examination instrument based on artificial intelligence according to a specific embodiment 2 of the present invention, Figure 5 To illustrate the main interface of the intelligent operating system of the portable Chinese medicine massage teaching and examination instrument based on artificial intelligence according to a specific embodiment 3 of the present invention, Figure 6 This is a cross-sectional diagram showing the techniques of a portable Chinese medicine massage teaching and examination instrument based on artificial intelligence according to a specific embodiment 3 of the present invention. Figure 7 To illustrate a 2D force distribution diagram formed by the perception data of the portable Chinese medicine massage teaching and examination instrument based on artificial intelligence according to a specific embodiment 3 of the present invention, Figure 8 To illustrate the result interface of the portable Chinese medicine massage teaching and examination instrument based on artificial intelligence involved in a specific embodiment 3 of the present invention. DETAILED DESCRIPTION

[0033] The present invention will be described in detail below in combination with specific embodiments with reference to the accompanying drawings; however, the description is exemplary and the present invention is not limited to the specific embodiments.

[0034] Example 1 This embodiment 1 explains the overall structure and functions of the teaching and examination instrument. Figure 1 (a) and Figure 1 (b) The artificial intelligence-based Chinese medicine massage teaching and examination instrument includes a 3D human skin simulation model ( Figure 2 ), Massage Table ( Figure 3 ), a data processing module, a host, and a display device, wherein the quasi-3D human skin simulation model can simulate a person's front or back facing up, and can also be switched to the front when the front is needed, and switched to the back when the back is needed, forming a freely switchable structure, which includes a simulated skin layer, a simulated fat layer and a simulated muscle layer from top to bottom, and is fixed on a massage table, and pressure sensors are distributed above the simulated fat layer. In a preferred embodiment, the pressure sensors are distributed as a flexible multi-dimensional sensor matrix and are connected to the input port of the data processing module, and the output port is connected to the external device interface of the host.

[0035] Taking the back side as an example, when the quasi-3D human skin simulation model is shown with the back side facing upward, for the convenience of explanation, Figure 3 As shown, it is simplified to be composed of two half sheets a and a half sheet b (in fact, it can be a double-layer structure or a multi-layer structure, not limited to a two-half sheet structure), and each half sheet is made of silicone and is combined or split by a detachable structure set on the half sheet, and the sensor (taking the flexible multi-dimensional sensor matrix as an example) is buried between the two half sheets by the combination ( Figure 3 A) and are distributed within 2cm-3cm of the acupuncture points and their surroundings, and then can be detachably fixed on the rigid support structure on the massage table to form a simulated three-dimensional surface of the back of the 3D human body. Figure 3 After fixing in the direction of the arrow in B Figure 3 The signal lines connecting the sensors to the data processing module are led out from the preset gap in the middle of the assembly.

[0036] The display device is connected to the host, and is used to display the dynamics of the massage operation, force distribution data, preset parameters, learning and evaluation information based on the intelligent operating system interface. The host has a memory, and the memory is installed with an intelligent algorithm application. A camera device (not shown) is set on or around the massage table to capture the massage action. The captured massage action is virtually imaged on the interface through the digital twin algorithm of the intelligent algorithm application to achieve the dynamics. The intelligent algorithm application also performs 3D dimension reduction on the sensor data received by the host into a 2D 2D force distribution map, and builds an artificial intelligence model based on the 2D force distribution map, which is used to identify the difference between the action and the preset technique standard according to the real-time massage action, so as to obtain the evaluation results of the action. Among them, the force distribution data includes the force size, technique rhythm, etc., and the acupoint name converted by the host according to the number.

[0037] The teaching and testing instrument also includes a network communication device (not shown) for communicating with a cloud server. The virtual imaging and / or 2D force distribution diagram is also uploaded to the cloud server via the network communication device for downloading and viewing by the teacher or examiner for manual evaluation, thereby achieving human-machine combined evaluation of teaching outcomes.

[0038] Example 2 This embodiment explains the local detailed structure and algorithm functions of the teaching and examination instrument of Embodiment 1.

[0039] A sliding mechanism (not shown) is provided on one side of the massage table through an extended fixed bracket, and the sliding mechanism includes a motor, a slider, and a rotary encoder. There are two camera devices installed on the slider of the sliding mechanism. The host controls the sliding mechanism according to the sensed data so that the two camera devices move synchronously to the side of the massage part to shoot, so as to capture the massage action. For each camera device to move independently, a dual motor and a corresponding dual rotary encoder are provided. The sliding mechanism can be additionally extended on the massage table of the teaching and examination instrument, or it can be built-in and fixed on the massage table of the teaching and examination instrument, unfolded when in use, and folded and stored in the teaching and examination instrument when not in use.

[0040] The method in which the host controls the sliding mechanism according to the sensed data comprises the following steps: S1 numbers each sensor, sets the sensing data to data with number information, and outputs the digital quantity increment of the rotary encoder that should be generated when the slider on the sliding mechanism moves to the side of the massage position. With number Build the mapping: , is a linear mapping function, After receiving the data with number information, the S2 host calculates the , and control the slider to move until the digital output increment of the rotary encoder is .

[0041] Since the technique will contact multiple sensors, the data received here with numbered information is based on the earliest received perception data. For a group of perception data that cannot be distinguished in order, one is selected from the data channel of the corresponding data processing module using a random algorithm.

[0042] The mapping function can be written as ,in ,but and The digital outputs before and after receiving the data with numbered information are the digital outputs relative to the initial position. , the solution is , is the linear coefficient, The value is positive or negative, indicating the direction of movement. For the independent movement, the above method is used to control the movement of each slider, and the relative distance between the two camera devices is kept at a preset value that can be set.

[0043] like Figure 4 As shown, the intelligent algorithm application also reduces the 3D dimension of the sensor data received by the host into a 2D force distribution map, and constructs an artificial intelligence model based on the 2D force distribution map, which specifically includes the following steps: Q1 establishes an XOY two-dimensional rectangular coordinate system, where the X direction is from the head to the feet of the human body, and the Y direction is the left and right direction. A blank model of the force distribution diagram is constructed in the two-dimensional rectangular coordinate system. The blank model is composed of multiple units corresponding to the positions of acupoints, and an array is formed according to the X and Y directions. It is stipulated that the units corresponding to the acupoints are projected on the two-dimensional rectangular coordinate system to obtain the array position.

[0044] The figure shows the corresponding units of the acupoints, represented by squares, including Yunmen, Youmen, Dabao, Shuifen, Qi Chong, Huantiao, Yanglingquan, and an additional site in the vicinity of Shuifen. Huantiao and Yanglingquan are reduced in dimension by projection in the above-mentioned rectangular coordinate system. The directions of other acupoints in the third dimension (i.e., the direction perpendicular to the XOY plane) are not much different, but in fact they are also converted into 2D through the above-mentioned projection method.

[0045] Q2 performs pseudo-color calibration on the sensed data and assigns the corresponding pseudo-color value to the corresponding unit to obtain a real-time 2D force distribution map; Q3: Let masters of different schools and students of advanced, intermediate and elementary levels perform massage operations on each human body part on two groups of 3D human skin simulation models, and collect the sensed data in the process of timed triggering. According to step Q2, multiple 2D force distribution maps of each school and each level of students are formed, and the multiple 2D force distribution maps are divided into a training set and a verification set. According to the three timed triggering time periods of the front section, middle section and back section of the completion process of a manual massage action on each human body part, the corresponding training set and verification set are respectively classified into the front section class, the middle section class and the back section class; The trigger acquisition timing can be 0.2s, 0.4s, and 0.8s after the start of the technique (timing starts when recognizable perception data is received) as the front section, middle section, and back section respectively, and the trigger acquisition timing is adjusted according to the characteristics of different techniques.

[0046] Q4 is a convolutional neural network (RES-NET) group with residual mechanism for each of the three time periods, which is divided into multiple groups. The output of each network group is fully connected and input into softmax for grading. There are 12 categories in total, namely, the master and four levels of high, medium and elementary. The validation set is used for verification until the recognition accuracy is stable, and the training of all groups is stopped to obtain a technique artificial intelligence model.

[0047] Q5 then builds other artificial intelligence models based on Q3-Q4.

[0048] In addition, masters from different schools and students at senior, intermediate and primary levels were asked to perform massage operations on each part of the human body for a specified duration of 1-10 minutes on the two groups of 3D human skin simulation models, and the perception data of the initial segment, i.e. the first third of the time, the intermediate segment, i.e. the middle third of the time, and the final segment, i.e. the remaining third of the time, were collected again at a re-timing trigger. The artificial intelligence model of whether the massage weight met the corresponding level qualification standards in the segments of different lengths was constructed based on Q3-Q5. That is, if a senior student is taken as an example, the prediction result is that it does not meet the weight standards of senior students, but meets the master level, which means that the student has made progress compared to the existing rating; on the contrary, if it meets the intermediate weight standards, it means that the technique has regressed.

[0049] Example 3 This embodiment explains the assessment and the intelligent operating system interface display.

[0050] We define the force distribution data to include a real-time 2D force distribution diagram and the current pressure converted from the perception data of each sensor. The preset parameters include any one of the following parameters or a combination thereof: Number of manipulations: record the cumulative number of times generated during the entire practice process, and each force is recorded as one force area: record the area generated by the real-time force of the manipulation, based on the pressure applied each time; Manipulation time: record the cumulative practice time of the manipulation practice from 0 to the end; Manipulation rate: rate = distance of manipulation movement / time taken.

[0051] The calculation method of the assessment results is: T1, within the prescribed 10 minutes, record the number of manipulations performed on the 3D manikin, the area of ​​force generated, the manipulation time, and the manipulation speed, and compare them with the corresponding data of the masters of each school, and use the percentage value as the score of this item. T2, within the prescribed time, collect the perception data of the front, middle, back, initial, intermediate and final stages of each manipulative technique, form a 2D force distribution map, and substitute it into the trained manipulative technique artificial intelligence model and the weight artificial intelligence model, to determine the corresponding level of the manipulative technique and the law of weight change to determine whether it is qualified, and score based on the level and whether it is qualified; T3 outputs the average of each score in T1 and the corresponding level and severity of pass or fail judgment in T2 as the final score.

[0052] Figure 5 Main interface: 1. After entering the function, select the practice method and time (later add the option of selecting a teacher); 2. After entering the method interface, click the start button to start practicing. If the start button is not clicked, the standard video starts the demonstration; 3. After selecting the parameters, enter the interface. After clicking the start button, the countdown will be 3 seconds, and the practice will start after it returns to zero. The smart button is the entrance for modeling and intelligent evaluation of scores.

[0053] Figure 6 This is a demonstration of back techniques: rubbing method: sports type technique: parameter display: the horizontal axis is pressure; the vertical axis is the number of times; one-finger Zen: sports type technique: parameter display: the horizontal axis is pressure; the vertical axis is the number of times; massage method: sports type technique: parameter display: the horizontal axis is pressure; the vertical axis is the number of times; rolling method: pressure type technique: parameter display: the horizontal axis is pressure; the vertical axis is time; point method: pressure type technique: parameter display: the horizontal axis is pressure; the vertical axis is time.

[0054] Figure 7 A 2D force map formed for the sensor data, and Figure 8 The performance scores given by the intelligent model of Example 2 and the performance calculation method of this embodiment are given, wherein the trajectory corresponds to the result of the performance calculated by the intelligent model. The higher the rank and the better the weight is, the higher the score.

Claims

1. A portable Chinese medicine massage teaching and examination instrument based on artificial intelligence, characterized in that: The invention comprises a quasi-3D human skin simulation model, a massage table, a data processing module, a host, and a display device, wherein the quasi-3D human skin simulation model is divided into a simulated skin layer, a simulated fat layer, and a simulated muscle layer from top to bottom, the quasi-3D human skin simulation model is fixed on the massage table, and sensors are distributed on the simulated fat layer, which can sense external forces and are connected to the input port of the data processing module, while the output port is connected to the external device interface of the host, the display device is connected to the host, and is used to display the dynamics of massage operations, force distribution data, and massage rhythm preset parameters, learning and evaluation information based on the intelligent operating system interface, The host has a memory in which an intelligent algorithm application is installed. A camera is provided on or around the massage table to capture the massage action. The captured massage action is virtually imaged on the interface through the digital twin algorithm of the intelligent algorithm application to achieve the dynamic state. The intelligent algorithm application also performs 3D dimension reduction on the perception data on the sensor received by the host into a 2D force distribution map, and constructs an artificial intelligence model based on the 2D force distribution map to identify the difference between the real-time massage action and the preset technique standard according to the real-time massage action, thereby obtaining the evaluation score of the action.

2. The teaching and examination instrument according to claim 1, characterized in that: One side of the massage table has a sliding mechanism, and the two cameras are installed on the sliding mechanism. The host controls the sliding mechanism according to the sensing data, so that the two cameras move synchronously or independently to one side of the massaged part to shoot, so as to capture the massage action.

3. The method according to claim 2, characterized in that The sliding mechanism comprises a motor, two sliders, and a rotary encoder, and the two camera devices are respectively mounted on different sliders. The method for the host to control the sliding mechanism according to the sensed data comprises the following steps: S1 numbers each sensor, sets the sensing data to data with number information, and outputs the digital quantity increment of the rotary encoder that should be generated when the slider on the sliding mechanism moves to the side of the massage position. With number Build the mapping: , is the mapping function; After receiving the data with number information, the S2 host calculates the , and control the slider to move until the digital output increment of the rotary encoder is .

4. The method according to claim 3, characterized in that: is a linear mapping function ,in ,but and The digital outputs before and after receiving the data with numbered information are the digital outputs relative to the initial position. , the solution is , is the linear coefficient, The value is signed to indicate the direction of movement.

5. The teaching and examination instrument according to claim 3 or 4, characterized in that: The force distribution data includes the current pressure converted from the sensing data of each sensor, and the acupoint name converted by the host according to the number; for the independent movement, there are two motors and two rotary encoders, and the relative distance between the two camera devices is maintained at a settable preset value.

6. The teaching and examination instrument according to claim 5, characterized in that: The intelligent algorithm application also reduces the 3D dimension of the sensor data received by the host into a 2D force distribution map, and constructs an artificial intelligence model based on the 2D force distribution map, which specifically includes the following steps: Q1 establishes an XOY two-dimensional rectangular coordinate system, where the X direction is the direction from the head to the feet of the human body, and the Y direction is the left-right direction. A blank model of the force distribution diagram is constructed in the two-dimensional rectangular coordinate system. The blank model is composed of a plurality of units corresponding to the positions of the acupoints, and an array is formed according to the X and Y directions. It is stipulated that the units corresponding to the acupoints are all projected on the two-dimensional rectangular coordinate system to obtain the array position; Q2 performs pseudo-color calibration on the sensed data and assigns the corresponding pseudo-color value to the corresponding unit to obtain a real-time 2D force distribution map; Q3: Let masters of different schools and students of advanced, intermediate and elementary levels perform massage operations on each human body part on two groups of 3D human skin simulation models, and collect the sensed data in the process of timed triggering. According to step Q2, multiple 2D force distribution maps of each school and each level of students are formed, and the multiple 2D force distribution maps are divided into a training set and a verification set. According to the three timed triggering time periods of the front section, middle section and back section of the completion process of a manual massage action on each human body part, the corresponding training set and verification set are respectively classified into the front section class, the middle section class and the back section class; Q4 builds a convolutional neural network group, which is divided into multiple martial arts groups. The output of each group of networks is fully connected and input into softmax for 12 classifications of time periods and grades, that is, it is divided into three time periods, each time period has a master and four levels of high, medium and elementary. The validation set is used for verification until the recognition accuracy is stable, and all groups are stopped from training to obtain a method artificial intelligence model; Q5 then builds other artificial intelligence models based on Q3-Q4.

7. The teaching and examination instrument according to claim 6, characterized in that: Masters from different schools and students at advanced, intermediate and elementary levels are asked to perform massage operations on each part of the human body for a specified duration on the two groups of 3D human skin simulation models. The perception data of the initial, intermediate and final segments of the collection time are re-timed to trigger the collection, and an artificial intelligence model of the weight of the massage in segments of different lengths is constructed according to Q3-Q5 to determine whether the weight of the massage meets the corresponding level qualification standards.

8. The teaching and examination instrument according to claim 7, characterized in that: The convolutional neural network is a convolutional neural network with a residual mechanism, and one or a corresponding plurality of convolutional neural networks with a residual mechanism are used for training each time period and / or each duration segment.

9. The teaching and examination instrument according to claim 7 or 8, characterized in that: The force distribution data also includes a real-time 2D force distribution diagram, and the preset parameters include any one of the following parameters or a combination thereof: Number of manipulations: record the cumulative number of times generated during the entire practice process, and each force is recorded as one force area: record the area generated by the real-time force of the manipulation, based on the pressure received each time; Manipulation time: record the cumulative practice time of the manipulation practice from 0 to the end; Manipulation rate: rate = distance of manipulation movement / time used; practice manipulation, manipulation practice time, reference to the master, start / end practice, practice countdown view; The content displayed on the intelligent operating system interface also includes the essentials of the technique: describing the operational requirements of the selected technique, and providing students with detailed text descriptions of the practical methods when practicing; technique stability: comparing the techniques of recorded expert teachers, and obtaining stability parameters after comparisons of frequency, pressure, speed and other aspects, calculated by combining three data; force statistics: recording the maximum force value, minimum force value, average pressure value, pressure stability, and the distribution of all pressures during the practice, recorded according to the pressure characteristics of each technique; and technique video / force distribution: capturing real-time techniques through a camera, converting them into digitized hand shapes, and displaying the recorded technique operation style in real time.

10. The teaching and examination instrument according to claim 9, characterized in that: The calculation method of the assessment results is: T1, within the specified time, record the number of manipulations performed on the 3D human skin simulation model, the area of ​​force generated, the manipulation time, and the manipulation speed, and compare them with the corresponding data of the masters of each school, and use the percentage value as the score of this item. T2, within the prescribed time, collect the perception data of the front, middle, back, initial, intermediate and final sections of each technique, form a 2D force distribution map, and substitute it into the trained technique artificial intelligence model and weight artificial intelligence model respectively, to determine the corresponding level of the technique and the law of weight change to determine whether it is qualified, and score based on the level and whether it is qualified or not. T3 outputs the average of each score in T1 and the corresponding level and severity of the pass or fail judgment in T2 as the final score.

11. The teaching and examination instrument according to claim 3, 4, 6-8 or 10, characterized in that: The teaching and testing instrument also includes a network communication device for communicating with a cloud server. The virtual imaging and / or 2D force distribution diagram is also uploaded to the cloud server through the network communication device for downloading and viewing by the teacher or examiner for manual evaluation.

Images