Estimation of the head tilt angle of toothbrush users

Abstract

The subject of this disclosure relates to a computer implementation method for estimating the head tilt angle of a user with a toothbrush while brushing teeth, which comprises the steps of: detecting operating parameters related to the toothbrush during the brushing time period using a toothbrush sensor; inputting the detected operating parameters into a machine learning model to estimate the user's head tilt angle at each point in time during the brushing time period; and generating the user's head tilt angle as the head tilt angle estimated by the first machine learning model.

Description

Technical Field

[0001] The subject matter of the present disclosure relates to a computer-implemented method for estimating the head tilt angle of a user of a toothbrush during brushing, a computer-implemented method for training a machine learning model for estimating a segment of a user's dental arch being brushed by a toothbrush user, a computer-implemented method for training a machine learning model for estimating the head tilt angle of a user of a toothbrush, and a non-transitory computer-readable medium.

Background Art

[0002] Modern toothbrushes include a body, a head with bristles, an inertial measurement unit (IMU) within the body, and a controller for monitoring and controlling various functions of the toothbrush. For example, the toothbrush can provide an alert when the head needs to be replaced after a certain number of uses, or an alert that the head needs to be replaced if the toothbrush is dropped.

Summary of the Invention

Problems to be Solved by the Invention

[0003] Such functions are controlled by algorithms stored in the memory device of the controller and executed by the processor of the controller. Some of those algorithms require input values related to various parameters of the toothbrush. Some inputs, such as angular velocity, can be directly detected from the IMU. However, some parameters are not readily available. For example, the head tilt angle of the user cannot be measured using the IMU.

[0004] The object of the subject matter of the present application is to improve the prior art.

Means for Solving the Problems

[0005] According to a first aspect of the present invention, a computer implementation method is provided for estimating the user's head tilt angle with respect to a toothbrush while brushing, the computer implementation method comprising the steps of: detecting operating parameters related to the toothbrush during the brushing period using a toothbrush sensor; inputting the detected operating parameters into a machine learning model to estimate the user's head tilt angle at each point in time during the brushing period; and generating the user's head tilt angle as the head tilt angle estimated by the machine learning model. Advantageously, the machine learning model enables the determination of the head tilt angle based on readily available operating parameters.

[0006] In one embodiment, the machine learning model has a long-term short-term memory recurrent neural network (LSTM).

[0007] In one embodiment, the machine learning model has an encoder between the input layer and the LSTM, a decoder between the LSTM and the output layer, the encoder has a convolutional neural network, and the classifier has a convolutional neural network.

[0008] In one embodiment, the operating parameters include acceleration and angular velocity, and the sensor is an inertial measurement unit.

[0009] According to another aspect of the present invention, a computer implementation method is provided for training a machine learning model for estimating the head tilt angle of a toothbrush user, the computer implementation method comprising: receiving a training dataset and ground truth for head tilt, wherein for each point in a brushing time period, the training dataset includes operating parameter values ​​and the ground truth for head tilt includes the head tilt angle of ground truth; using the machine learning model to estimate the head tilt angle at each point in the brushing time period using the individual operating parameter values ​​as input; calculating the loss between the estimated brushing angle and the head tilt angle of ground truth; and adjusting the parameter settings of the machine learning model to minimize the loss.

[0010] In one embodiment, the computer implementation method further includes the steps of capturing a video of the user brushing their teeth with a toothbrush, a marker on the user's head, and a marker on the toothbrush during a sample period, detecting the operating parameter values ​​at each point in time during the sample period using the sensor on the toothbrush, calculating the head tilt angle using the orientation of the marker at each point in time during the sample period, and generating the ground truth of the head tilt by generating the head tilt angle related to the operating parameter value at each point in time, and transmitting the ground truth of the head tilt angle to the machine learning model.

[0011] In one embodiment, the step of calculating the loss between the estimated head tilt angle and the ground truth head tilt angle is to determine the ground truth unit vector vgt and the estimated head tilt angle unit vector veba by calculating the cosine component and sine component of the ground truth head tilt angle and the estimated head tilt angle, respectively, for each point in the brushing time period. A step of calculating the loss L using TIFF2026521500000002.tif1957, wherein . represents the inner product between vgt and veba.

[0012] In one embodiment, the computer implementation method further includes the steps of: training a machine learning model (60) to estimate the segments of the user's dental arch being brushed by the toothbrush user; receiving a training dataset and the ground truth of the segments, wherein for each time point in the brushing time period, the training dataset includes operating parameter values ​​and the ground truth head tilt angle of the segments indicates the segments of the dental arch being brushed; and using the machine learning model, estimating the brushing segments at each time point in the brushing time period using individual operating parameter values ​​as input. The method includes the steps of: calculating the loss between the estimated brushing segment and the segment indicated by the label above; adjusting the parameter settings of the machine learning model to minimize the loss; reinitializing the parameters related to the final layer of the machine learning model; and using the machine learning model to estimate the head tilt angle at each point in the brushing time period using individual operating parameter values ​​as input, calculating the loss between the estimated head tilt angle and the head tilt angle of the ground truth, and fine-tuning the machine learning model by adjusting the parameter settings of the machine learning model to minimize the loss.

[0013] In one embodiment, the computer implementation method further includes the steps of capturing a video of a user brushing their teeth with a toothbrush over a sample period, detecting the operating parameter values ​​at each point in time during the sample period using the toothbrush's sensor, displaying the video to the user of the user interface, receiving multiple user inputs indicating the brushing segment currently being brushed by the toothbrush in the video, and generating ground truth of segments by labeling each point in time during the sample period using the segments from the multiple user inputs, and transmitting the ground truth to the machine learning model.

[0014] In one embodiment, inputting detected operating parameters into a machine learning model to estimate the segment of the dental arch in contact with the bristles at each point in time during the brushing period involves inputting segments of the dental arch from a list, which includes the maxillary anterobuccal, maxillary anterolingual, maxillary right buccal, maxillary right occlusal surface, maxillary right lingual, maxillary left buccal, maxillary left occlusal surface, maxillary left lingual, mandibular anterobuccal, mandibular anterolingual, mandibular right buccal, mandibular right occlusal surface, mandibular right lingual, mandibular left buccal, mandibular left occlusal surface, and mandibular left lingual.

[0015] According to one aspect of the present invention, a temporary or non-temporary computer-readable medium is provided on which instructions are stored, and when these instructions are executed by a computer, the computer causes the computer to perform a computer implementation method of any aspect or embodiment described above.

[0016] These and other aspects of the present invention will become apparent from the embodiments described below and will be explained with reference to those embodiments. [Brief explanation of the drawing]

[0017] [Figure 1] This figure shows a schematic block diagram of a toothbrush according to one or more embodiments. [Figure 2] Figure 1 shows a diagram illustrating the upper and lower dental arches of the user of the toothbrush. [Figure 3] This figure shows a block diagram of a machine learning model trained to estimate the head tilt angle of a toothbrush user, according to one or more embodiments. [Figure 4] This figure shows a block diagram of a machine learning model trained to estimate the segments of the dental arch being brushed by the toothbrush shown in Figure 1, according to one or more embodiments. [Figure 5] This figure shows a schematic configuration of the training laboratory used to collect training data for ground truth, which is used to train the machine learning model shown in Figure 3. [Figure 6]Figure 5 shows a schematic diagram of the toothbrush used in the training laboratory. [Figure 7] Figure 5 shows a schematic diagram of the head marker used in the training laboratory. [Figure 8] This figure shows a flowchart summarizing the steps related to a computer implementation method for estimating the user's head tilt angle while brushing their teeth using the machine learning model shown in Figure 3, according to one or more embodiments. [Figure 9] This figure shows a flowchart summarizing the steps related to a computer implementation method for training the machine learning model shown in Figure 3, according to one or more embodiments. [Modes for carrying out the invention]

[0018] Embodiments of the present invention are best understood with reference to the accompanying drawings.

[0019] At least some of the exemplary embodiments described in this book can be constructed, in part or in whole, using dedicated special-purpose hardware. As used in this book, terms such as "component", "module", or "unit" are not limited to, but can include, a circuit, a field-programmable gate array (FPGA), or an application-specific integrated circuit (ASIC) in the form of a discrete element or an integrated element that performs a specific task or provides a related function. In some embodiments, the described elements can be configured to be disposed on a tangible, persistent, addressable storage medium and configured to be executed on one or more processors. These functional elements can include, in some embodiments, components such as software components, object-oriented software components, class components, task components, processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. Exemplary embodiments are described with reference to the components, modules, and units discussed in this book, but such functional elements can be integrated into fewer elements or separated into additional elements. Although various optional function combinations are described in this book, it should be understood that the described functions can be combined in any suitable combination. In particular, the functions of any exemplary embodiment can be appropriately combined with the functions of any other embodiment, provided that such combinations are not mutually exclusive. In this book, the terms "having" or "including" mean including the specified elements, but do not exclude the existence of other elements.

[0020] Referring to FIG. 1, the toothbrush 10 includes an elongated body forming a handle 12 and a head 14 provided at one end of the handle 12. The toothbrush also includes an energy storage module 16, a controller 18, and a sensor 20 within the handle.

[0021] The energy storage module 16 can be a battery. The battery can be a secondary battery or a rechargeable battery. The controller can include a processor 22 and a storage device 24. The computer-implemented method described below can be embodied as electronic data defining instructions stored in the storage device as a non-transitory computer-readable medium, which, when executed by the processor, causes the processor to execute an individual computer-implemented method. The instructions are loaded into the storage device and can thus be embodied as a transitory computer-readable medium.

[0022] The sensor 20 can be an inertial measurement unit (IMU) sensor. The IMU sensor can be configured to detect rotational speed and acceleration. The rotational speed and acceleration can be detected in three axes, which can be orthogonal coordinate axes including the x-axis, y-axis, and z-axis. The energy storage module 16 is connected to the controller 18 and the sensor 20 and can supply power to both of them.

[0023] The head 14 can include bristles 26 for brushing the user's teeth.

[0024] Referring to FIG. 2, the user's teeth are classified into an upper dental arch 30 and a lower dental arch 32. The upper and lower dental arches should each include the following tooth types: central incisors 34, lateral incisors 35, canines 36, premolars 37, and molars 38. Each tooth type other than the central incisors 34 and lateral incisors 35 includes a buccal side, a lingual side, and an occlusal surface. The central incisors and lateral incisors include a buccal side and a lingual side but do not have an occlusal surface.

[0025] The dental arch can be divided into different segments. Tooth types can be grouped within the segments. There can be 16 segments. The maxillary dental arch 30 may have 8 segments, and the mandibular dental arch 32 may also have 8 segments. The segments can include 2 segments on the anterior side 42 of each arch, 3 segments on the right side 44 of each arch, and 3 segments on the left side 46 of each arch. The central incisors 34 and lateral incisors 35 are located on the anterior side 42 of each arch. The boundary between the incisors (e.g., lateral incisors) and the canines 36 may be the boundary between the anterior and left / right 42, 44 segments.

[0026] Thus, the segment includes the anterobuccal side of the maxilla, the anterolingual side of the maxilla, the right buccal side of the maxilla, the right occlusal surface of the maxilla, the right lingual side of the maxilla, the left buccal side of the maxilla, the left occlusal surface of the maxilla, the left lingual side of the maxilla, the anterobuccal side of the mandible, the anterolingual side of the mandible, the right buccal side of the mandible, the right occlusal surface of the mandible, the right lingual side of the mandible, the left buccal side of the mandible, the left occlusal surface of the mandible, and the left lingual side of the mandible.

[0027] In other embodiments, the number of segments may differ, and the segments may contain different numbers of teeth. For example, there may be segment definitions in which each segment contains a single tooth.

[0028] Referring to Figure 3, a machine learning model is provided that uses the detected motion parameters as input to estimate the angle of the user's head tilt relative to the toothbrush at each point in time during brushing. This machine learning model can be called the first machine learning model 50.

[0029] The first machine learning model may include a recurrent neural network, which may be a long-term short-term memory recurrent neural network (LSTM). The first machine learning model 50 has an input 51, an encoder 52, an LSTM 53, a decoder 54, and an output 55. The encoder 52 includes a convolutional neural network 56 (or convolutional layer) and a batch normalization layer 57. The decoder 54 includes a convolutional neural network 58 (or convolutional layer) and a normalization L2 layer 59.

[0030] In encoder 52, the input operating parameters are filtered (linearly) by a convolutional layer and scaled by a batch normalization layer. This filtered signal is then fed into an LSTM layer that updates its internal state (also called the hidden state). The hidden state of the LSTM is then used as input to the decoder, which maps the hidden state to two channels (using a convolutional layer with 2 output channels and a kernel size of 1). As a final step, these two channels are normalized to represent vectors of length 1.

[0031] Referring to Figure 4, a second machine learning model 60 is provided that uses motion data as input to estimate the segment of the dental arch that the bristles are in contact with at each time point in the brushing time period.

[0032] The second machine learning model 60 includes an input 61, an encoder 62, an LSTM 63, a classifier 64, and an output 65. The encoder 62 includes a convolutional neural network 66 (or convolutional layer) and a batch normalization layer 67. The classifier 64 includes a first group of three layers, which includes a convolutional neural network 68 (convolutional layer), a batch normalization layer 69, and an activation function 70, and a second group, which includes the same three layers in the same order as the first group. The activation function 70 can be a ReLU activation function. The classifier 64 also includes a convolutional neural network 71 (or convolutional layer) and a softmax layer 72. The softmax layer includes multiple nodes representing a probability distribution, where the sum of the probabilities of the multiple nodes is 1.

[0033] In order to train the first and second machine learning models, the second machine learning model is trained first.

[0034] The ground truth of a segment can be generated by capturing a video of a user brushing their teeth with a toothbrush over a sample period, detecting the operating parameter values ​​at each point in time during the sample period using the toothbrush's sensors, displaying the video to the toothbrush user using a user interface, receiving multiple user inputs indicating the segment of the dental arch that the toothbrush is currently brushing in the video, and then having the toothbrush user manually label each point in time during the sample period using the segments from the multiple user inputs. IMU data, or the operating parameters at each point in time, can be incorporated as a training dataset, and the ground truth data includes labels indicating the segment of the dental arch being brushed.

[0035] The method for training a machine learning model includes the steps of: receiving a training dataset and ground truth for the segments; using the machine learning model to estimate the brushing segments for each point in the brushing time period, using individual operating parameter values ​​as input; calculating the loss between the estimated brushing segments and the labeled segments; and adjusting the parameter settings of the machine learning model to minimize the loss. Parameter adjustment can be performed using an optimization algorithm. Parameter settings may refer to weights in the machine learning model as a whole.

[0036] Referring to Figures 5 through 7, another aspect relates to a computer implementation for training a first machine learning model to estimate the head tilt angle of a toothbrush user. This method first requires the collection of training data and the generation of corresponding ground truth data. The training data can be collected in laboratory 76 where the ground truth is constructed.

[0037] The laboratory includes a toothbrush 10, a headset 78, a camera 80, and a computing system 82. The toothbrush 10 includes several markers 84 fixed to it. The markers are detected by the camera 80, and as a result, the orientation of the toothbrush can be detected. The headset may also include several markers 86 detected by the camera 80 to estimate the user's head tilt angle.

[0038] This method includes the steps of: capturing a video of the user brushing their teeth with a toothbrush during a sample period, the user wearing a marker on their head, detecting the motion parameter values ​​at each point in time during the sample period using the sensor on the toothbrush, calculating the head tilt angle using the orientation of the marker at each point in time during the sample period, generating the ground truth head tilt angle related to the motion parameter value for each point in time, thereby generating the ground truth of head tilt; and transmitting the ground truth of head tilt angle to the machine learning model.

[0039] This method includes the steps of: receiving a training dataset and ground truth data for head tilt, wherein for each time point in the brushing time period, the training dataset includes operating parameter values ​​and the ground truth data for head tilt includes the ground truth head tilt angle; using the machine learning model described above, estimating the head tilt angle at each time point in the brushing time period using individual operating parameter values ​​as input; calculating the loss between the estimated brushing angle and the ground truth head tilt angle; and adjusting the parameter settings of the machine learning model to minimize the loss. The adjustment is performed using an optimization algorithm, and the parameter settings may refer to the weights of the machine learning model.

[0040] The step of calculating the loss between the estimated brushing angle and the ground truth brushing angle includes determining the ground truth unit vector vgt and the estimated brushing angle unit vector veba by calculating the cosine component and sine component of the ground truth brushing angle and the estimated brushing angle, respectively, for each point in the brushing time period. A step of calculating the loss L using TIFF2026521500000003.tif1028, wherein . represents the inner product between vgt and veba, and a step of .

[0041] In the training procedure, IMU data is used as input, and the brushing angle is used as the ideal / ground truth that we want to obtain. Therefore, when the AI ​​model predicts the brushing angle, it can use this ideal brushing angle and the predicted brushing angle and minimize the difference between them. Since it is desirable to predict the angle, it is desirable that the metric being minimized (also called the loss function) converges. If we simply use the mean absolute angle difference, we may be troubled by the cyclical definition of these angles. An angle of -179 degrees is only 2 degrees away from an angle of +179 degrees. To solve this problem, it is possible to predict the cosine and sine components (essentially vectors of length 1) of the brushing angle and similarly convert the ground truth to the cosine and sine of the ground truth angle.

[0042] While it is possible to train the first machine learning model using random weight initialization, it should be noted that in certain situations, it may be better to start with a pre-trained second machine learning model and utilize transfer learning. This is because the ground truth data and the training data for training the second machine learning model do not need to be obtained in a laboratory. Instead of a camera used in the laboratory, the user's smartphone can be used outside the laboratory.

[0043] To train a first machine learning model, this method includes the steps of: providing a second machine learning model trained using the computer implementation method described above; reinitializing parameters related to the final layer of the machine learning model; and fine-tuning the machine learning model by using the machine learning model to estimate the brushing angle for each point in the brushing time period using individual operating parameter values ​​as input; calculating the loss between the estimated brushing angle and the brushing angle of ground truth; adjusting the parameter settings of the machine learning model to minimize the loss.

[0044] Referring to Figure 8, a computer implementation method for estimating the user's head tilt angle with respect to a toothbrush during brushing can be summarized as including the steps of: S100, which uses a toothbrush sensor to detect operation parameters related to the toothbrush during the brushing period; S102, which inputs the detected operation parameters into a machine learning model to estimate the user's head tilt angle at each point in time during the brushing period; and S104, which generates the user's head tilt angle as the head tilt angle estimated by the first machine learning model.

[0045] Referring to Figure 9, a computer implementation for training a machine learning model to estimate the head tilt angle of a toothbrush user can be summarized as including the steps of: receiving a training dataset and ground truth data for head tilt, wherein for each time point in a brushing time period, the training dataset includes operating parameter values ​​and the head tilt angle of the ground truth; using a machine learning model to estimate the head tilt angle for each time point in a brushing time period, using the individual operating parameter values ​​as input; calculating the loss between the estimated brushing angle and the head tilt angle of the ground truth, and adjusting the parameter settings of the machine learning model to minimize the loss, in a step S206.

[0046] Although the present invention has been illustrated and described in detail in the figures and description, such illustrations and description are merely illustrative and not limiting. The present invention is not limited to the disclosed embodiments.

[0047] Other variations of the disclosed embodiments can be understood and implemented by those skilled in the art practicing the claimed invention from the drawings, description and appended claims. In the claims, the word “has” does not exclude other elements or steps, and the indefinite article “a” or “an” does not exclude plurality. One processor or other unit can perform the functions of multiple items described in the claims. Thus, the mere fact that certain means are described in different dependent claims does not mean that combinations of these means cannot be used advantageously. Any reference numerals in the claims should not be construed as limiting the scope of the invention.

[0048] The various features disclosed in this book can be summarized in the following appendix.

[0049] Note 1. A computer implementation method for estimating the user's head tilt angle with respect to a toothbrush while brushing teeth, comprising the steps of: detecting operation parameters related to the toothbrush during the brushing time period using a toothbrush sensor; inputting the detected operation parameters into a machine learning model to estimate the user's head tilt angle at each point in time during the brushing time period; and generating the user's head tilt angle as the head tilt angle estimated by the machine learning model.

[0050] Note 2. The computer implementation method described in Note 1, wherein the machine learning model has a long-term memory recurrent neural network (LSTM).

[0051] Note 3. The computer implementation method described in Note 2, wherein the machine learning model has an encoder between the input layer and the LSTM, a decoder between the LSTM and the output layer, the encoder has a convolutional neural network, and the classifier has a convolutional neural network.

[0052] Note 4. The computer implementation method described in Note 1, wherein the above operating parameters include acceleration and angular velocity, and the above sensor is an inertial measurement unit.

[0053] Note 5. A computer implementation method for training a machine learning model to estimate the head tilt angle of a toothbrush user, comprising the steps of: receiving a training dataset and ground truth of head tilt, wherein at each point in the brushing time period, the training dataset includes operating parameter values ​​and the ground truth of head tilt data includes the ground truth head tilt angle; using the machine learning model to estimate the head tilt angle for each point in the brushing time period using individual operating parameter values ​​as input; calculating the loss between the estimated brushing angle and the ground truth head tilt angle; and adjusting the parameter settings of the machine learning model to minimize the loss.

[0054] Note 6. The computer implementation method described in Note 5, comprising the steps of: capturing a video of the user brushing their teeth with a toothbrush during a sample period, and markers on the user's head and the toothbrush; detecting the operating parameter values ​​at each point in time during the sample period using the sensor on the toothbrush; calculating the head tilt angle at each point in time during the sample period using the orientation of the markers; generating the ground truth of the head tilt angle related to the operating parameter value for each point in time; and transmitting the ground truth of the head tilt angle to the machine learning model.

[0055] Note 7. The step of calculating the loss between the estimated head tilt angle and the head tilt angle of the ground truth is to determine the ground truth unit vector vgt and the estimated head tilt angle unit vector veba by calculating the cosine component and sine component of the head tilt angle in the ground truth and the estimated head tilt angle, respectively, for each point in the brushing time period, The computer implementation method described in Appendix 5, comprising the steps of calculating loss L using TIFF2026521500000004.tif928, where . represents the inner product between vgt and veba.

[0056] Note 8. Steps include: training a machine learning model (60) to estimate the segments of a user's dental arch being brushed by a toothbrush user; receiving a training dataset and ground truth of segments, wherein for each time point in the brushing time period, the training dataset includes operating parameter values ​​and the ground truth head tilt angle of the segments indicates the segments of the dental arch being brushed; using the machine learning model to estimate the brushing segments for each time point in the brushing time period using individual operating parameter values ​​as input; and using the estimated brushing segments and The computer implementation method described in Appendix 5, comprising the steps of: calculating a loss between the segments indicated by the labels; adjusting the parameter settings of the machine learning model to minimize the loss; reinitializing the parameters related to the final layer of the machine learning model; and using the machine learning model to estimate the head tilt angle for each point in the brushing time period using individual operating parameter values ​​as input, calculating the loss between the estimated head tilt angle and the head tilt angle of the ground truth, and fine-tuning the machine learning model by adjusting the parameter settings of the machine learning model to minimize the loss.

[0057] Note 9. The computer implementation method described in Note 8, comprising the steps of: capturing a video of a user brushing their teeth with a toothbrush over a sample period; detecting the operating parameter values ​​at each point in time during the sample period using the toothbrush's sensor; displaying the video to the user of the user interface; receiving multiple user inputs indicating the brushing segment currently being brushed by the toothbrush in the video; and generating ground truth for the segments by labeling each point in time during the sample period using the segments from the multiple user inputs; and transmitting the ground truth to the machine learning model.

[0058] Note 10. The computer implementation method described in Note 9, wherein the step of inputting the detected operating parameters into a machine learning model in order to estimate the segment of the dental arch in contact with the bristles at each point in time during the brushing time period comprises inputting segments of the dental arch from a list, which include the maxillary anterobuccal, maxillary anterolingual, maxillary right buccal, maxillary right occlusal surface, maxillary right lingual, maxillary left buccal, maxillary left occlusal surface, maxillary left lingual, mandibular anterobuccal, mandibular anterolingual, mandibular right buccal, mandibular right occlusal surface, mandibular right lingual, mandibular left buccal, mandibular left occlusal surface, and mandibular left lingual.

Claims

1. In a computer implementation method for estimating the tilt angle of a user's head while brushing their teeth, The steps include: using the toothbrush sensor to detect operating parameters related to the toothbrush during the brushing time period; The steps include inputting the detected motion parameters into a machine learning model to estimate the user's head tilt angle at each point in time during the brushing period, A computer implementation method comprising the steps of generating the user's head tilt angle as the head tilt angle estimated by the machine learning model.

2. The computer implementation method according to claim 1, wherein the machine learning model has a long-term short-term memory recurrent neural network (LSTM).

3. The computer implementation method according to claim 2, wherein the machine learning model has an encoder between the input layer and the LSTM, a decoder between the LSTM and the output layer, the encoder has a convolutional neural network, and the classifier has a convolutional neural network.

4. The aforementioned operating parameters include acceleration and angular velocity, and the sensor is an inertial measurement unit. A computer implementation method according to any one of claims 1 to 3.

5. In a computer implementation method for training a machine learning model to estimate the head tilt angle of a toothbrush user, A step of receiving a training dataset and ground truth data of head tilt, wherein, for each point in the brushing time period, the training dataset includes operating parameter values ​​and the ground truth data of head tilt includes the ground truth head tilt angle. The steps include using the aforementioned machine learning model to estimate the head tilt angle for each point in time during the brushing period, using individual operating parameter values ​​as input, The steps include calculating the loss between the estimated brushing angle and the head tilt angle of the ground truth, A computer implementation method comprising the steps of adjusting the parameter settings of the machine learning model to minimize the loss.

6. During the sample period, a video of the user brushing their teeth with a toothbrush, and markers on the user's head and the toothbrush were captured. Using the toothbrush sensor, the operating parameter values ​​at each point in time during the sample period are detected. At each point in the aforementioned sample period, the head tilt angle is calculated using the orientation of the marker, and For each point in time, generate the ground truth of the head tilt angle associated with the aforementioned operating parameter value. This involves the step of generating ground truth for head tilt, The computer implementation method according to claim 5, further comprising the step of transmitting the ground truth of the head tilt angle to the machine learning model.

7. The step of calculating the loss between the estimated head tilt angle and the head tilt angle of the ground truth includes determining the ground truth unit vector vgt and the estimated head tilt angle unit vector veba by calculating the cosine component and sine component of the head tilt angle in the ground truth and the estimated head tilt angle, respectively, for each point in the brushing time period. A computer implementation method according to claim 5 or 6, comprising the step of calculating the loss L using, wherein represents the inner product between vgt and veba.

8. Steps include training a machine learning model to estimate the segments of a user's dental arch that are brushed by the toothbrush user, A step of receiving a training dataset and ground truth of segments, wherein, for each point in the brushing time period, the training dataset includes operating parameter values, and the ground truth of segments includes labels indicating the segments of the dental arch to be brushed. The steps include using the aforementioned machine learning model to estimate brushing segments for each point in the brushing time period, using individual operating parameter values ​​as input, The steps include: calculating the loss between the estimated brushing segment and the segment indicated by the label; The steps include adjusting the parameter settings of the machine learning model to minimize the loss, The steps include: reinitializing the parameters related to the final layer of the machine learning model; Using the aforementioned machine learning model, the head tilt angle for each point in the brushing time period is estimated using individual operating parameter values ​​as input. The loss between the estimated head tilt angle and the ground truth head tilt angle is calculated, and Adjust the parameter settings of the aforementioned machine learning model to minimize the loss. The computer implementation method according to any one of claims 5 to 7, further comprising the step of fine-tuning the machine learning model.

9. During the sample period, we captured videos of the user brushing their teeth with a toothbrush. Using the toothbrush sensor, the operating parameter values ​​at each point in time during the sample period are detected. Display a video to the user interface, The video receives multiple user inputs indicating the brushing segment currently being brushed by the toothbrush, and Using the segments from the multiple user inputs, each point in time during the sample period is labeled. This involves the step of generating the ground truth of the segment, The computer implementation method according to claim 8, further comprising the step of transmitting the ground truth to the machine learning model.

10. The step of inputting the detected operating parameters into a machine learning model in order to estimate the segment of the dental arch that the bristles are in contact with at each point in time during the aforementioned brushing time period is as follows: The computer implementation method according to claim 9, comprising inputting a segment of the dental arch from a list, wherein the list includes the maxillary anterobuccal, maxillary anterolingual, maxillary right buccal, maxillary right occlusal surface, maxillary right lingual, maxillary left buccal, maxillary left occlusal surface, maxillary left lingual, mandibular anterobuccal, mandibular anterolingual, mandibular right buccal, mandibular right occlusal surface, mandibular right lingual, mandibular left buccal, mandibular left occlusal surface, and mandibular left lingual.

Images