Radio frequency therapeutic apparatus and motion trail identification system thereof

Abstract

The invention discloses a radio frequency therapeutic apparatus and a motion trail recognition system thereof, the radio frequency therapeutic apparatus is provided with a control module, and the control module is electrically connected with an acceleration sensor and a data processing module; the system runs on the radio frequency therapeutic instrument and comprises a data preprocessing module and a deep learning model. The data preprocessing module filters the data through a Kalman filter and sends the processed data to the deep learning model, and the deep learning model analyzes and learns the data sent by the data preprocessing module and judges and recognizes a motion track formed by operation behaviors of a user through the model obtained through training. The device can adjust the working power according to the motion trail, and the purpose of controlling the output power of the radio frequency therapeutic apparatus in real time is achieved.

Description

Technical Field

[0001] The present invention relates to the field of equipment motion trajectory data processing, and in particular to a radio frequency therapeutic device motion trajectory recognition system. Background Art

[0002] The radio frequency therapy device is used to treat or beautify the user's face. When using it, the working end needs to be facing the user's skin and moved along the user's skin.

[0003] Existing radiofrequency therapeutic devices cannot obtain the user's operation in time when in use, especially during the user's facial beauty treatment operation. In addition, it is difficult for the user to observe effectively during the facial beauty treatment process. In most cases, the user operates by feeling. This limitation ultimately leads to unsatisfactory accuracy of user operation and unsatisfactory beauty treatment effect.

[0004] In addition to operating based on feeling, some radio frequency therapy devices or users also use simple timers to control the timing and specific methods of operation. However, this method lacks corresponding effective feedback mechanism and intelligent judgment mechanism. When the user uses the device to move or stop within a small range, if the radio frequency therapy device is not operated in time, the user may be burned by the device.

[0005] The existing feedback mechanism of radio frequency therapy devices is to obtain skin temperature in real time through temperature sensors, and then control the power or opening and closing of the working end to prevent users from being burned. However, since the acquisition of skin temperature will be affected by the temperature of the peripheral components of the device, the skin temperature finally obtained by the temperature sensor will have a significant difference from the actual skin temperature.

[0006] Therefore, there is an urgent need for a radiofrequency therapeutic device and a motion trajectory recognition system thereof that can solve one or more of the above problems. Summary of the invention

[0007] In order to solve one or more problems existing in the prior art, the present invention provides a radio frequency therapeutic device and a motion trajectory recognition system thereof. The technical solution adopted by the present invention to solve the above problems is: a radio frequency therapeutic device, the radio frequency therapeutic device is provided with a control module, the control module is electrically connected to an acceleration sensor and a data processing module, the control module is used to control the operation of the device, the acceleration sensor obtains the acceleration data of the device in the X, Y, and Z directions, and the data processing module performs processing on the acceleration data.

[0008] and a corresponding motion trajectory recognition system, the system comprising: a data preprocessing module and a deep learning model;

[0009] The data preprocessing module filters the data through a Kalman filter and sends the processed data to the deep learning model, and the deep learning model analyzes and learns the data sent by the data preprocessing module, and judges and recognizes the motion trajectory formed by the user's operation behavior through the trained model;

[0010] The deep learning model processes the input data through a convolutional neural network algorithm and maps the processing results to data labels, which include: normal, large circle, small circle, jitter and stop. The large circle indicates that the user makes circular motion larger than a preset diameter range during use; the small circle indicates that the user makes circular motion smaller than a preset diameter range during use; the jitter indicates that the skin area contacted by the radio frequency therapy device has hardly changed during use while the device itself has moved; the stop indicates that the user stops moving in a certain area during use and the stop time is greater than a preset time value; and the normal indicates the user's normal sliding operation except for the large circle during use.

[0011] Furthermore, the convolutional neural network algorithm structure of the deep learning model includes: an input layer, a convolution layer, an activation layer, an average pooling layer and at least two connection layers, and the connection layer maps the results to the data labels.

[0012] Furthermore, the data preprocessing module uses a segmented sliding window to intercept the data of the radiofrequency therapy device, and the segmented sliding window is: intercepting a period of n seconds of data every x seconds, and the deep learning model marks the intercepted data with the data label.

[0013] Further, when the data tag is the normal or large circle, the system controls the radio frequency therapeutic apparatus to continue to operate at the set power;

[0014] When the data label is any one or more of the small circle, the shaking, and the stopping, the system controls the radio frequency therapeutic device to reduce the operating power or stop the operation and outputs a warning signal to the user.

[0015] Furthermore, the data preprocessing module filters the acquired data through a Kalman filter. During the filtering process, a state vector x is defined for the acquired position, velocity and acceleration data. The state vector Among them, p x 、p y 、p z is the coordinate data of the position, v x 、v y 、v z is the velocity data of the position in the X, Y, and Z directions, a x 、ay 、a z is the acceleration data of the position in the X, Y, and Z directions;

[0016] Define the stateful prediction equation: is the predicted value at time k, x k―1 is the actual value at time k-1, F is the state transfer matrix, Among them, △t is the time difference between time K and time K-1;

[0017] Define the prediction equation with covariance: in, is the predicted state covariance matrix, F T is the transpose of the state transfer matrix F, Q is the process noise covariance matrix, and the initial state covariance matrix

[0018]

[0019] is the estimated value of the initial uncertainty of the state variables;

[0020] An observation matrix H is set for the acceleration data. The Kalman gain is Where R is the measurement noise covariance matrix, H T is the transpose of the observation matrix H;

[0022] Define the state update equation as Among them, z k is the observed value, x k To update the predicted value;

[0023] The covariance update equation is Where I is the identity matrix, P k is the updated covariance;

[0024] After filtering, the data preprocessing module updates the corresponding predicted value x k and update the covariance P k Sent to the deep learning model for subsequent processing.

[0025] The technical effect achieved by the present invention is: by running the above-mentioned system on the above-mentioned radio frequency therapy device, the sensor data is filtered, and the filtered data is input into the convolutional neural network for classification and prediction, and the motion trajectory formed by the current and subsequent user operations is judged by predicting the data N seconds ago; the device can adjust the working power according to the motion trajectory, so as to achieve the purpose of real-time control of the output power of the radio frequency therapy device and reduce the effect of user burns. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a schematic block diagram of the present invention;

[0027] Figure 2 This is the normal data waveform example of the present invention;

[0028] Figure 3 The data waveform diagram of the large circle of the present invention;

[0029] Figure 4 The data waveform diagram of the small circle of the present invention;

[0030] Figure 5 A data waveform diagram of the jitter of the present invention;

[0031] Figure 6 This is an example of the data waveform of the stop according to the present invention. DETAILED DESCRIPTION

[0032] In order to make the above-mentioned objects, features and advantages of the present invention more understandable, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from this description, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific embodiments disclosed below.

[0033] like Figure 1 As shown, the present invention discloses a radio frequency therapeutic device, the radio frequency therapeutic device is provided with a control module, the control module is electrically connected with an acceleration sensor and a data processing module, the control module is used to control the operation of the device, the acceleration sensor obtains the acceleration data of the device in the X, Y, and Z directions, and the data processing module performs processing on the acceleration data;

[0034] The radio frequency therapeutic apparatus runs a motion trajectory recognition system, which includes: a data preprocessing module and a deep learning model, the data processing module is used for the operation of the data preprocessing module, and the deep learning model runs in the control module or the data processing module;

[0035] The data preprocessing module filters the data through a Kalman filter and sends the processed data to the deep learning model, and the deep learning model analyzes and learns the data sent by the data preprocessing module, and judges and recognizes the motion trajectory formed by the user's operation behavior through the trained model;

[0036] The deep learning model processes the input data through a convolutional neural network algorithm and maps the processing results to data labels, which include: normal, large circle, small circle, jitter and stop. The large circle indicates that the user makes circular motions larger than a preset diameter range (for example, a diameter range of 2 cm) during use. The small circle indicates that the user makes circular motions smaller than a preset diameter range (for example, a diameter range of 2 cm) during use. The jitter indicates that the skin area contacted by the radiofrequency therapeutic device during use has hardly changed while the device itself has moved. The stop indicates that the user stops moving in a certain area during use and the stop time is greater than a preset time value (for example, the preset time value is 1.5 seconds). The normal indicates the normal sliding operation of the user except for the large circle during use.

[0037] It should be noted that, with respect to the jitter, if the change in the skin area contacted by the radiofrequency therapeutic device during use by the user is within a preset diameter range (eg, 0-0.5 cm), it is considered to be almost unchanged.

[0038] like Figure 2 As shown, Figure 2 Each line in is a schematic diagram of a group of normal motion trajectories, that is, the normal corresponding to the data label, Figure 2 From top to bottom, there are four groups of schematic diagrams, and each group of schematic diagrams of normal motion trajectories includes three schematic diagrams, which are the X-axis acceleration schematic diagram (the horizontal axis is time, and the vertical axis is the X-axis acceleration x_acc), the Y-axis acceleration schematic diagram (the horizontal axis is time, and the vertical axis is the Y-axis acceleration y_acc) and the Z-axis acceleration schematic diagram (the horizontal axis is time, and the vertical axis is the Z-axis acceleration z_acc);

[0039] like Figure 3 As shown, Figure 3 A set of motion trajectory schematic diagram examples of the large circle of the data label;

[0040] like Figure 4 As shown, Figure 4 A set of motion trajectory schematic diagram examples of the small circle of the data label;

[0041] like Figure 5 As shown, Figure 5 A set of motion trajectory schematic diagram examples of the jitter of the data tag;

[0042] like Figure 6 As shown, Figure 6 Each line in is a schematic diagram of a set of stop motion trajectories, that is, the stop corresponding to the data label, Figure 6 There are three groups of schematic diagrams from top to bottom, and each group of schematic diagrams of the stop motion trajectory includes three schematic diagrams;

[0043] Combination Figure 2-Figure 6 It can be seen that when the radio frequency therapy device performs any movement corresponding to the data tags, the change of its three-axis acceleration has a clear trend. The deep learning model can determine the device motion trajectory n seconds ago based on the sensor data n seconds ago and predict the next motion trajectory, thereby realizing the recognition and judgment of the user's current operation.

[0044] Figure 2 , Figure 3 and Figure 4-Figure 6 By comparison, it can be seen that the change trends, amplitudes and peak intervals of the three-axis acceleration of the normal, large circle and small circle, the jitter and the stop are obviously different. For example, the three-axis acceleration data of the large circle and the small circle have obvious differences in peak intervals, and the peak interval of the small circle is relatively smaller; the frequency and trend of acceleration changes of the first group of normal three-axis acceleration data are obviously different from those of the three-axis acceleration data of the large circle. Therefore, the deep learning model can effectively identify the input data and map it to the corresponding data label.

[0045] Specifically, the convolutional neural network algorithm structure of the deep learning model includes: an input layer, a convolution layer, an activation layer, an average pooling layer and at least two connection layers, and the connection layer maps the result to the data label. In particular, according to experimental results, the average pooling layer is generally preferred rather than the maximum pooling layer, which achieves better results.

[0046] Furthermore, the data preprocessing module uses a segmented sliding window to intercept the data of the radiofrequency therapeutic device, and the segmented sliding window is: intercepting a period of n seconds (such as 1.5 seconds) of data every x seconds (such as 0.1 seconds), and the deep learning model marks the intercepted data with the data label;

[0047] For example, according to the sampling frequency of the three-axis acceleration sensor, for example, the sampling frequency is 400 data samples per second, then when intercepting data, 1.5 seconds of data is intercepted to ensure that the data volume can support the correct conclusion; every 0.1 second interval, that is, 40 data are intercepted again, and the long segment of data collected when collecting data is intercepted into multiple segments of data with a total length of 600 points at intervals of 40 points at the starting point;

[0048] The purpose of using the segmented sliding window for data capture is to avoid excessive data processing burden while ensuring that the amount of data is sufficient to support correct recognition conclusions.

[0049] Specifically, when the data label is the normal or large circle, the system controls the radio frequency therapeutic apparatus to maintain the set power and continue to operate;

[0050] When the data label is any one or more of the small circle, the shaking, and the stopping, the system controls the radio frequency therapeutic device to reduce the operating power or stop the operation and outputs a warning signal to the user.

[0051] Specifically, the data preprocessing module filters the acquired data through a Kalman filter. During the filtering process, a state vector x is defined for the acquired position, velocity and acceleration data. The state vector Among them, p x 、p y 、p z is the coordinate data of the position, v x 、v y 、v z is the velocity data of the position in the X, Y, and Z directions, a x 、a y 、a z is the acceleration data of the position in the X, Y, and Z directions;

[0052] Define the stateful prediction equation: is the predicted value at time k, x k―1 is the actual value at time k-1, F is the state transfer matrix, Among them, △t is the time difference between time K and time K-1;

[0053] Define the prediction equation with covariance: in, is the predicted state covariance matrix, F T is the transpose of the state transfer matrix F, Q is the process noise covariance matrix, and the initial state covariance matrix

[0054]

[0055] is the estimated value of the initial uncertainty of the state variables;

[0056] An observation matrix H is set for the acceleration data. The Kalman gain is Where R is the measurement noise covariance matrix, H T is the transpose of the observation matrix H;

[0058] Define the state update equation as Among them, z k is the observed value, x k To update the predicted value;

[0059] The covariance update equation is Where I is the identity matrix, P k is the updated covariance;

[0060] After filtering, the data preprocessing module updates the corresponding predicted value x k and update the covariance P k The data is sent to the deep learning model for subsequent processing. The deep learning model maps the processing structure to the data tag. Finally, the radio frequency therapy device controls its own output power according to the data tag.

[0061] To summarize, by running the above system on the above radio frequency therapy device, the sensor data is filtered, and the filtered data is input into the convolutional neural network for classification and prediction. The motion trajectory formed by the current and subsequent user operations is judged by predicting the data N seconds ago; the device can adjust the working power according to the motion trajectory, so as to achieve the purpose of real-time control of the output power of the radio frequency therapy device and reduce the effect of user burns.

[0062] The above-described embodiments only express one or more implementation modes of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the patent of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A radiofrequency therapeutic apparatus, characterized in that: The radio frequency therapeutic device is provided with a control module, which is electrically connected to the acceleration sensor and the data processing module. The control module is used to control the operation of the device. The acceleration sensor obtains the acceleration data of the device in the X, Y and Z directions, and the data processing module processes the acceleration data.

2. A motion trajectory recognition system, characterized in that: The system is applied to the radiofrequency therapeutic apparatus as claimed in claim 1, and the system comprises: a data preprocessing module and a deep learning model; The data preprocessing module filters the data through a Kalman filter and sends the processed data to the deep learning model, and the deep learning model analyzes and learns the data sent by the data preprocessing module, and judges and recognizes the motion trajectory formed by the user's operation behavior through the trained model; The deep learning model processes the input data through a convolutional neural network algorithm and maps the processing results to data labels, which include: normal, large circle, small circle, jitter and stop. The large circle indicates that the user makes circular motion larger than a preset diameter range during use; the small circle indicates that the user makes circular motion smaller than a preset diameter range during use; the jitter indicates that the skin area contacted by the radio frequency therapy device has hardly changed during use while the device itself has moved; the stop indicates that the user stops moving in a certain area during use and the stop time is greater than a preset time value; and the normal indicates the user's normal sliding operation except for the large circle during use.

3. The motion trajectory recognition system according to claim 2, characterized in that: The convolutional neural network algorithm structure of the deep learning model includes: an input layer, a convolution layer, an activation layer, an average pooling layer and at least two connection layers, and the connection layer maps the results to the data labels.

4. The motion trajectory recognition system according to claim 2, characterized in that: The data preprocessing module uses a segmented sliding window to intercept the data of the radiofrequency therapeutic device. The segmented sliding window is: intercepting a period of n seconds of data every x seconds, and the deep learning model marks the intercepted data with the data label.

5. The motion trajectory recognition system according to claim 2, characterized in that: When the data label is the normal or large circle, the system controls the radio frequency therapeutic apparatus to continue to operate at the set power; When the data label is any one or more of the small circle, the shaking, and the stopping, the system controls the radio frequency therapeutic device to reduce the operating power or stop the operation and outputs a warning signal to the user.

6. The motion trajectory recognition system according to claim 2, characterized in that: The data preprocessing module filters the acquired data through a Kalman filter. During the filtering process, a state vector x is defined for the acquired position, velocity and acceleration data. The state vector Among them, p x 、p y 、p z is the coordinate data of the position, v x 、v y 、v z is the velocity data of the position in the X, Y, and Z directions, a x 、a y 、a z is the acceleration data of the position in the X, Y, and Z directions; Define the stateful prediction equation: is the predicted value at time k, x k―1 is the actual value at time k-1, F is the state transfer matrix, Among them, △t is the time difference between time K and time K-1; Define the prediction equation with covariance: in, is the predicted state covariance matrix, F T is the transpose of the state transfer matrix F, Q is the process noise covariance matrix, and the initial state covariance matrix is the estimated value of the initial uncertainty of the state variables; An observation matrix H is set for the acceleration data. The Kalman gain is Where R is the measurement noise covariance matrix, H T is the transpose of the observation matrix H; Define the state update equation as Among them, z k is the observed value, x k To update the predicted value; The covariance update equation is Where I is the identity matrix, P k is the updated covariance; After filtering, the data preprocessing module updates the corresponding predicted value x k and update the covariance P k Sent to the deep learning model for subsequent processing.

Images