WO2025083058A1

WO2025083058A1 - Replaceable component type determination

Info

Publication number: WO2025083058A1
Application number: PCT/EP2024/079204
Authority: WO
Inventors: Wilhelmus Andreas Marinus Arnoldus Maria Van Den Dungen
Original assignee: Koninklijke Philips NV
Current assignee: Koninklijke Philips NV
Priority date: 2023-10-20
Filing date: 2024-10-16
Publication date: 2025-04-24
Anticipated expiration: 2026-04-20

Abstract

Proposed are schemes, solutions, concepts, designs, methods and systems pertaining to determining a type of a replaceable component to a personal care device that vibrates due to an actuator while in use. Specifically, a data set describing an operating parameter of the personal care device (during vibration) is used to generate a spectrogram image. The spectrogram image is input to a type classification machine learning algorithm, which outputs a predicted type (e.g., a predicted make, model, version, variety and/or brand) of the replaceable component of the personal care device. Accordingly, operating settings of the personal care device may be adjusted based on the predicted type, thus reducing occurrences of failures of the personal care device and improving efficacy of the personal care device.

Description

REPLACEABLE COMPONENT TYPE DETERMINATION

FIELD OF THE INVENTION

The present invention relates to determining a type of a replaceable component of a personal care device.

BACKGROUND OF THE INVENTION

Personal care devices (e.g., toothbrush, shaver, skin scrubber, etc.) may have different sorts of replaceable components (e.g., brush head, shaver head, etc.). For each replaceable component there may be many different types. That is, there may be replaceable components of different models, makes, versions, varieties or brands.

Knowing the type of a replaceable component of a personal care device can be beneficial for operation of the personal care device. The settings of operation of the personal care device may be altered in response to the type of the replaceable component, so that performance of the personal care device can be appropriately adapted. Indeed, adaptation of the operation settings may increase safety of the device, and reduce occurrences of faults and failures of the replaceable component and the personal care device itself. Moreover, appropriate adaptation of settings may also increase user safety.

There is therefore a need for a means for automatically determining the type of a replaceable component of a personal care device.

SUMMARY OF THE INVENTION

According to examples in accordance with an aspect of the invention, there is provided a method for determining a type of a replaceable component of a personal care device. The personal care device comprises an actuator that causes the personal care device to vibrate while in use. The method comprises: obtaining a data set describing an operating parameter of the personal care device during vibration of the personal care device; generating a spectrogram image based on at least part of the data set; and processing, with an type classification machine learning algorithm, the spectrogram image to predict a type of the replaceable component of the personal care device. Proposed is a method for determining a type of a replaceable component of a personal care device that vibrates due to movement of an actuator. Specifically, a data set describing an operating parameter of the personal care device during vibration is used to generate a spectrogram image. The spectrogram image is input to a type classification machine learning algorithm, which outputs a predicted type of the replaceable component of the personal care device. Accordingly, operating settings of the personal care device may be adjusted based on the predicted type, thus reducing occurrences of failures of the personal care device and improving efficacy of the personal care device.

The personal care device comprises an actuator that causes the personal care device to vibrate while in use. The actuator may be a component of the personal care device that is configured to affect a personal care function (e.g., a motor to drive movement of a brush head of a toothbrush), or may simply be a component that is configured to perform a different function, but incidentally causes a vibration of the personal care device. That is, the actuator may be any component that affects mechanical movement while in use, thereby causing the personal care device to vibrate.

Furthermore, the data set is obtained responsive to the vibration of the personal care device. In other words, the data set is obtained/acquired/generated as the personal care device vibrates due to movement of the actuator. Thus, features indicative of the vibration response of the personal care device may be present in the data set.

The data set of the personal care device describes an operating parameter during vibration of the personal care device. That is, the data set contains values indicative of a system vibrational response of the personal care device measured with, for example, an acceleration (in one or more dimensions), motor current, sound/noise while the device is in an operational mode (i.e., while the actuator is moving and thus causing vibration of the personal care device). The vibration response contained within the data set reflects the type of the replaceable component, as attaching different types of the replaceable component to the personal care device will result in a change in the vibration response (to movement of the actuator) of the personal care device.

From (at least a part of) the data set, a spectrogram image is generated. That is, the data set is processed using known methods in order to generate the spectrogram image. For example, the data set may be processed using fast Fourier transforms (FFTs) in order to generate the spectrogram image. The spectrogram image contains a plurality of FFTs, each corresponding to a point in time. Each FFT represents the strength of frequencies of the vibration response of the personal care device at a point in time captured by the data set. Subsequently, the spectrogram image is provided to the type classification machine learning algorithm. The type classification machine learning algorithm processes the spectrogram image and may identify various features indicative of the type of the replaceable component. Accordingly, a predicted type can be provided. The type may be indicative of the make, model, variety, brand, or version of the replaceable component.

To be clear, the personal care device may be, for example, an electric toothbrush/mouthpiece, an electric shaver, or a skin scrubber. Of course, the invention may be applied to various other personal care devices configured to provide a personal care function for a user. The personal care device may be handheld and/or portable.

The replaceable component may be any component of the personal care device whose type impacts operation of the personal care device. For example, the replaceable component may be a brush head, a shaver head, or any other component that may come into contact with a surface of a user, or a component thereof.

In some embodiments, the type classification machine learning algorithm may be a spectral image classifier. A spectral image classifier classifies features in an image into different classes based on their spectral signature. Accordingly, the algorithm may be well adapted to process/analyse the spectrogram image. More particularly, the type classification machine learning algorithm may be a CNN classifier or an FCNN classifier.

Furthermore, the type classification machine learning algorithm may be a CNN classifier comprising 3-6 convolutional layers.

The fewer convolutional layers in a CNN, the less computationally complex it is, requiring fewer memory and processing resources. However, they also become less powerful. 3-6 convolutional layers may provide a balance between size of the CNN classifier, and efficacy of the CNN classifier.

The data set may comprise accelerometer data describing an acceleration in at least one axis of the personal care device.

It has been shown that a data set comprising accelerometer data in just one dimension/one axis is sufficient to facilitate prediction of a type of a replaceable component of a personal care device. This may be relatively simply data to generate, and hardware to generate such data is present in many exiting personal care devices. Furthermore, many personal care devices comprise inertial motion units (IMUs) that capture acceleration data in three dimensions, with such data potentially further improving an accuracy of the prediction.

In the case that the accelerometer data describes acceleration in 3 axis, the method may further comprise calculating norm vector data describing an absolute vector length of acceleration based on the accelerometer data; and serializing the accelerometer data of each of the 3 axis and the norm vector data into a single data segment.

Alternatively, or additionally, the data set may comprise sound data describing a sound produced by the personal care device.

Sound/noise/audio data has been shown to provide sufficient information regarding vibration of the personal care device so as to facilitate replaceable component type prediction.

Alternatively, or additionally, the data set may comprise current data describing a motor current of the actuator.

Similarly, current data has also been shown to provide sufficient information regarding vibration of the personal care device so as to facilitate replaceable component type prediction. The current data may be acquired from the motor itself, or may be acquired via, for example, a battery to which the motor is connected. The motor may be the component of the actuator that affects movement of the actuator, thereby causing vibration of the personal care device.

Specifically, the obtained data set may comprise values describing the operating parameter of the personal care device spanning up to one second.

That is, the data set comprises values spanning up to one second, or is a signal that lasts for up to one second. By restricting the length of time spanned by the data set, processing and analysis may be simplified. Of course, the data set may span more or less time, but approximately one second enables accurate prediction of the type, whilst being of reasonable complexity.

The obtained data set may comprise values describing the operating parameter of the personal care device sampled at a rate of at least 800Hz. In some embodiments, the obtained data set may comprise values describing the operating parameter of the personal care device sampled at a rate of at least 1.6kHz.

It may be beneficial to sample the operating parameter at a rate of at least 800Hz in order to ensure that unique features of vibration of the personal care device are present in the dataset. 1.6kHz may be preferable to ensure no unique features are missed, although with increased data set size comes increased complexity.

The method may further comprise normalizing at least part of the data set. Accordingly, variations in the data set due to, for example differences (material, shape, condition, etc.) between personal care devices, or the context (e.g., position, environment, etc.) in which the personal care device is kept when acquiring the data set, may be accounted for when predicting the type of the replaceable component. This may result in a more accurate type prediction across personal care devices.

Example embodiments of the invention may provide that the data set describing the operating parameter is obtained responsive to the personal care device being in a static state.

The static state may be any state in which the personal care device is not being moved or vibrated by external means (e.g., due to action of the user). For example, the static state may correspond to when the personal care device is in a charging state, or in an idle state in the hands of a user. This may ensure that detection of unique vibration-related features in the dataset can be performed by the method.

More particularly, the data set describing the operating parameter may be obtained during a single operational mode of the actuator.

Indeed, the single operational mode may correspond to normal use of the personal care device (e.g., a single operation setting of the actuator of an electric toothbrush). This may mean that the user does not notice when the data set is being gathered, and may be gathered during typical use of the personal care device. In essence, this means that the actuator does not need to be controlled in a way that induces vibration of the personal care device in a specific way (e.g., a vibration sweep) in order to capture the data set. In other words, the actuator is merely controlled in a manner in which it would usually operate. Accordingly, the invention provides for the acquisition of the data set without operating the actuator outside normal operation conditions, settings and/or ranges.

Alternatively, the data set describing the operating parameter may be obtained while changing an operational mode of the actuator to vary an amplitude and/or duty cycle of motion of actuation (and therefore a resultant vibration response of the personal care device).

In contrast to a vibration sweep that is caused by an actuator operating in a mode or range outside typical operation conditions, an amplitude sweep may be performed. By obtaining the data set during the amplitude sweep, additional unique vibration response- related features may be present in the data set, which may be detected for prediction of the type of the replaceable component. In other words, as an alternative to a frequency/vibration sweep outside the normal operational range of the actuator, an amplitude and/or duty cycle sweep can be used to further enhance features present in the data set.

Predicting the type of the replaceable component of the personal care device may comprise classifying, with the type classification machine learning algorithm, the predicted type of the replaceable component of the personal care device into one of a plurality of type classes. In other words, the type of the replaceable component may be predicted to be in one of many classes reflecting the type of replaceable component. Each of the plurality of type classes may reflect the replaceable component being a certain model, make, version, variety, and/or brand.

The method may further comprise generating the type classification machine learning algorithm. Generating the type classification machine learning algorithm may comprise: generating, for a plurality of replaceable components of the personal care device each having a respective known type, a test data set describing an operating parameter of the personal care device during vibration of the personal care device; and training the type classification machine learning algorithm using a training algorithm configured to receive an array of training inputs and respective known outputs, wherein a training input comprises the test data sets of each of the plurality of replaceable components, and wherein a respective known output comprises the known type.

Accordingly, a type classification machine learning algorithm may be provided that is capable of accurately predicting a type of the replaceable component.

Each test data set may comprise data generated during user operation of the personal care device. This may be achieved during user tests.

Alternatively, or in addition, each test data set may comprise synthetic generated data. Such data may be valuable for increasing the volume of training data sets that the type classification machine learning algorithm may be trained upon. With increasing amounts of data, the algorithm may be increasingly accurate and robust.

According to another aspect of the invention, there is provided a computer program comprising computer program code means adapted, when said computer program is run on a computer, to implement a method of a proposed embodiment.

According to a further aspect of the invention, there is provided a personal care device. The personal care device comprises: a replaceable component; an actuator configured to cause the personal care device to vibrate while in use; a sensor for obtaining a data set describing an operating parameter of the personal care device during vibration of the personal care device; a processor configured to: generate a spectrogram image based on at least part of the data set; and process, with a type classification machine learning algorithm, the spectrogram image to predict a type of the replaceable component of the personal care device.

These and other aspects of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter.

BRIEF DESCRIPTION OF THE DRAWINGS

For a better understanding of the invention, and to show more clearly how it may be carried into effect, reference will now be made, by way of example only, to the accompanying drawings, in which:

Figure 1 presents a flow diagram of a method for determining a type of a replaceable component of a personal care device according to an embodiment;

Figure 2 is a personal care device according to an example embodiment;

Figure 3 provides a simplified block diagram of a computer within which one or more parts of an embodiment may be employed.

DETAIEED DESCRIPTION OF THE EMBODIMENTS

The invention will be described with reference to the Figures.

It should be understood that the Figures are merely schematic and are not drawn to scale. It should also be understood that the same reference numerals are used throughout the Figures to indicate the same or similar parts.

It should also be understood that the detailed description and specific examples, while indicating exemplary embodiments of the apparatus, systems and methods, are intended for purposes of illustration only and are not intended to limit the scope of the invention. These and other features, aspects, and advantages of the apparatus, systems and methods of the present invention will become better understood from the following description, appended claims, and accompanying drawings. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Proposed are schemes, solutions, concepts, designs, methods and systems pertaining to determining a type of a replaceable component of a personal care device that vibrates due to a mechanical actuator while in use. Specifically, a data set describing an operating parameter of the personal care device (during operation) is used to generate a spectrogram image. The spectrogram image is input to a type classification machine learning algorithm, which outputs a predicted type of the replaceable component of the personal care device. Accordingly, operating settings of the personal care device may be adjusted based on the predicted type, thus reducing or preventing the occurrence of faults and failures of the personal care device, as well as improving performance and safety of the personal care device by appropriate selection of settings.

Disclosed embodiments provide a method for replaceable component type assessment that is based on vibration-related operating parameters, such as data sets describing operating parameters obtained from microphones, accelerometers, actuator motor current, or current of batteries connected to the actuator. A short sample of the data set is converted to a spectrogram image and a neural network (e.g., a CNN style network) is used to classify the signal, and thus the type of the replaceable component, based on the features it finds in the spectrogram image. That is, simplification of processing is achieved by converting the data set to a spectrogram image, which can be effectively processed by a machine learning algorithm to identify features for determination of a type of the replaceable component. The disclosed method thus provides a sensitive and accurate type prediction means.

Furthermore, disclosed embodiments may also be implemented with hardware of existing personal care devices, negating the need for costly and complex additional components and sensor. That is, additional sensors (e.g., NFC sensors) are not required to implement the invention. This is due to a reliance on an operating parameter signal holding vibration response information, which may already be produced by sensors already implemented in personal care devices for various other functions. For example, the solution can make use of signals from 3 axis or single axis accelerometer, current of a motor of an actuator, a microphone, or a load sensor signal - many of which are commonly present in personal care devices.

By effectively determining the replaceable component type, a user experience may be improved. Indeed, type prediction can be used to adjust an operation of the personal care device to prevent potential faults due to improper operation when a replaceable component of a certain type is attached (e.g., a fraudulent replaceable component, or a replaceable component made of weaker materials). In addition, the adjustment of the operation may be used to enhance the operation of the personal care device to provide an improved user experience, including an improved performance and effectiveness of the device. Furthermore, in the case of medicalclass personal care devices, this information may be used to help prevent unsafe situations and provide traceable evidence when needed.

Taking the example of a toothbrush, disclosed embodiments may be implemented in existing hardware. This is the case even when an inertial motion unit (IMU) sensor is not present, due to the ability of the invention working with many vibration-dependent operating parameters (e.g., motor current, sound/noise signals, or a load sensor signal). Some embodiments provide that ~ 1 second of operating parameter values are converted to a spectrogram image, and fed to a small CNN model, in order to provide replaceable component (e.g., a brush head) type prediction. Thus, disclosed embodiments may be implemented as new software installed on existing devices due to limited memory and processing requirements.

Nevertheless, it is worth noting that the invention may be implemented on a wide range of personal care devices and replaceable components. The type of shaver head of a shaving device may be assessed, for example. The type of scrubbing head of a skin scrubbing device may also benefit from assessment. Thus, it is worth noting that while toothbrushes and brush heads may particularly benefit from the invention, the invention may also provide advantages to many other personal care devices.

Furthermore, the method of determining the type may be executed in the handle of the personal care device, on an external processor, or even in the cloud or on edge processing systems.

The data set used to predict the type of the replaceable component describes an operating parameter of the personal care device (as the personal care device vibrates responsive to movement of the actuator during normal operation). The operating parameter that can be used relate to a vibration response of the personal care device to movement of the actuator, such as acceleration (in one or more dimension), sound/noise, and motor/system current.

In some embodiments, the data set is obtained responsive to the personal care device being in a static state. For example, the static state may be when the personal care device is in a charging state, indicative of the personal care device being on a charging station, or otherwise static. The static state may also be when the user is handling the personal care device, but is not applying any load to the personal care device (e.g., holding a toothbrush, but not applying pressure to teeth, or holding a shaver, but not holding a shaving head against any hair). This may ensure that features related to motion/vibration caused by a user or other sources of motion and/or vibration contained in the data set are minimised, and in turn that the type classification machine learning algorithm can accurately predict the type.

Nonetheless, the data set may also be obtained in use. That is, the dataset may also be obtained while the personal care device is actively being used to perform an associated personal care function. However, a more complex machine learning algorithm may be required (e.g., increased layers), with a larger data set (e.g., acquired at a higher rate, with greater bit depth and for a longer span of time), in order to ensure accuracy of the predicted type. The data set may span a time of approximately 1 second, but may span a longer period of time to improve accuracy of prediction, or a shorter period of time to decrease computational complexity of prediction.

The dataset may be sampled at a rate of at least 800Hz, and preferably 1.6kHz to capture the frequency range containing unique features. Note that the rate may depend on the chosen operating parameter. For example, it may be preferable that acceleration data is captured at 1.6kHz, while sound and current may be captured at 16kHz. Indeed, the operational speed of the personal care device may influence the frequency response of the personal care device, thus impacting the rate at which data should be sampled. For example, a toothbrush operating at 200Hz may have data sampled at a rate 1.6kHz, while a toothbrush operating at 500Hz may require a higher sampling rate.

Furthermore, capturing the dataset at different bit depths (i.e. resolution) may also impact the computational complexity and accuracy of prediction. For sound data, 16 bit data may be captures, acceleration data may be captured at 12 bit, while current may be captured 8 or 16 bit.

Of course, the chosen operating parameter, data length, rate, and resolution of the data set may depend on the chosen personal care device upon which the invention is implemented. The skilled person would fully appreciate the range of selected parameters of the data set, and would be able to adapt said parameters for the chosen application.

Moreover, the data set may be further pre-processed before being used to predict the type of the replaceable component. Particularly advantageously, the data set may be normalized, so that variations in amplitude (e.g., caused by production variations in the personal care device and/or replaceable component) are accounted for. Thus, differences between individual personal care devices and replaceable components may be mitigated.

Specifically, when the operating parameter is acceleration, the acceleration data (e.g., from a 3-axis accelerometer) data format can be improved by pre-processing. By serializing or concatenation, the values of acceleration for each different axis (XYZ) may also include a norm (N) calculation as a fourth set of values. The 4 sets of values (XYZN) sampled over a time span of 2.5 second at 1600Hz will result in 16000 data points (similar to a microphone sampling at 16Khz for 1 second). This 3-axis accelerometer data format may be advantageous compared to, for example, single axis accelerometer data or data from a microphone or current data, as it also contains vibration direction information in the serialized data format. Normalization may be performed for each of the set of values XYZN to preserve directional information in the signal. Once the data set describing the operating parameter is obtained, and optionally pre- processed, the data set is then used to generate a spectrogram image, representing the signal strength of various frequencies in the vibration response of the personal care device over time, contained in the data set. Accordingly, a low resolution (e.g., 128x128 pixel) spectrogram image may be generated, which has been shown to be used to perform classification of type with very high accuracy. Of course, a higher resolution image may be generated to ensure accuracy of the type prediction, but this may require greater computational resources.

The generation of the spectrogram may be achieved using known methods, such as applying multiple fast Fourier transforms (FFTs) to the data set. Various applications and implementations of the FFTs to the dataset will be readily appreciated by the skilled person.

Once the spectrogram image is generated, it is provided to a type classification machine learning algorithm (e.g., a trained neural network). The type classification machine learning algorithm processes the spectrogram image to predict a type of the replaceable component.

The structure of an artificial NN (or, simply, neural network (NN)) is inspired by the human brain. Neural networks are comprised of layers, each layer comprising a plurality of neurons. Each neuron comprises a mathematical operation. In particular, each neuron may comprise a different weighted combination of a single type of transformation (e.g. the same type of transformation, sigmoid etc. but with different weightings). In the process of processing input data, the mathematical operation of each neuron is performed on the input data to produce a numerical output, and the outputs of each layer in the neural network are fed into the next layer sequentially. The final layer provides the output.

There are several types of neural network, such as convolutional neural networks (CNNs), fully connected neural networks (FCNNs) and recurrent neural networks (RNNs). Examples of the invention typically implement the type classification machine learning algorithm as a spectral image classifier, such as a CNN or an FCNN.

Specifically, the type classification machine learning algorithm may be a classical convolutional neural network (CNN) classifier with a limited number of layers. It has been shown that a CNN with 3-6 convolutional layers provides effective processing of the spectrogram image. Further, these small CNN based algorithms can easily be used in embedded firmware that may be present in existing personal care devices. Alternatively, as the sample sizes are small, cloud processing is also possible.

For a given spectrogram image (corresponding to a data set acquired during vibration of the personal care device, and thus associated with a given replaceable component), the type classification machine learning algorithm can predict a type of the replaceable component. The algorithm is trained on previous spectrograms associated with known types (e.g., known makes, models, versions and/or brands) of replaceable components.

Methods of training a machine-learning algorithm are well known. Typically, such methods comprise obtaining a training dataset, comprising training input data entries and corresponding training output data entries. An initialized machine-learning algorithm is applied to each input data entry to generate predicted output data entries. An error between the predicted output data entries and corresponding training output data entries is used to modify the machine-learning algorithm. This process can be repeated until the error converges, and the predicted output data entries are sufficiently similar to the training output data entries. This is commonly known as a supervised learning technique.

For example, weightings of the mathematical operation of each neuron may be modified until the error converges. Known methods of modifying a neural network include gradient descent, backpropagation algorithms and so on.

The training input data entries for the type classification machine learning algorithm correspond to example test data sets describing an operating parameter of the personal care device during vibration of the personal care device. The training output data entries correspond to a known type of the replaceable component, for each of the plurality of test data sets. That is, the machine learning algorithm is trained using a training algorithm configured to receive an array of training inputs and respective known outputs, wherein a training input comprises one of the test data sets, and a respective known output comprises the known type. In this way, the machine learning algorithm is trained to output a predicted type when provided with a data set describing an operating parameter of the personal care device during vibration of the personal care device.

In some cases, the type classification machine learning algorithm may be trained to classify the spectrogram image (and consequently the replaceable component) into one of a plurality of type classes. In one embodiment, the classification may be performed into an a first type class and a second type class, with a probability score provided per classification. The probability score provided per classification may be used to identify the type class that the replaceable component most likely belongs in, and therefore the likely type of the replaceable component.

The plurality of type classes may reflect a different model group, make group, version group, variety group, and/or brand group of the replaceable component. That is, the type class may specify that the replaceable component is of a particular model, make, version, variety and/or group. It will be appreciated that the type classification machine learning algorithm may be (at least partially) pre-trained before deployment for use in detecting the type of a replaceable component. Nevertheless, calibration and/or retraining may be performed response to use by an end-user to ensure effective type prediction.

The test data sets may be obtained from real user tests. That is, the test data sets may be obtained by sensing the data set describing the operating parameter (e.g., measuring acceleration, sound and/or motor current) as a user uses the personal care device, and with a known type of the replaceable component. Such data sets may be acquired when in use, for example by requiring input from a user as to the type of the replaceable component, or by inferring a type from context. In this way, real user loading states may be reflected in the test data set, such that the machine learning algorithm may be more likely to provide an accurate prediction in unseen conditions.

Additionally or alternatively, the test data sets may be synthetic generated data. In other words, the test data sets may be obtained by sensing the data set describing the operating parameter (e.g., measuring acceleration, sound and/or motor current) as a robotic testing setup uses the personal care device, and with a known type of the replaceable component. The robotic test setup may be based on real user data/loads to reflect real- world use cases. This may enable the generation of large sets of test data, spanning a wide variety of test conditions, in a relatively cheap and quick manner (compared to test data sets obtained from user operation). In turn, by providing more test data sets, an accuracy and efficacy of the machine learning algorithm may be improved. Furthermore, the test data set may be generated using data from computer simulations.

Turning to Fig 1., there is presented a flow diagram of a method for determining a type of a replaceable component of a personal care device.

To be clear, the type of the replaceable component relates to the specific version, make, model, variety and/or brand of the replaceable component. Each of these factors may impact the way in which the personal care device operates, and therefore determining the type may be important for appropriately adapting operating parameters.

The method relates to a personal care device that comprises an actuator that causes the personal care device to vibrate while in use. The personal care device may be a device that comprises an actuator that makes a motion, that results in a vibrational response of the personal care device. Also, the personal care device may vibrate, but the vibration may not form part of its function. Essentially, the personal care device comprises an actuator that may cause mechanical action that results in (a measurable) vibration of the personal care device, whether intended for the personal care function of the device or otherwise. Nevertheless, in some cases, the personal care device may be a vibrational personal care device that vibrates to perform a function.

The personal care device itself may be, for example, an electric toothbrush, a shaver, a skin scrubber, etc. The replaceable component may be any component of said personal care devices that deteriorate over time and/or with use of the personal care device, such that it may need replacement by a user. The deterioration over time or use may result in a reduced efficacy of the personal care function of the device. Thus, the user may need to swap an older replaceable component out for a new replaceable component. As a result, it may be useful to determine the type of the replaceable component that the user has attached, so that operating settings may be appropriately adapted/selected.

In step 110, a data set is obtained that describes an operating parameter of the personal care device during vibration of the personal care device.

The data set may comprise one or more of accelerometer data describing an acceleration in one axis of the personal care device, accelerometer data describing an acceleration in three axes of the personal care device, sound data describing a sound produced by the personal care device, and/or current data describing a motor current of the actuator. Other operating parameters for capturing features indicative of vibration of the personal care device may also be contemplated. However, the above described operating parameters may be particularly suitable for measuring vibration of the personal care device.

In some cases, the obtained data set may comprise values describing the operating parameter of the personal care device spanning up to one second. Of course, values spanning a greater range of time may be obtained (and subsequently used to generate a spectrogram image), but one second may be sufficient to provide a data set having extractable vibration- related features for prediction of the type of the replaceable component. Furthermore, the obtained data set may comprise values describing the operating parameter of the personal care device sampled at a rate of at least 800Hz to clearly capture vibration-related features.

Step 110 may also comprise an (optional) sub-step of pre-processing the data set. For example, the data set may be normalized in order to account for any variations between the same type of personal care device (e.g., material differences, slight faults, different environmental conditions). This may comprise altering all of the values in the data set so that they are referenced between 0 and 1, or -1 and 1, for example.

Furthermore, in the case that the data set comprises accelerometer data describing acceleration in three axis, then the pre-processing may involve calculating (using known methods) norm vector data describing an absolute vector length of acceleration using the accelerometer data. Such norm vector data may then be concatenated with serialized accelerometer data (in each of the 3 axis) to generate a data set comprising a single data segment of accelerometer data in each of the 3 axis and norm vector data.

Step 110 (of obtaining the data set) may be performed responsive to the personal care device being in a static state. That is, when the personal care device does not experience any load (i.e. is not in a use state) the data set may be obtained. This may indicate that the personal care device freely vibrates. Accordingly, vibrational noise in the obtained data set may be reduced. Namely, data artifacts or user induced artifacts in the data may be reduced (i.e. artifacts resulting from user induced motion, or other motion, of the personal care device). The static state may correspond to the personal care device being in a charging state, or merely being held idly in the hand of a user.

Furthermore, step 110 may be performed during a single operational mode of the actuator. In other words, the data set may be obtained responsive to/during regular/normal vibration of the personal care device. Normal vibration may correspond to a mode of vibration that the personal care device usually experiences during performance of a personal care function of the device. On the contrary to methods of frequency shifting, this may not require additional complex driving of the actuator of the personal care device.

Conversely, step 110 may be performed while changing an operational mode of the actuator to vary an amplitude and/or duty cycle of vibration of actuation. This may be considered to be a vibration amplitude sweep, which may enrich the data set with additional unique features for prediction of the type of the replaceable component.

Then, in step 120, a spectrogram image is generated based on/using (at least part of) the data set. The spectrogram image may be generated using a variety of known methods, both analog and digital. Most suitably may be to generate the spectrogram by applying a plurality of FFTs to the data set.

More particularly, each FFT is applied to the data set at a different point in time captured by the data set to generate spectrogram data. Then, from the spectrogram data, an image may be generated representing the spectrogram data.

Accordingly, a spectrogram image is obtained containing a spectrum of frequencies present in the data set over the time spanned by the data set.

In step 130, the spectrogram image is provided to a type classification machine learning algorithm. The type classification machine learning algorithm may be any Al-based (e.g., neural network) algorithm configured to process spectrogram images to identify and classify features, so as to determine a type.

The spectrogram image is processed with the type classification machine learning algorithm. Accordingly, a predicted type of the replaceable component is determined.

Predicting the type of the replaceable component of the personal care device may comprise classifying, with the type classification machine learning algorithm, the type of the replaceable component of the personal care device into one of plurality of type classes (each corresponding to a certain make, model, version, brand and/or variety of the replaceable component). In essence, the type classification machine learning algorithm may identify features of the spectrogram that indicate the replaceable component to belonging to one of said classes, and assign the predicted type based on the identified features.

The type classification machine learning algorithm may be a spectral image classifier, such as a CNN classifier or an FCNN classifier. In the case that it is a CNN classifier, it may comprise 3-6 convolutional layers.

In some cases, the method may further comprise generating the type classification machine learning algorithm. This may further comprise generating, for a plurality of replaceable components of the personal care device each having a respective known type, a test data set describing an operating parameter of the personal care device during vibration of the personal care device. Each test data set may be generated during use by a user, or may be synthetically generated. The synthetically generated data may include data generated using a robotic testing setup, or computer simulated vibration response data.

The type classification machine learning algorithm is then trained using a training algorithm configured to receive an array of training inputs and respective known outputs, wherein a training input comprises one of the plurality of test data sets, and wherein a respective known output comprises the known type.

Moving on, Fig. 2 depicts a personal care device 200 according to an example embodiment. In this case, an electric toothbrush is shown. However, it is worth noting that this is for illustrative purposes only, and the personal care device 200 may be any type of device used for personal care (not just oral care).

The personal care device 200 comprises a replaceable component 210, which in this case is a toothbrush head. The personal care device 200 further comprises an actuator 220 configured to cause the personal care device 200 to vibrate while in use, a sensor 230, and a processor 240. The sensor 230 obtains a data set describing an operating parameter of the personal care device 200 during vibration of the personal care device 200. The data set may be any data set described above, and acquired using any method described above. While here it is depicted that the sensor 230 is provided as part of the personal care device 200, the sensor 230 may instead be provided as part of an external device. For example, the sensor 230 may be a microphone provided on a charger base/dock. The sensor 230 may also be a load sensor that measures the force exerted by a surface of the personal care device on a surface of the user. Alternatively, or additionally, the sensor 230 may fetch a previously stored data set from memory, and provide said dataset to the processor 240.

The processor 240 is configured to generate a spectrogram image based on at least part of the data set, process, with a type classification machine learning algorithm, the spectrogram image to predict a type of the replaceable component 210 of the personal care device 200. The processor 240 may be configured to perform any of the method steps described above in relation to Fig. 1.

Similarly to the sensor, while the processor 240 is depicted as forming part of the personal care device 200, it may also be provided externally to the personal care device 200. For example, the processor 240 may be connected wirelessly to the sensor 230 in order to receive the data set. The processor 240 may be provided in the form of a smartphone, or the processor 240 may be implemented on the cloud or other external processing service.

The resulting predicted type of the replaceable component 210 may be used to provide an output indicative of the type of the replaceable component 210. Such an output may be provided to a user. Alternatively, or additionally, the predicted type may be used to alter operating parameters of the personal care device 200 to ensure potential faults are reduced and effective operation of the personal care device 200. Other uses of the predicted type would be apparent to the skilled person.

Accordingly, the invention may facilitate an improved user experience, and improved delivery of a personal care function.

Figure 3 illustrates an example of a computer 300 within which one or more parts of an embodiment may be employed. Various operations discussed above may utilize the capabilities of the computer 300. For example, one or more parts of a system for controlling a handheld device may be incorporated in any element, module, application, and/or component discussed herein. In this regard, it is to be understood that system functional blocks can run on a single computer or may be distributed over several computers and locations (e.g. connected via internet), such as a cloud-based computing infrastructure. The computer 300 includes, but is not limited to, PCs, workstations, laptops, PDAs, palm devices, servers, storages, microcontroller units, Al accelerators, integrated processors and the like. Generally, in terms of hardware architecture, the computer 300 may include one or more processors 310, memory 320, and one or more I/O devices 330 that are communicatively coupled via a local interface (not shown). The local interface can be, for example but not limited to, one or more buses or other wired or wireless connections, as is known in the art. The local interface may have additional elements, such as controllers, buffers (caches), drivers, repeaters, and receivers, to enable communications. Further, the local interface may include address, control, and/or data connections to enable appropriate communications among the aforementioned components.

The processor 310 is a hardware device for executing software that can be stored in the memory 320. The processor 310 can be virtually any custom made or commercially available processor, a central processing unit (CPU), a digital signal processor (DSP), a tensor processing unit (TSP) specifically designed for neural processing, a dedicated Al accelerator/processing init or an auxiliary processor among several processors associated with the computer 300, and the processor 310 may be a semiconductor based microprocessor (in the form of a microchip) or a microprocessor.

The memory 320 can include any one or combination of volatile memory elements (e.g., random access memory (RAM), such as dynamic random access memory (DRAM), static random access memory (SRAM), etc.) and non-volatile memory elements (e.g., ROM, erasable programmable read only memory (EPROM), electronically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), tape, compact disc read only memory (CD-ROM), disk, diskette, cartridge, cassette or the like, etc.). Moreover, the memory 320 may incorporate electronic, magnetic, optical, and/or other types of storage media. Note that the memory 320 can have a distributed architecture, where various components are situated remote from one another, but can be accessed by the processor 310.

The software in the memory 320 may include one or more separate programs, each of which comprises an ordered listing of executable instructions for implementing logical functions. The software in the memory 320 includes a suitable operating system (O/S) 340, compiler 360, source code 350, and one or more applications 370 in accordance with exemplary embodiments. As illustrated, the application 370 comprises numerous functional components for implementing the features and operations of the exemplary embodiments. The application 370 of the computer 300 may represent various applications, computational units, logic, functional units, processes, operations, virtual entities, and/or modules in accordance with exemplary embodiments, but the application 370 is not meant to be a limitation.

The operating system 340 controls the execution of other computer programs, and provides scheduling, input-output control, file and data management, memory management, and communication control and related services. It is contemplated by the inventors that the application 370 for implementing exemplary embodiments may be applicable on all commercially available operating systems.

Application 370 may be a source program, executable program (object code), script, or any other entity comprising a set of instructions to be performed. When a source program, then the program is usually translated via a compiler (such as the compiler 360), assembler, interpreter, or the like, which may or may not be included within the memory 320, so as to operate properly in connection with the O/S 340. Furthermore, the application 370 can be written as an object oriented programming language, which has classes of data and methods, or a procedure programming language, which has routines, subroutines, and/or functions, for example but not limited to, C, C++, C#, Pascal, BASIC, API calls, HTML, XHTML, XML, ASP scripts, JavaScript, FORTRAN, COBOL, Perl, Java, ADA, .NET, and the like.

The I/O devices 330 may include input devices such as, for example but not limited to, a mouse, keyboard, scanner, microphone, camera, etc. Furthermore, the I/O devices 330 may also include output devices, for example but not limited to a printer, display, etc. Finally, the I/O devices 330 may further include devices that communicate both inputs and outputs, for instance but not limited to, a NIC or modulator/demodulator (for accessing remote devices, other files, devices, systems, or a network), a radio frequency (RF) or other transceiver, a telephonic interface, a bridge, a router, etc. The I/O devices 330 also include components for communicating over various networks, such as the Internet or intranet.

If the computer 300 is a PC, workstation, intelligent device or the like, the software in the memory 320 may further include a basic input output system (BIOS) (omitted for simplicity). The BIOS is a set of essential software routines that initialize and test hardware at startup, start the O/S 340, and support the transfer of data among the hardware devices. The BIOS is stored in some type of read-only-memory, such as ROM, PROM, EPROM, EEPROM or the like, so that the BIOS can be executed when the computer 300 is activated.

When the computer 300 is in operation, the processor 310 is configured to execute software stored within the memory 320, to communicate data to and from the memory 320, and to generally control operations of the computer 300 pursuant to the software. The application 370 and the O/S 340 are read, in whole or in part, by the processor 310, perhaps buffered within the processor 310, and then executed.

When the application 370 is implemented in software it should be noted that the application 370 can be stored on virtually any computer readable medium for use by or in connection with any computer related system or method. In the context of this document, a computer readable medium may be an electronic, magnetic, optical, or other physical device or means that can contain or store a computer program for use by or in connection with a computer related system or method.

The application 370 can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer- based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. In the context of this document, a "computer-readable medium" can be any means that can store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer readable medium can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium.

The proposed control method(s) of Fig. 1, and the system(s) of Fig. 2, may be implemented in hardware or software, or a mixture of both (for example, as firmware running on a hardware device). To the extent that an embodiment is implemented partly or wholly in software, the functional steps illustrated in the process flowcharts may be performed by suitably programmed physical computing devices, such as one or more central processing units (CPUs) or graphics processing units (GPUs). Each process - and its individual component steps as illustrated in the flowcharts - may be performed by the same or different computing devices. According to embodiments, a computer-readable storage medium stores a computer program comprising computer program code configured to cause one or more physical computing devices to carry out a control method as described above when the program is run on the one or more physical computing devices.

Storage media may include volatile and non-volatile computer memory such as RAM, PROM, EPROM, and EEPROM, optical discs (like CD, DVD, BD), magnetic storage media (like hard discs and tapes). Various storage media may be fixed within a computing device or may be transportable, such that the one or more programs stored thereon can be loaded into a processor. To the extent that an embodiment is implemented partly or wholly in hardware, some of the blocks shown in the block diagrams of Fig 2. may be separate physical components, or logical subdivisions of single physical components, or may be all implemented in an integrated manner in one physical component. The functions of one block shown in the drawings may be divided between multiple components in an implementation, or the functions of multiple blocks shown in the drawings may be combined in single components in an implementation. Hardware components suitable for use in embodiments of the present invention include, but are not limited to, conventional microprocessors, application specific integrated circuits (ASICs), and field-programmable gate arrays (FPGAs). One or more blocks may be implemented as a combination of dedicated hardware to perform some functions and one or more programmed microprocessors and associated circuitry to perform other functions.

Variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit may fulfil the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. If a computer program is discussed above, it may be stored/distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. If the term "adapted to" is used in the claims or description, it is noted the term "adapted to" is intended to be equivalent to the term "configured to". Any reference signs in the claims should not be construed as limiting the scope.

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.

Claims

CLAIMS:

1. A method for determining a type of a replaceable component of a personal care device, the personal care device comprising an actuator that causes the personal care device to vibrate while in use, the method comprising: obtaining (110) a data set describing an operating parameter of the personal care device during vibration of the personal care device; generating (120) a spectrogram image based on at least part of the data set; and processing, with a type classification machine learning algorithm, the spectrogram image to predict (130) a replaceable component type.

2. The method of claim 1, wherein the type classification machine learning algorithm is a spectral image classifier, and optionally wherein the type classification machine learning algorithm is a CNN classifier or an FCNN classifier.

3. The method of claim 2, wherein the type classification machine learning algorithm is a CNN classifier comprising 3-6 convolutional layers.

4. The method of any preceding claim, wherein the data set comprises accelerometer data describing an acceleration in at least one axis of the personal care device.

5. The method of claim 4, wherein the accelerometer data describes acceleration in 3 axis, and wherein the method further comprises: calculating norm vector data describing an absolute vector length of acceleration based on the accelerometer data; and serializing the accelerometer data of each of the 3 axis and the norm vector data into a single data segment.

6. The method of any preceding claim, wherein the data set comprises sound data describing a sound produced by the personal care device.

7. The method of any preceding claim, wherein the data set comprises current data describing a motor current of the actuator.

8. The method of any preceding claim, wherein the obtained data set comprises values describing the operating parameter of the personal care device spanning up to one second, and wherein the obtained data set comprises values describing the operating parameter of the personal care device sampled at a rate of at least 800Hz, and optionally at a rate of at least

I.6kHz.

9. The method of any preceding claim, further comprising normalizing at least part of the data set.

10. The method of any preceding claim, wherein the data set describing the operating parameter is obtained responsive to the personal care device being in a static state.

I I. The method of claim 10, wherein the data set describing the operating parameter is obtained (110) during a single operational mode of the actuator, or wherein the data set describing the operating parameter is obtained (110) while changing an operational mode of the actuator to vary an amplitude and/or duty cycle of vibration of actuation.

12. The method of any preceding claim, wherein determining (130) the predicted replaceable component type comprises: classifying, with the type classification machine learning algorithm, the replaceable component into one of a plurality of type classes based on the spectrogram image, and optionally wherein each of the plurality of type classes reflect a different model group, make group, version group, variety group, and/or brand group of the replaceable component.

13. The method of any preceding claim, further comprising generating the type classification machine learning algorithm, comprising: generating, for a plurality of replaceable components of the personal care device each having a respective known type, a test data set describing an operating parameter of the personal care device during vibration of the personal care device; training the type classification machine learning algorithm using a training algorithm configured to receive an array of training inputs and respective known outputs, wherein a training input comprises one of the test data sets, and wherein a respective known output comprises the known type, and optionally wherein each test data set comprises data generated during user operation of the personal care device, and/or synthetic generated data.

14. A computer program comprising computer program code means adapted, when said computer program is run on a computer, to implement the method of any of claims 1-13.

15. A personal care device, comprising: a replaceable component (210); an actuator (220) configured to cause the personal care device to vibrate while in use; a sensor (230) for obtaining a data set describing an operating parameter of the personal care device during vibration of the personal care device; a processor (240) configured to: generate a spectrogram image based on at least part of the data set; and process, with a type classification machine learning algorithm, the spectrogram image to predict a type of the replaceable component.