A method for identifying operation mode of facial massager

Abstract

The present invention proposes an operation mode recognition method for a facial massager, which relates to the field of pattern recognition technology. A camera system generates a user's facial skin picture, and an image analysis module is used to recognize the characteristic value of the skin state; various sensor data sets of the facial massager under different operation modes are collected, and various sensor data sets are clustered respectively to form a pattern mapping list; the characteristic value of the skin state is input into a pattern recognition neural network model, and the optimal value of multiple types of sensor data is output; the operation mode corresponding to the optimal value of multiple types of sensor data is queried according to the pattern mapping list, and the optimal operation mode is recognized. The massage mode analysis process of the present invention is faster, and the final massage mode recognized corresponds accurately to the user's skin state. The massage operation mode of the facial massager can be automatically recognized according to the user's skin state to achieve the best massage effect.

Description

Technical Field

[0001] The invention provides an operation mode recognition method of a facial massager, and relates to the technical field of pattern recognition. Background Art

[0002] A facial massager is a beauty instrument designed to improve blood circulation in the facial skin, relieve muscle tension, promote the absorption of skin care products, and improve facial skin through various massage techniques. The operating modes of the facial massager include: vibration massage technology, which uses high-frequency vibration to stimulate facial muscles and skin, promote blood circulation and lymphatic detoxification; intelligent sensor technology, which combines pressure, temperature and other sensors to intelligently adjust the massage intensity and mode to meet the needs of different skin conditions; facial roller massage: imitates traditional facial massage techniques, massages facial acupuncture points through roller design, promotes blood circulation and other modes.

[0003] The intelligent sensor technology of the facial massager can adjust the massage mode according to the user's skin condition. The skin detection sensors (such as capacitive sensors, pressure sensors, temperature sensors, etc.) integrated in the massager collect skin data in real time, including skin moisture content, oil level, temperature, elasticity, pore size, etc. The collected skin data is analyzed to evaluate the skin condition, analyze the user's skin condition, and recommend a suitable massage mode. However, the analysis process of the facial massager in the prior art is too complicated and slow, resulting in a poor user experience. At the same time, the massage mode does not correspond accurately to the user's skin condition, and there is a phenomenon of over-massage or insufficient massage. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes an operation mode recognition method of a facial massager, comprising the following steps:

[0005] S1, the camera system generates a user's facial skin picture, and the image analysis module identifies the characteristic value of the skin condition;

[0006] S2, collecting various sensor data sets of the facial massager in different operation modes, and clustering various sensor data sets respectively to form a mode mapping list;

[0007] S3, inputting the characteristic values ​​of the skin state into a pattern recognition neural network model, and outputting optimal values ​​of multiple types of sensor data;

[0008] S4. Query the operating modes corresponding to the optimal values ​​of the multiple types of sensor data according to the mode mapping list, and identify the optimal operating mode.

[0009] Furthermore, the step S2 comprises:

[0010] S21, setting the maximum area radius and point number threshold for each type of sensor data set;

[0011] S22, traversing all data in the sensor data set, and marking all data as cluster core points or contour points according to the maximum area radius and the point number threshold;

[0012] S23. Segment the sensor data set according to the labeling result to generate a plurality of final segmentation regions and data point sets within the regions.

[0013] Further, the sensor data set is preliminarily segmented according to the marking result of step S22 to obtain multiple preliminary segmentation areas, and the average of the maximum lengths from the clustering core points of all preliminary segmentation areas to each included point in the area is used as the truncation length; based on the truncation length, the boundary area point set of every two preliminary segmentation areas is determined, and the average of the local density of the sub-point sets in all the boundary area point sets is used as the density critical value to generate multiple final segmentation areas.

[0014] Furthermore, in step S3, the characteristic values ​​of the skin state {tx1…tx i …tx n} is input into the pattern recognition neural network model, n represents the number of characteristic values ​​of the skin state, and the output value of each neuron in the hidden layer of the pattern recognition neural network model is calculated:

[0015] ;

[0016] ;

[0017] Among them, s j represents the output intermediate value of the jth neuron in the hidden layer, represents the threshold of the jth neuron in the hidden layer; b j represents the output value of the jth neuron in the hidden layer, j=1,2,…p, p is the number of neurons in the hidden layer; {w j1 …w ji …w jn} are the n connection weights of the jth neuron in the hidden layer.

[0018] Furthermore, in step S3, the hidden layer output sequence {b1, ...b j ,…b p} is input to the output layer of the pattern recognition neural network model, then the output of each neuron in the output layer is:

[0019] ;

[0020] ;

[0021] Among them, sk represents the output intermediate value of the kth neuron in the output layer, represents the threshold of the kth neuron in the output layer, y k represents the output value of the kth neuron in the output layer. The p connection weights of the kth neuron in the output layer are {v k1 …v kj …v kp}, k = 1, 2, ... q, q represents the number of neurons in the output layer, and the output sequence {y1 ... y k …y q The q values ​​in} represent the optimal values ​​of q types of sensor data respectively.

[0022] Furthermore, in step S4, the priority value F(c) of each type of sensor data is determined according to the overlap between the data range of each type of sensor data and the data range of each type of sensor data in the pattern mapping list. m ):

[0023] ;

[0024] Among them, c m is the data range of the mth sensor data in the pattern mapping list, N is the number of sensor data categories remaining excluding the mth sensor data and the sensor data categories with determined priorities, and the function The data range c represents the hth and mth sensor data. h With c m The number of overlapping data, D(c m ) represents the difference between the maximum and minimum values ​​in the data range of the mth type of sensor data.

[0025] Furthermore, in step S4 and step S22, if a data has at least Z data in the area with a maximum area radius of R, the data is marked as a cluster core point; if a data does not have any other data in the area with a maximum area radius of R, the data is marked as a noise point and deleted; if a data is located in the area with a maximum area radius of R of the cluster core point, the data is marked as a contained point.

[0026] Furthermore, the various types of sensor data include sensor data of a force sensor of the massage head and sensor data of a posture sensor; the sensor data of the force sensor of the massage head also includes: massage force, speed, and frequency sensor data.

[0027] Furthermore, the characteristic values ​​of the skin condition include: erythema index, texture parameter, stratum corneum water content and skin oil content.

[0028] Compared with the prior art, the present invention has the following beneficial technical effects:

[0029] The camera system generates a user's facial skin picture, and the image analysis module identifies the characteristic value of the skin state; collects various sensor data sets of the facial massager in different operating modes, and clusters each sensor data set to form a pattern mapping list; inputs the characteristic value of the skin state into the pattern recognition neural network model, and outputs the optimal value of multiple types of sensor data; queries the operating mode corresponding to the optimal value of multiple types of sensor data according to the pattern mapping list, and identifies the optimal operating mode. Compared with the massage method of the facial massager in the prior art, the massage mode analysis process of the present invention is faster, and the final massage mode identified accurately corresponds to the user's skin state. The massage operation mode of the facial massager can be automatically identified according to the user's skin state to achieve the best massage effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0031] Figure 1 A schematic flow chart of an operating mode recognition method of a facial massager of the present invention;

[0032] Figure 2 This is a schematic diagram of the mode mapping list;

[0033] Figure 3 It is a schematic diagram of the process of clustering various sensor data sets respectively according to the present invention;

[0034] Figure 4 It is a structural schematic diagram of the operation mode recognition system of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0036] In the drawings of the specific embodiments of the present invention, in order to better and more clearly describe the working principles of the various components in the system, the connection relationship of the various parts in the device is shown, which only clearly distinguishes the relative position relationship between the various components, and cannot constitute a limitation on the signal transmission direction, connection sequence and structural size, size and shape of each part within the component or structure.

[0037] Example 1

[0038] like Figure 1 The flowchart of the operation mode recognition method of the facial massager of the present invention is shown, which includes the following steps:

[0039] S1. The user's facial skin picture generated by the camera system is used to identify the characteristic value of the skin condition through the image analysis module.

[0040] Collect a large number of user facial skin pictures, collect a large number of user facial skin pictures, these pictures should cover different skin types and conditions. Standardize the collected user facial skin pictures, including cropping, scaling, rotation correction, etc., to ensure that all pictures have a uniform format and size.

[0041] Extract characteristic values ​​of skin condition from standardized skin images, including erythema index, texture parameters, etc.

[0042] In a preferred embodiment, a skin detector can also be used to detect the skin condition. By detecting the oil secretion status on the skin surface and accurately analyzing the distribution of stratum corneum water at different depths through a microscope, characteristic values ​​such as stratum corneum water content and skin oil content can be obtained.

[0043] In a preferred embodiment, in addition to the erythema index, texture parameters, stratum corneum water content and skin oil content, other skin information (such as skin quality, age, gender, etc.) should also be combined as other characteristic values ​​of the skin state.

[0044] S2. Collect various sensor data sets of the facial massager under different operating modes, and cluster the various sensor data sets respectively to form a mode mapping list.

[0045] The various types of sensor data include sensor data of the force sensor of the massage head, sensor data of the posture sensor, etc. The sensor data of the force sensor of the massage head also includes sensor data such as massage strength, speed, and frequency.

[0046] Various types of sensor data under different operating modes are collected to form corresponding categories of sensor data sets, such as massage frequency data sets, massage intensity data sets, etc., and cluster analysis is performed on each type of sensor data set to form a mode mapping list.

[0047] like Figure 2 As shown, clustering various sensor data sets includes the following steps:

[0048] S21. For each type of sensor data set, set the maximum area radius R and the point number threshold Z.

[0049] S22, traverse all data in the sensor data set, and mark all data as cluster core points or contour points.

[0050] If a data has at least Z data in the area with the maximum area radius of R, the data is marked as a cluster core point; if a data does not have any other data in the area with the maximum area radius of R, the data is marked as a noise point and deleted; if a data is located in the area with the maximum area radius of R of the cluster core point, the data can be included by the cluster core point and is marked as an included point.

[0051] S23. Segment the sensor data set according to the labeling result to generate a plurality of final segmentation regions and data point sets within the regions.

[0052] The sensor data set is preliminarily segmented according to the marking result of step S22 to obtain multiple preliminary segmentation areas, and the average of the maximum lengths from the cluster core points of all preliminary segmentation areas to each inclusion point in the area is used as the truncation length; the boundary area point set of each two preliminary segmentation areas is determined based on the truncation length, if the length between the inclusion point of one of the two preliminary segmentation areas and the inclusion point of the other preliminary segmentation area is less than the truncation length, the two inclusion points belong to the boundary area point set of the two preliminary segmentation areas; the average of the local density of the sub-point sets in all the boundary area point sets is used as the density critical value to generate multiple final segmentation areas.

[0053] The specific steps include:

[0054] Step 231: perform preliminary segmentation of the sensor data set according to the labeling result to obtain multiple preliminary segmentation regions, and take the average of the maximum lengths of the cluster core points of all preliminary segmentation regions to the included points in the included points in the region as the truncation length. :

[0055] ;

[0056] Among the M preliminary segmented areas, It is the maximum length from the cluster core point of the preliminary segmentation area B to each included point J in the preliminary segmentation area.

[0057] Step 232: determine the boundary area point set of each two preliminary segmentation areas based on the cutoff length. If the length between the included point of one of the two preliminary segmentation areas and the included point of the other preliminary segmentation area is less than the cutoff length, the two included points belong to the boundary area point set of the two preliminary segmentation areas.

[0058] Assume the boundary region point set L mn for:

[0059] ;

[0060] Among them, I and J belong to the initial segmentation area B m ,B n When the length between the two included points is less than the cutoff length When , it is considered that these two included points belong to the preliminary segmentation area B m ,B n The boundary area points are concentrated.

[0061] Step 233: taking the mean of the local density of the sub-point sets in all the boundary area point sets as the density critical value.

[0062] The local density including point I for:

[0063] ;

[0064] Among them, d IJ is the length of the included point I and other included points J in the sub-point set, K is the number of other included points in the sub-point set, is the cut-off length.

[0065] The mean of the local density of all sub-point sets of all boundary area point sets is calculated as the density critical value.

[0066] Step 234: take the data points of the sub-point set whose local density in the boundary area point set is less than the density critical value as discrete points. For each discrete point, find the cluster core point with the closest length to the discrete point from the M cluster core points, and add the discrete point to the segmentation area to which the cluster core point belongs, thereby generating M final segmentation areas.

[0067] S24. Cluster the sensor data sets of other categories in turn according to the above steps to obtain multiple final segmentation regions and data point sets within each region under each category of sensor data set, and mark different operation modes for the data point sets within each final segmentation region.

[0068] Finally, the data points of each type of sensor data set belonging to the same mode are imported into the same column in the mode mapping list, such as Figure 3 As shown, for example, in the sensor data set formed by sensor data category A, the data subset with a value range of [A1, A2] belongs to operation mode one; the data subset with a value range of [A3, A4] belongs to operation mode two; the data subset with a value range of [A5, A6] belongs to operation mode three; and so on.

[0069] S3. Input the characteristic value of the skin state obtained in step S1 into the pattern recognition neural network model, and output the optimal value of the multi-class sensor data.

[0070] The characteristic values ​​of the skin state {tx1…tx i …tx n} is input into the pattern recognition neural network model, n represents the number of eigenvalues ​​of the skin state, and the output of each neuron in the hidden layer of the pattern recognition neural network model is calculated:

[0071] ;

[0072] ;

[0073] Among them, s j represents the output intermediate value of the jth neuron in the hidden layer, represents the threshold of the jth neuron in the hidden layer; b j represents the output value of the jth neuron in the hidden layer, j=1,2,…p, p is the number of neurons in the hidden layer; {w j1 …w ji …w jn} are the n connection weights of the jth neuron in the hidden layer.

[0074] The hidden layer output sequence {b1,…b j ,…b p} is input to the output layer of the pattern recognition neural network model, then the output of each neuron in the output layer is:

[0075] ;

[0076] ;

[0077] Among them, s k represents the output intermediate value of the kth neuron in the output layer, represents the threshold of the kth neuron in the output layer, y k represents the output value of the kth neuron in the output layer. The p connection weights of the kth neuron in the output layer are {v k1 …v kj …v kp}, k = 1, 2, ... q, q represents the number of neurons in the output layer, and the output sequence {y1 ... y k…y q The q values ​​in} represent the optimal values ​​of q types of sensor data respectively.

[0078] S4. According to the mode mapping list obtained in step S2, query the operation mode corresponding to the optimal values ​​of the multiple types of sensor data output in step S3 to identify the optimal operation mode.

[0079] According to the mode mapping list, the range where the optimal value of the q-type sensor data is located is queried, and the corresponding operation mode is found according to the range where the optimal value of the q-type sensor data is located.

[0080] Preferably, if the operating modes corresponding to the optimal values ​​of all sensor data are the same, the operating mode is directly selected; if the operating modes corresponding to the optimal values ​​of the sensor data are different, the final operating mode needs to be determined according to the priority rule.

[0081] In a preferred embodiment, the following formula is used to determine the priority value of each type of sensor data according to the data range of each type of sensor data in the mode mapping list and the overlap of the data range of each type of sensor data:

[0082] ;

[0083] Among them, c m is the data range of the mth sensor data in the pattern mapping list, N is the number of sensor data categories remaining excluding the mth sensor data and the sensor data categories with determined priorities, and the function The data range c represents the hth and mth sensor data. h With c m The number of overlapping data, D(c m ) represents the difference between the maximum and minimum values ​​in the data range of the mth type of sensor data, for example, A2-A1, F(c m ) is the calculated c m Priority value.

[0084] Example 2

[0085] The present invention also proposes an operation mode recognition system for implementing the above-mentioned operation mode recognition method of the facial massager, such as Figure 4 As shown, the operation mode recognition system includes: a camera system, an image analysis module, a sensor data processing unit, a pattern recognition neural network model and a recognition unit.

[0086] The user's facial skin picture generated by the camera system is used to identify the characteristic values ​​of the skin condition through the image analysis module.

[0087] The image analysis module performs specific processing and analysis on skin images to indirectly evaluate changes in stratum corneum water content, involving technologies such as color model conversion, region segmentation, and image binarization.

[0088] The sensor data processing unit collects various sensor data sets of the facial massager in different operating modes, and clusters various sensor data sets to form a mode mapping list. The collection of various sensor data sets is mainly for historical data statistics. The more statistical data, the better the later clustering analysis effect.

[0089] The input parameter of the pattern recognition neural network model is the characteristic value of the skin state, and the output parameter is the optimal value of the corresponding multi-class sensor data. According to the complexity of the problem and the characteristics of the data, a suitable pattern recognition neural network model is selected, such as a multi-layer perceptron neural network model, a convolutional neural network or a recurrent neural network. In this embodiment, since the input is a characteristic value rather than an image or sequence data, a multi-layer perceptron neural network model is preferred.

[0090] The recognition unit queries the operating mode corresponding to the optimal value of the multi-class sensor data output by the pattern recognition neural network model according to the pattern mapping list, and recognizes the optimal operating mode.

[0091] Once the optimal operating mode is identified, the facial massager system can adjust the parameters of the relevant sensors and the massage method to provide the most personalized skin massage experience. For example, high-frequency vibration is used to stimulate facial muscles and skin to promote blood circulation and lymphatic detoxification; combined with pressure, temperature and other sensors, the massage intensity is adjusted to meet the needs of different skin conditions; facial acupuncture points are massaged to promote blood circulation and other modes.

[0092] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.

[0093] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for identifying an operating mode of a facial massager, characterized in that: The steps include: S1, the camera system generates a user's facial skin picture, and the image analysis module identifies the characteristic value of the skin condition; S2. Collect various sensor data sets of the facial massager in different operation modes, and cluster the various sensor data sets respectively to form a mode mapping list, including: S21, setting the maximum area radius and point number threshold for each type of sensor data set; S22, traverse all data in the sensor data set, and mark all data as cluster core points or contour points according to the maximum area radius and the point number threshold; if a data has at least Z data in the area with the maximum area radius of R, then the data is marked as a cluster core point; if a data does not have any other data in the area with the maximum area radius of R, then the data is marked as a noise point and deleted; if a data is located in the area with the maximum area radius of R of the cluster core point, then the data is marked as a contained point; S23, segmenting the sensor data set according to the labeling result to generate a plurality of final segmentation regions and data point sets within the regions; Perform preliminary segmentation on the sensor data set according to the marking result of step S22 to obtain multiple preliminary segmentation areas, and use the average of the maximum lengths from the cluster core points of all preliminary segmentation areas to each of the included points in the area as the truncation length; determine the boundary area point set of every two preliminary segmentation areas based on the truncation length, use the average of the local density of the sub-point sets in all the boundary area point sets as the density critical value, and generate multiple final segmentation areas; S24, clustering the sensor data sets of other categories in turn according to the above steps to obtain multiple final segmentation regions and data point sets within each region under each category of sensor data set, and marking different operation modes for the data point sets within each final segmentation region; Finally, the data point sets belonging to the same mode of each type of sensor data set are imported into the same column in the mode mapping list; S3, inputting the characteristic values ​​of the skin state into a pattern recognition neural network model, and outputting optimal values ​​of multiple types of sensor data; S4. Query the operating modes corresponding to the optimal values ​​of the multiple types of sensor data according to the mode mapping list, and identify the optimal operating mode; if the operating modes corresponding to the optimal values ​​of all sensor data are the same, directly select the operating mode; if the operating modes corresponding to the optimal values ​​of the sensor data are different, it is necessary to determine the final operating mode based on the priority values ​​of each type of sensor data.

2. The operating mode recognition method according to claim 1, characterized in that: In step S3, the characteristic values ​​of the skin state {tx1…tx i …tx n } is input into the pattern recognition neural network model, n represents the number of characteristic values ​​of the skin state, and the output value of each neuron in the hidden layer of the pattern recognition neural network model is calculated: ； ； Among them, s j represents the output intermediate value of the jth neuron in the hidden layer, represents the threshold of the jth neuron in the hidden layer; b j represents the output value of the jth neuron in the hidden layer, j=1,2,…p, p is the number of neurons in the hidden layer; {w j1 …w ji …w jn } are the n connection weights of the jth neuron in the hidden layer.

3. The operating mode recognition method according to claim 2, characterized in that: In step S3, the hidden layer output sequence {b1, ...b j ,…b p } is input to the output layer of the pattern recognition neural network model, then the output of each neuron in the output layer is: ； ； Among them, s k represents the output intermediate value of the kth neuron in the output layer, represents the threshold of the kth neuron in the output layer, y k represents the output value of the kth neuron in the output layer. The p connection weights of the kth neuron in the output layer are {v k1 …v kj …v kp }, k = 1, 2, ... q, q represents the number of neurons in the output layer, and the output sequence {y1 ... y k …y q The q values ​​in} represent the optimal values ​​of q types of sensor data respectively.

4. The operating mode recognition method according to claim 1, characterized in that: The various types of sensor data include sensor data of a force sensor of the massage head and sensor data of a posture sensor; The sensing data of the force sensor of the massage head also includes: massage force, speed, and frequency sensing data.

5. The operating mode recognition method according to claim 1, characterized in that: The characteristic values ​​of the skin condition include: erythema index and texture parameter.

Images