GB2637165A

GB2637165A - Heater apparatus and methods

Info

Publication number: GB2637165A
Application number: GB2400435.0A
Authority: GB
Inventors: Cutter Ry
Original assignee: Jemella Ltd
Current assignee: Jemella Ltd
Priority date: 2024-01-12
Filing date: 2024-01-12
Publication date: 2025-07-16
Also published as: GB202400435D0

Abstract

A hair drying or styling apparatus 1 comprising a heater 6 with a plurality of independently controllable heating zones (642, fig.6b), a memory (30, fig.2) that stores a trained classifier that classifies the hair among a plurality of reference hair classes, and a controller (28, fig.2) that: determines a power demand based on a temperature of the heating zones or the power needed to maintain each at a target temperature; determines a loaded hair tress size based on a drop of the temperature of the heating zones or an increase in the power output needed to maintain them at the target temperature, and an amount of hair engaged; uses the stored trained classifier to determine the type of hair by classifying based on the determined power demand and hair tress size; and controls the temperature or the power output of the heating zones which are engaged with the hair, based on the hair type. The apparatus may receive user input from a mobile phone or smart watch. The controller may also determine grip pressure, motion of the apparatus, and hair diagnostic information. A method of generating the classifier using a machine learning algorithm is also disclosed.

Description

HEATER APPARATUS AND METHODS

Field of the Invention

The present invention relates to heating apparatus and methods. The heaters can be used for styling hair. Such styling of the hair may be performed by a user in respect of their own hair, for example, or by a hair stylist. The invention has particular, but not exclusive, relevance to a styling device comprising one or more low thermal mass heaters.

Background to the Invention

Heated hair styling tools use heat to increase the temperature of hair to a desired styling temperature. For example, a hair straightener having a heated plate applies heat directly via conduction to heat the hair, which may be either wet or dry. to achieve the desired temperature for styling. The hair may be heated to a temperature that is particularly suitable for styling hair (for example, to or beyond a hair glass transition phase temperature). At lower temperatures, the user may have to make many passes with the hair straightener over the hair to achieve a desired styling effect, whereas at higher temperatures, there is a risk of causing permanent damage to the hair.

Similarly, a heated brush or hair dryer can also be used to style hair by heating air which in turn heats the hair to a temperature suitable for styling. The hair is typically styled from wet, for example after the user has washed their hair, although the hair could also be styled from dry.

Existing hair styling appliances typically use heaters that provide a certain amount of thermal energy to the hair styling appliance.

The amount of thermal energy provided to the hair styling appliance correspond to one mode of operation so that the hair of most users of the hair styling appliance may be heated to a temperature that is particularly suitable for styling hair.

However, there is a need for improvements to such existing hair styling appliances, as users with hair types which do not correspond to that of most users feel a more frustrating experience and/or have their hair damaged when styling.

The present invention aims to address or at least partially ameliorate one or more of the above problems.

Summary of the Invention

Aspects and embodiments of the invention are set out in the appended claims. These and other aspects of the invention, and aspects and embodiments which are useful in understanding the invention set out in the appended claims, are also described in the disclosure herein.

Any feature in one aspect of the disclosure may be applied to other aspects of the disclosure, in any appropriate combination. In particular, method aspects may be applied to device and computer program aspects, and vice versa.

Furthermore, features implemented in hardware may generally be implemented in software, and vice versa. Any reference to software and hardware features herein should be construed accordingly.

In one aspect the invention provides apparatus for drying or styling hair, the apparatus comprising: a heater comprising a hair contacting surface for heating hair of a user that contacts the hair contacting surface by conduction, the heater comprising a plurality of independently controllable heater electrodes that define a plurality of independently controllable heating zones of the hair contacting surface; a memory that stores a trained classifier configured to classify the hair of the user of the apparatus into a class among a plurality of reference hair classes; and a controller coupled to the memory and configured to control the apparatus to: determine a power demand of the apparatus, based on a temperature of the plurality of independently controllable heating zones or based on the power output needed to maintain each heating zone of the plurality of heating zones at a target temperature; determine a hair tress size which has been loaded by the user into the apparatus, by determining, based on a drop of the temperature of one or more heating zones of the plurality of independently controllable heating zones or based on an increase in the power output needed to maintain the one or more heating zones at the target temperature, whether hair is engaged with the one or more heating zones, and, if it is, by determining an amount of hair which is engaged with the one or more heating zones, the hair tress size being determined based on the amount of hair engaged with the one or more heating zones across all of the one or more heating zones; use the stored trained classifier to determine the type of hair of the user by classifying the hair into a reference hair class, based on the determined power demand and the determined hair tress size; and cause the apparatus to control the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones which are engaged with the hair, based on the determined hair type.

The apparatus may further comprise communication circuitry such that the apparatus is configured to receive user input information from an external processing device. The user input information may be input on a user interface associated with the user and configured to enable the user to input the input information. The external processing device may be part of a mobile phone or a smart watch and may be configured to run an application, remote from the apparatus.

The controller may be further configured to control the apparatus to use the stored trained classifier such that the plurality of reference hair classes into which the hair of the user is to be classified is narrowed down to a subset of the plurality of reference hair classes, based on the received user input information. The user input information may comprise at least one of: an indication by the user of the hair type they believe they have; a result from the user answering a questionnaire about their hair and/or the environment.

The controller may be configured to: determine a grip pressure based on data associated with a pressure sensor configured to measure a pressure between movable arms of the apparatus; and cause the apparatus to use the stored trained classifier to determine the type of hair of the user, further based on the grip pressure.

The controller may be configured to: determine a motion of the apparatus when the user is styling or drying their hair, based on data associated with one or more motion sensors of the apparatus; and cause the apparatus to use the stored trained classifier to determine the type of hair of the user, further based on the motion of the apparatus.

The one or more motion sensors of the apparatus may comprise any one or more of an accelerometer, a gyrometer, a magnetometer, an inclination sensor. The motion of the hair styling or drying appliance may comprises: a rotation of the apparatus when styling or drying the hair; and/or a speed of movement of the apparatus when styling or drying the hair.

The controller may be configured to: process further sensor data to determine diagnostic information about the hair being styled or dried; and cause the apparatus to use the stored trained classifier to determine the type of hair of the user, further based on the determined diagnostic information. The diagnostic information may comprise at least one of: a level of moisture in the hair, a humidity ambient to the hair being styled or dried, a temperature ambient to the hair being styled or dried, a geographic location of the hair being styled or dried.

The controller may be configured to control the temperature or the power output of the one or more heating zones by capping the increase in the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones based on the determined type of hair, to avoid the hair to be burnt.

The plurality of reference hair classes may comprise nine reference classes, referred to as (1 a, 1 b, 1 c, 2a, 2b, 2c, 3a, 3b, 3c), the class la corresponding to fine, straight blonde hair and the class 3c corresponding to thick, curly dark hair. The controller may be configured to set the target temperature or the power output of the one or more heating zones of the plurality of controllable heating zones which are engaged with the hair, based on the determined type of hair with reference to the nine reference classes. The controller may be configured set the target temperature or the power output of the one or more heating zones of the plurality of controllable heating zones so as to enable a better drying or styling of the hair which is loaded in the apparatus.

The trained classifier may comprise a machine learning algorithm comprising a convolutional neural network.

The apparatus may be a hair straightener, a hair dryer, a hot paddle brush, a hot round brush, a heater roller, or a hair curler.

In another aspect the invention provides a method for generating a classifier configured to classify hair of a user of apparatus for drying or styling hair, into a class among a plurality of reference hair classes, the method comprising: obtaining a plurality of annotated training streams of data; and training the classifier by applying a machine learning algorithm to the obtained training streams of data. The annotation indicates the class of hair associated with each training stream of data, and the plurality of annotated training streams of data comprises, for each class of hair among the plurality of reference hair classes: a training stream of data corresponding to a power demand of the apparatus when styling or drying hair belonging to the class of hair; and a training stream of data corresponding to a hair tress size which is typically loaded in the apparatus when styling or drying hair belonging to the class of hair.

The classifier may be trained to minimize a classification loss between the annotated class of hair associated with each training stream of data and a classification of the hair determined by the classifier. The classification loss may comprise a similarity metric of Lp-norm, p being an integer greater or equal to 1, such as an average absolute deviation or a least mean square distance.

The plurality of training streams of data may comprises, for each class of hair among the plurality of reference hair classes: a training stream of data corresponding to sensor data indicative of a grip pressure between movable arms of the apparatus when styling or drying hair belonging to the class of hair; and/or one or more training streams of data corresponding to sensor data indicative of a motion of the apparatus when the user is styling or drying hair belonging to the class of hair. The motion of the hair styling or drying appliance may comprise: a rotation of the apparatus when styling or drying the hair; and/or a speed of movement of the apparatus when styling or drying the hair; and/or a training stream of data corresponding to sensor data indicative of diagnostic information about the hair being styled or dried.

The classifier may be trained to classify the hair of the user into a subset of the plurality of reference hair classes, based on user input information. The user input information may comprise at least one of: an indication of a suspected use's hair type; a result from a questionnaire about their hair and/or the environment.

The machine learning algorithm may comprise a convolutional neural network.

In another aspect the invention provides a method of producing apparatus for drying or styling hair, wherein the apparatus comprises a heater comprising a hair contacting surface for heating hair of a user that contacts the hair contacting surface by conduction, the heater comprising a plurality of independently controllable heater electrodes that define a plurality of independently controllable heating zones of the hair contacting surface, a memory and a controller coupled to the memory, the method comprising: obtaining a classifier; and storing the obtained classifier in the memory of the appliance. The controller is configured to control the apparatus to: use the stored trained classifier to determine the type of hair of the user by classifying the hair into a reference hair class; and cause the apparatus to control the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones which are engaged with the hair, based on the determined hair type.

The storing may comprise transmitting the generated classifier to the apparatus via a network, the apparatus receiving and storing the classifier. The classifier may be generated, stored and/or transmitted in the form of one or more of: a data representation of the classifier; executable code for applying the classifier.

In another aspect the invention provides a computer program or a computer program duct.

Brief Description of the Drawings

Embodiments of the invention will now be described, by way of example only, and with reference to the drawings in which: Figure la shows an overview of an exemplary hair styling device; Figure 1 b shows a hair styling device in use; Figure 2 is a block diagram illustrating the main electronic components of the hair styling device shown in Figure 1; Figure 3a is an exploded view of a heater forming part of the hair styling device shown in Figure 1; Figure 3b is an assembled partially transparent view of the heater shown in Figure 3a; Figure 4a schematically illustrates the heating zones on the heating surface of the heater shown in Figure 3; Figure 4b schematically illustrates an alternative arrangement of heating zones; Figure 5 schematically illustrates a further alternative arrangement of heating zones that are of different sizes and shapes; Figure 6a illustrates the way in which the heating zones may be formed on a tubular substrate for use in a curling tong or the like; Figure 6b illustrates the way in which the heating zones may be arranged on a curved substrate which may be used on a heated brush; Figure 7 illustrates a tress of hair that partly overlaps with zones Z2 and Z4 of a heater; Figure 8 shows a flow chart illustrating an example method for generating a classifier according to the disclosure; Figure 9 shows an example architecture of a classifier according to the disclosure; Figure 10 shows another example architecture of a classifier according to the

disclosure;

Figure 11 shows a flow chart illustrating an example method for producing a hair styling appliance according to the disclosure; and Figure 12 shows a flow chart illustrating an example method for operating a hair styling appliance according to the disclosure.

Overview of Hair Styling Device Figure 1a illustrates a hand-held (portable) hair styler 1 (or hair styling appliance in the present disclosure). The hair styler 1 includes a first movable arm 4a and a second movable arm 4b, which are coupled at proximal ends thereof to a shoulder 2. The first arm 4a bears a first heater 6a at its distal end, and the second arm 4b bears a second heater 6b at its distal end. The first and second heaters 6a, 6b oppose one another and are brought together as the first and second arms 4a, 4b are moved from an open configuration to a closed configuration. As shown in Figure 1 b, during use, a tress of hair 40 is sandwiched between the two arms 4 so that the users hair is in contact with, and therefore heated by, outer heating surfaces of the heaters 6a, 6b. Therefore, as the user pulls the hair styler 1 along the tress of hair 40, the tress of hair 40 is heated by conductive heating to a suitable temperature to facilitate styling.

A user interface 11 is provided to allow the user to input information about them to the device and/or for the device to output information to the user. The user interface 11 may have a dial, button or touch display for allowing the user to input information to the device 1 and the user interface 11 may have an indicator light, display, sound generator or haptic feedback generator for outputting information to the user. In this embodiment, the user interface 11 also comprises a control button or switch 14 to enable the user to turn the device 1 on or off; and an indicator light 15 to show whether the power is on.

A printed circuit board assembly (not shown) may be provided at any suitable location within the housing of the device 1 and carries the control circuitry for controlling the operation of the device 1 and for controlling the interaction with the user via the user interface 11. In this example, electrical power is provided to the device 1 by means of a power supply located at an end of the device, via a power supply cord 3. The power supply may be an AC mains power supply. However, in an alternative embodiment the power supply may comprise one or more DC batteries or cells (which may be rechargeable, e.g., from the mains or a DC supply via a charging lead), thereby enabling the device 1 to be a cordless product.

In use, the device 1 is turned on, energising the heaters 6 to cause them to heat up. The user then opens the first and second arms 4a, 4b and, normally starting from the roots of the hair (i.e. near the scalp), a length or tress of hair 40 (which may be clumped) is introduced between the arms 4a, 4b, transversely across the heaters 6a, 6b. The user then closes the arms 4a, 4b so that the length of hair 40 is held between the first and second arms 4a, 4b and then the user pulls the hair through the closed arms (as illustrated in Figure 1 b). The outer (hair contacting) surface of the heaters 6 is flat in this embodiment and so the hair styler 1 can be used to straighten the user's hair. The hair styling device 1 shown in Figure 1 can also be used to curl the hair by turning the device 1 through approximately 180 degrees or more after clamping the hair between the arms 4a, 4b and before moving the device 1 along the tress of hair 40.

Hair has a relatively high thermal mass and when in contact with the heating surface of the heater 6 the hair absorbs a significant amount of the heat energy. The heaters 6 must quickly supply the lost heat energy back to the heating surface otherwise the temperature of the heating surface will drop and potentially impact on the quality of the thermal styling.

If the temperature of the heaters 6 fall below the glass transition temperature of the hair, the hair will not retain the styled shape.

However, if the hair is heated to a temperature that is too high, the hair can undergo significant damage.

Furthermore, different hair types require different amount of heat energy for hair styling (because of different thickness, quality, condition, thermal mass of hair). Typically, fine straight hair requires less heat than thick curly hair for styling. Depending on the type of hair, the glass transition temperature may be a temperature in the range of approximately 160°C-200°C.

As such, the device 1 must be able to control the temperature so that the heating surface of the heaters 6 remains within a particular temperature range. Furthermore, it must maintain the temperature range both when hair is frequently and quickly loaded and unloaded onto the heating surface, and when hair is held on the heating surface for a prolonged period of time.

Control Circuitry Figure 2 is a simplified block diagram of control circuitry 15 that controls the operation of the hair styler device 1 shown in Figure 1. As shown, the control circuitry 15 comprises a power supply 21 that, in this embodiment, derives power from a battery power source. A mains power supply input may be provided to charge the battery via an AC to DC converter (not shown), which may be external or internal to the device 1.

Alternatively, the power supply 21 may derive power from an AC mains supply input.

In this example, power is provided to the heaters 6 for heating the user's hair. The power supplied to the heaters 6 is controlled by a controller 28 having a microprocessor 29.

The power supplied to the heaters 6 is controlled by drive circuitry 23 (which may include one or more power semiconductor switching devices (triacs)) which controls the application of an AC mains voltage, or a DC voltage derived from the AC mains or from a battery, to the heaters 6 in accordance with instructions from the microprocessor 29.

The microprocessor 29 is coupled to a memory 30 (which is typically a non-volatile memory) that stores processor control code for implementing one or more control methods that control the heating of the heaters 6 in accordance with a desired operating temperature of the heaters 6 and sensed temperatures of the heaters obtained from temperature measurement circuitry 25.

The microprocessor 29 allows for complex control of the heaters 6. For example, the controller 28 may be configured to adjust the power delivered to heaters by using an on/off triac based upon the output of the temperature measurement circuitry 25.

The memory 30 may store a number of transfer functions such as: simple on-off control means or bang-bang control means; proportional-integral-derivative (PI D) control means; fuzzy logic; feed back control means; feed forward control means.

The controller 28 comprises means to measure the input voltage or alternatively to detect the speed at which the heaters 6 heat up, so as to detect the type of input voltage. A high input voltage would lead to a faster heat up of the heaters 6 and hence a control loop can read appropriately. The input voltage and/or speed of heat up can also be used to detect a failure.

The controller 28 may comprise means to detect the use of the hair styling appliance and control the power supply to the heaters accordingly. This feature helps to reduce power consumption and improve safety. For example, the controller 28 may comprise means to reduce the temperature of the heaters when they are not active and then rapidly heat them up when they are about to be used. The controller 28 may allow a heater to power down to a standby temperature if a user momentarily places the hair styling appliance on a table. The controller 28 may then power up the heater to an operating temperature when the hair styling appliance is picked up to be used. If the controller 28 detects that the hair styling appliance has not been used fora longer period of time, then the control means may shut down the hair styling appliance. This enables the hair styling appliance to meet the mandatory requirement of the safety standard that the appliance must turn off after 30 minutes whether it is being used or not.

Detection of use may be achieved by detecting the opening and closing of the first movable arm and a second movable arm of the hair styling appliance, or through the use of one or more motion detection devices to detect the motion of the hair styling appliance or the used of a capacitive touch system.

The one or more motion detection devices are shown in Figure 2 and referred to with numerical reference 31. The one or more motion detection devices 31 may comprise any one of an accelerometer, a gyrometer, a magnetometer, an inclination sensor.

The hair styler 1 includes a pressure sensor 32 to measure a pressure between the first movable arm 4a and the second movable arm 4b, e.g., when styling the hair.

The hair styler 1 includes a hair diagnostic sensor 33 configured to output data indicative of diagnostic information about the hair being styled or dried. The diagnostic information comprises at least one of: a level of moisture in the hair, a humidity ambient to the hair being styled or dried, a temperature ambient to the hair being styled or dried, a geographic location of the hair being styled or dried.

The temperature measurement circuitry 25 may be temperature sensors such as thermistors or they may use circuitry that senses the resistance of heater electrodes that are used to heat the heaters 6, which resistance depends on the temperature of the heater electrode.

Figure 2 also shows that the user interface 11 is coupled to the microprocessor 29, for example to provide one or more user controls, input and/or output indications such as a visual indication or an audible alert.

Finally, the control circuitry includes communications circuitry 27 to allow the device to communicate with a remote sensor, a remote server, or a remote application (e.g., on a mobile telephone). The communications circuitry 27 may use, for example, Bluetooth, Wi-Fi and/or 3GPP communication protocols to communicate with the remote device.

The hair styling appliance described in the present disclosure may further comprise means for providing a polyphonic sound. The means may provide a particular sound brand or jingle when switching on and/or off. The means may provide a sound to indicate particular events, such as reaching a desired operating temperature and/or sleep mode.

The hair styling appliance described in the present disclosure may comprise lighting means. The lighting means may provide a pleasing aesthetic appearance as well as indicate temperature or other events. The lighting means may comprise an electroluminescent backlight as it enables wide angle, wide area viewing. Alternatively or additionally, the lighting means may comprise an LED lighting with a suitable light-pipe and/or optical diffuser.

Heaters The heaters 6a, 6b are low thermal mass heaters and can therefore heat up and cool down quickly. Figures 3a and 3b show an exemplary embodiment of such heaters 6a, 6b, which comprise a stack of thin layers. Referring in particular to Figure 3a, the heaters 6a, 6b include an upper dielectric (electrically insulating) layer 62, an electrode layer 63 that has a plurality of separate heater electrodes 64, and a lower dielectric layer 66 which electrically insulates the heater electrodes 64 from other components mounted behind the heater 6a, 6b. The three layers 62, 63 and 66 are bonded together either through an adhesive layer (pressure set or thermoset) or through diffusion bonding of the contacting materials (e.g. melting them together) and define a heater 6 that is very thin (the three layers have an overall thickness of between 30pm to 1000pm in the case of low voltage operation (less than about 40 Volts) and 0.8mm to 2.0mm in the case of AC operation) and with very low thermal mass. The upper surface of the layer 62 provides the hair contacting surface of the heater 6, although a non-stick coating may be applied to the upper surface of the layer 62 to facilitate the passage of the user's hair over the heating surface. The bonded layers 62, 63 and 66 define a flexible heater 6 and rigidity of the heater is provided in the illustrated embodiment by mounting the heater layers 62, 63 and 66 into a rigid support 68 which forms a base. These layers may be mounted onto the rigid support after the layers themselves have been bonded together or they may be bonded one at a time (or multiple at a time) onto the rigid support 68. If a flexible heater is desired, then there is no need for the rigid support 68 or if a support is used, this may be a non-rigid support.

In the illustrated embodiment, there are ten heater electrodes 64 that each snake across and back across the width of the heater 6, folding twice such that they each cross the width three times. The ends of each of the heater electrodes 64 are electrically connected through the lower dielectric layer 66 to electrical connections within the rigid support 68, which connect to an electrical connector 70. Drive circuitry 23 that is mounted within one of the arms 4 connects to the heater electrodes 64 via the electrical connector 70 and applies electrical power to the individual heater electrodes 64 to control the heat generated by each heater electrode 64. The electrical connector 70 extends from a surface of the rigid support 68 facing away from the surface layer 62 (shown in Figures 3a and 3b as extending directly away from the upper layer 62, but it could also be provided as extending in a perpendicular direction).

Each of the heater electrodes 64 thus creates an individual heating zone 642 on the hair contacting surface of the heater 6, which spans the width (which we shall refer to as the x-direction) of the heater 6 and the heater electrodes 64 are arranged sequentially one after the other along the length (the y-direction) of the heater 6.

Figures 4a and 4b show schematic views of different arrangements of such heating zones 642. Figure 4a shows an arrangement corresponding to that of Figures 3a and 3b, in which the heating zones 642-1 to 642-10 are arranged along the y-direction only.

Figure 4b shows an alternative arrangement, in which heating zones 642-1 to 642-16 are arranged in both the x-and y-directions. Such an arrangement of heating zones 642 can be provided by arranging two sets of heater electrodes 64 like those shown in Figure 3a side by side in the width, x-direction. The heaters 6 may be separated in this way into any number of heating zones 642 and may comprise any number of heating zones along the x-and y-directions. In particular, whilst Figure 4b shows two zones along the x-direction, a greater number of zones in the x-direction could also be provided. The heating zones 642 of the heaters 6a, 6b can be operated (heated) independently, which can help to reduce hot/cold spots when using very low thermal mass heaters 6 such as those shown in Figure 3.

The heating zones illustrated in Figure 4 are all the same size. Of course, different sized heating zones 642 may be provided, as illustrated in Figure 5, which shows a heater 6 having seven different sized heating zones (labelled Z1 to Z7). The way in which the heater electrodes 64 would be arranged to define these different sized zones would be understood by the skilled reader and will not be described in detail here.

The heating zones 642 described above form part of a heater having a flat hair contacting surface. The heater is not limited to flat hair contacting surfaces and can be configured for use a tubular form (as illustrated in Figure 6a) for example for use in a hair curler device or in a curved form (as illustrated in Figure 6b) for example for use in a heated hair brush. The heater surface may have a corrugated or ribbed shape to provide a hair crimping device.

The temperature of each heating zone 642 is independently controllable. Each heating zone 642 can be set to a target temperature. The target temperature of each heating zone 642 may be different.

A separate temperature sensor may be provided for sensing the temperature of each heating zone 642 which is fed back to the microprocessor 29 to allow the microprocessor 29 to control the delivery of power to the heater electrode 64 of the corresponding heating zone 642.

Alternatively, if the heater electrodes 64 are formed of a material having a Positive Temperature Coefficient(PTC) or a Negative Temperature Coefficient (NTC) (such that its resistance varies with its temperature), then the temperature of each heating zone 642 can be determined by determining the resistance of the corresponding heater electrode 64.

The microcontroller 28 controls the heating in order to reduce the difference between the actual temperature of the heating zone 642 and the target temperature for that heating zone 642.

Heating Zone Sizing When the user loads a tress of hair 40 onto the heaters 6, some parts of the heater will be loaded with hair whilst other parts will not be loaded with hair.

Upon loading with hair, and using the already described temperature measurement circuitry 25, more power is supplied to the heater 6 to ensure that all regions on the hair contacting surface can be retained within and/or recovered back to the desired operating temperature limits.

Figure 7 shows a tress of hair 40 overlying heating zones Z2, Z3 and Z4, with heating zone Z3 being fully loaded with hair whilst heating zones Z2 and Z4 being only partially loaded with hair.

Input streams of data Tress Load The Figure 7 extent of the heating zones gives the microcontroller 28 an indication of the size of the tress of hair 40 introduced between the arms of the styling appliance.

The microcontroller 28 is configured to determine a hair tress size which has been loaded by the user into the apparatus.

In order to determine the hair tress size, the microcontroller 28 determines, based on a drop of the temperature of one or more heating zones of the plurality of independently controllable heating zones or based on an increase in the power output needed to maintain the one or more heating zones at the target temperature, whether hair is engaged with the one or more heating zones. If it is determined that hair is engaged with the one or more heating zones, the microcontroller 28 determines an amount of hair which is engaged with the one or more heating zones. The microcontroller 28 then determines the hair tress size based on the amount of hair engaged with the one or more heating zones across all of the one or more heating zones.

The size of the tress of hair 40 in turn gives an indication of the amount of hair introduced between the arms of the styling appliance.

Additionally or alternatively, the styler may comprise sensing means to determine the amount of hair introduced in the styling apparatus. The sensing means may include capacitive sensing means to detect the amount of hair between the heaters.

Motion: Rotation and Speed As already stated, the one or more motion detection devices 31, comprising any one of an accelerometer, a gyrometer, a magnetometer, an inclination sensor, are configured to enable the microprocessor 29 to determine motion of the styling appliance. The motion may comprise the rotation and/or the speed of the styling appliance.

Grip Pressure The hair styler 1 includes a pressure sensor 32 to enable the microcontroller 28 to measure a pressure between the first movable arm 4a and the second movable arm 4b, e.g., when styling the hair.

Power Demand As already stated, the controller 28 comprises means to measure the power demand.

In a preferred example, the microcontroller 28 determine a power demand of the apparatus, based on a temperature of the plurality of independently controllable heating zones or based on the power output needed to maintain each heating zone of the plurality of heating zones at a target temperature.

Determination of class of hair based on the streams of data Hair types have already been broadly categorised into nine reference classes (la, lb, lc, 2a, 2b, 2c, 3a, 3b, 3c).

The class la corresponds to fine, straight blonde hair and the class 3c corresponds to thick, curly dark hair.

As explained in greater detail below, the memory 30 comprises a trained classifier, comprising a trained machine learning algorithm, which determines the class of hair of the user, based on their use of the hair styling appliance.

As explained in greater detail below, the microcontroller 28 is configured to use the stored trained classifier to determine the type of hair of the user by classifying the hair into a reference hair class, based on the determined power demand and the determined hair tress size and control the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones which are engaged with the hair, based on the determined hair type.

The microcontroller 28 is configured to control the temperature or the power output of the one or more heating zones by capping the increase in the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones based on the determined type of hair, to avoid the hair to be burnt. Therefore even users with hair types which do not correspond to that of most users do not get their hair damaged when styling or drying their hair.

As explained in greater detail below, the controller 28 is configured set the target temperature or the power output of the one or more heating zones of the plurality of controllable heating zones which are engaged with the hair, based on the determined type of hair with reference to the nine reference classes. The controller 28 is configured set the target temperature or the power output of the one or more heating zones of the plurality of controllable heating zones, based on the determined type of hair with reference to the nine reference classes, such that even users with hair types which do not correspond to that of most users have a better styling or drying experience.

Generating the classifier Figure 8 shows a flow chart illustrating an example method 100 according to the disclosure. The method 100 is for generating a classifier configured to determine the type of hair of the user of the hair styling appliance. In Figure 8, the method 100 comprises: obtaining, at S1, a plurality of annotated training streams of data; and training, at S2, the classifier by applying a machine learning algorithm to the obtained training streams of data.

The learning process is typically computationally intensive and may involve large volumes of training data.

As explained in more detail below, the machine learning step S2 involves inferring the type of hair, such as the class of hair among the nine reference classes mentioned above, based on the training data, and encoding the determined class in the form of the classifier.

The training data are annotated. In other words, during the training and in the streams of data, the class of hair is known. As such, in the method 100 of Figure 8, the annotation indicates the class of hair associated with each training stream of data. A domain specialist may manually annotate the streams data with ground truth annotation corresponding to the class of hair during the training.

In the method 100 of Figure 8, the plurality of annotated training streams of data comprises, for each class of hair among the plurality of reference hair classes: a training stream of data corresponding to a power demand of the apparatus when styling or drying hair belonging to the class of hair; and a training stream of data corresponding to a hair tress size which is typically loaded in the apparatus when styling or drying hair belonging to the class of hair.

For example,

for a la class of hair, a training stream of data corresponding to the power demand of the appliance shows less power demand than that for a 3c class of hair, as the la class of hair requires less heat than the 3c class of hair; and for a 1a class of hair, a training stream of data corresponding to the tress load in the appliance shows tress load than that for a 3c class of hair, as the la class of hair is thinner and straighter than the 3c class of hair.

Training at S2 the classifier comprises training the machine learning algorithm to minimise a classification loss, for the classification of the hair in a class of the plurality of classes of hair, such as a similarity metric of Lp-norm, p being an integer greater or equal to 1, such as an average absolute deviation or a least mean square distance from the class of hair as determined by the classifier during the training to the known class of hair. Other ways of calculating the classification loss are envisaged.

Referring back to Figure 8, the classifier is built by applying the machine learning algorithm to the training data. Any suitable machine learning algorithm may be used for building the classifier. For example, approaches based on a convolutional neural network may be used.

As explained above, the classifier is primarily trained to determine the type of hair of the user based on the power demand and the hair tress size. However, additional, optional training streams of data may be fed to the classifier during training, such as one or more training streams of data corresponding to sensor data indicative of a motion of the apparatus when the user is styling or drying hair belonging to the class of hair. The motion of the hair styling or drying appliance comprises a rotation of the apparatus when styling or drying the hair and/or a speed of movement of the apparatus when styling or drying the hair.

Additionally, the classifier may be trained using a training stream of data corresponding to sensor data indicative of diagnostic information about the hair being styled or dried. The diagnostic information may comprise at least one of: a level of moisture in the hair, a humidity ambient to the hair being styled or dried, a temperature ambient to the hair being styled or dried, a geographic location of the hair being styled or dried.

Figure 9 shows a non-limiting example architecture of a classifier according to the disclosure. The classifier of Figure 9 is generated based on the training data obtained at Si. In Figure 9, the training streams of data comprise: a training stream of data corresponding to a power demand of the apparatus when styling or drying hair belonging to the class of hair; and a training stream of data corresponding to a hair tress size which is typically loaded in the apparatus when styling or drying hair belonging to the class of hair.

a training stream of data corresponding to sensor data indicative of a grip pressure between movable arms of the apparatus when styling or drying hair belonging to the class of hair; a stream of data corresponding to the rotation of the hair styling appliance which is typical for the given class of hair; and a stream of data corresponding to the speed of motion of the hair styling appliance when in use which is typical for the given class of hair.

For example,

fora 1a class of hair, a training stream of data corresponding to the grip pressure of the appliance shows less grip pressure than that fora 3c class of hair, as the 1a class of hair requires less grip pressure for styling than the 3c class of hair; for a la class of hair, a training stream of data corresponding to the rotation of the hair styling appliance shows less rotation of the hair styling appliance than that for a 3c class of hair, as the 1a class of hair is usually easier to curl than the 3c class of hair; and for a 1a class of hair, a training stream of data corresponding to the speed of motion of the hair styling appliance shows more speed of motion of the hair styling appliance than that for a 3c class of hair, as the 1 a class of hair is usually quicker to style than the 3c class of hair.

In the non-limiting example of Figure 9, in, the example classifier comprises one or more layers: an appliance input layer 101, such that the input layer 101 E Il85, each dimension of the appliance input layer 101 corresponding to a training stream of data; a hidden layer 102, such that the hidden layer 102 E 1W5; and an output layer 103 such that the output layer 104 c 1W9, each dimension of the output layer 103 corresponding to a class of hair.

Other configurations with other layers may also be envisaged, and other architectures are also envisaged for the classifier. For example, deeper architectures may be envisaged and/or an architecture of the same shape as the architecture described above that would generate an output layer 103 with sizes different from those already discussed may be envisaged.

For example, an architecture using user input is shown in Figure 10.

As shown in Figure 10, the classifier comprises one or more layers: an appliance input layer 101, such that the input layer 101 c Il85, each dimension of the input layer 101 corresponding to a stream of training data; a user input 104; a hidden layer 102, such that the hidden layer 102 c -85; and an output layer 103 such that the output layer 104 E -S3, the total dimension of the output layer 103 corresponding to a narrowed down subset of the classes of hair.

The user input 104 enables the classifier to narrow down the possible hair classes of the user (for example here from nine reference classes to only three possible classes) and to deliver a more confident prediction of the hair class.

The input in the user input 104 may result from the user simply indicating to the classifier the class of hair they believe they have. A confidence interval may be taken around the class of hair indicated by the user. For example, if the user indicates that they have a 2c class of hair, the classifier can narrow down the possible class of hair of the user to a narrowed down subset of the nine reference classes of hair which comprises e.g., one class around the input class, such as either class 2b, 2c and 3a. It should be understood that other dimensions of the subset of the classes of hair (e.g., two classes around the input class) are also envisaged.

Alternatively or additionally, the input in the user input 104 may result from the user answering a questionnaire, asking questions to the user about their hair (e.g., general questions or more specific questions such as "has the hair been moisturised?" or "has the hair been bleached?"...), in order to determine a narrowed down subset of the classes of hair. The questionnaire may ask more advanced questions, e.g., asking questions about the environment (such as room temperature or humidity as non-limiting examples) to account for the influence of the environment on different hair classes.

The input in the user input 104 may be performed on the user interface 11 and/or a remote application (e.g., on a mobile telephone) connected to the communications circuitry 27.

After it has been trained, the classifier is used for determining the class of hair of the user of the hair styling appliance, with a confidence in the determined classification (in % of confidence in the classification).

Based on the confidence of the classification, a matrix can be built to apply different settings (including e.g., a target temperature of the heaters) to the hair styling appliance (see Table 1 reproduced below).

Confidence in In class In class In class In class In class In class In class In class In class classification la lb lc 2a 2b 2c 3a 3b 3c Greater than 85% Setting Setting Setting Setting Setting Setting Setting Setting Setting la lb lc 2a 2b 2c 3a 3b 3c Between Setting Setting Setting Group 6 Setting Group 8 Setting Group 10 Setting 55% and Group 4 Group 4 Group 85% 11 Setting Group 5 Setting Group 7 Setting Group 9 Setting Group 11 Between Setting Group 1 Setting Group 2 Setting Group 3 15% and 54% Smaller than 15% Default setting for all classes of hair

Table 1

When the confidence in the classification by the classifier is greater than a great degree of confidence, e.g., greater than 85% in Table 1, each class of hair has its own setting, for example a target temperature of 160°C for class 1 a and a target temperature of 200°C for class 3c.

When the confidence in the classification is great, e.g., a degree of confidence between 55% and 85% in Table 1, small overlapping groups of classes of hair have their own setting, for example a target temperature of 170°C if the determined class of hair is between class 1 a and class 1 b (Setting Group 4) and a target temperature of e.g., 190°C if the determined class of hair is between class 3b and class 3c (Setting Group 11).

When the confidence in the classification is not great, e.g., a degree of confidence between 15% and 84% in Table 1, three main groups of classes of hair have their own setting, for example a target temperature of e.g., 175°C if the determined class of hair is between classes la and lc (Setting Group 1), a target temperature of e.g., 180°C if the determined class of hair is between classes 2a and 2c (Setting Group 2), and a target temperature of e.g., 185°C if the determined class of hair is between classes 3a and class 3c (Setting Group 3) forclass 3c. If there is no confidence in the classification, the target temperature is set to the default temperature for all of the classes of hair. Other temperatures and groupings than those mentioned above are envisaged.

Computer system and hair styling appliance A computer system (not shown in the Figures) may execute the deep learning algorithm to generate the classifier to be stored on the memory 30 of the hair styling appliance. The computer system may communicate and interact with multiple such hair styling appliances. The computer system may conventionally comprise a memory, a processor and a communications interface. The computer system may be configured to communicate with one or more hair styling appliances, via the communications interface and a link (e.g. Wi-Fi connectivity, but other types of connectivity may be envisaged).

The memory of the computer system is configured to store data, for example for use by the processor. In some examples the data stored on the memory may comprise the training data and/or the deep learning algorithm.

In some examples, the training data may correspond to actual observed data streams on the hair styling appliance, or the training data may be generated, for example in a laboratory.

The training may be performed at the computer system separate, optionally remote, from hair styling appliance.

However, if sufficient processing power is available locally then the classifier learning could be performed (at least partly) by the microprocessor 29 of the hair styling appliance.

The classifier is arranged to produce the determination of the class of hair more easily, after it is stored in the memory 30 of the hair styling appliance, even though the process 100 for generating the classifier from the training data may be computationally intensive.

After it is configured, the hair styling appliance may provide determination of a class of hair of the user of the hair styling appliance, by applying the learned classifier during use of the hair styling appliance, using the same input data streams as during the training.

Hair styling appliance manufacture As illustrated in Figure 11, the method 200 of producing the hair styling appliance configured to determine the class of hair of the user using the hair styling appliance comprises: obtaining, at S31, a classifier generated by the method 100 according to any aspects of the disclosure; and storing, at S32, the obtained classifier in the memory 30 of the hair styling appliance.

The classifier may be generated and stored using any suitable representation, for example as a data description comprising data elements specifying classification conditions and their classification outputs. Such a data description could be encoded e.g. using XML or using a bespoke binary representation. The data description is then interpreted by the microprocessor 29 running on the appliance when applying the classifier. Alternatively, the deep learning algorithm may generate the classifier directly as executable code (e.g. machine code, virtual machine byte code or interpretable script). This may be in the form of a code routine that the appliance can invoke to apply the classifier.

The appliance may be connected temporarily to the computer system to transfer the generated classifier (e.g. as a data file or executable code) or transfer may occur using a storage medium (e.g. memory card). In a preferred approach, the classifier is transferred to the appliance from the computer system over the communications link (this could include transmission over the Internet from a central location of the computer system to a local network where the appliance is located using the communications circuitry 27, e.g., via a mobile phone connected to the appliance via the Bluetooth, Fi and/or 3GPP communication protocols).

Thanks to the communications circuitry 27, the classifier could be installed as part of a firmware update of device software, or independently. Installation of the classifier may be performed once (e.g. at time of manufacture or installation) or repeatedly (e.g. as a regular update). The latter approach can allow the classification performance of the classifier to be improved over time, as new training data become available. The latter approach can allow updating the setting for different hair types.

Applying the classifier to perform classification Figure 12 shows a flow chart illustrating an example method 300 for operating the hair styling appliance, using the classification of the hair of the user of the hair styling appliance. The method 300 is performed by the hair styling appliance.

The method 300 comprises: obtaining, at S41, streams of data during use of the hair styling appliance; classifying, at S42, the hair of the user, by applying the learned algorithm to the obtained streams of data; and applying, at S43, settings to the hair styling appliance, based on the classifying.

The obtained streams of data correspond to the observed streams of data during use of the hair styling appliance and which are similar to the training streams of data.

As already stated, the streams of data primarily comprise a power demand and a hair tress size which has been loaded by the user, and the trained classifier is configured to determine the type of hair of the user by classifying the hair into a reference hair class, based on the determined power demand and the determined hair tress size, so that the appliance can control the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones which are engaged with the hair, based on the determined hair type.

However, additional, optional streams of data may be fed to the classifier during use to have a better classification. Optional additional streams of data may include: one or more training streams of data corresponding to sensor data indicative of a motion of the apparatus when the user is styling or drying hair belonging to the class of hair. The motion of the hair styling or drying appliance comprises a rotation of the apparatus when styling or drying the hair and/or a speed of movement of the apparatus when styling or drying the hair, as explained above.

a stream of data corresponding to the grip pressure that the user applies to the hair styling appliance when styling or drying.

The communication circuitry may receive the user input information from an external processing device of a mobile phone of the user, and the stored trained classifier can narrow down the plurality of reference hair classes into which the hair of the user is to be classified, based on the received user input information. The input may comprise at least one of: an indication by the user of the hair type they believe they have and/or a result from the user answering a questionnaire about their hair and/or the environment, as explained above.

The controller 28 may configured to process further sensor data to determine diagnostic information about the hair being styled or dried, and the diagnostic information comprises at least one of: a level of moisture in the hair, a humidity ambient to the hair being styled or dried, a temperature ambient to the hair being styled or dried, a geographic location of the hair being styled or dried.

After the class of hair has been determined by the appliance with a percentage of confidence, the settings corresponding to Table 1 already discussed may be applied to the appliance. The microcontroller controls the temperature or the power output of the one or more heating zones by capping the increase in the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones based on the determined type of hair, to avoid the hair to be burnt.

The applied settings correspond to the determined class of hair, such that the settings are tailored to the type of hair of the user. The user has thus a better styling experience, a better styling result and the hair is less likely to get damaged when styling.

Modifications and alternatives Detailed embodiments and some possible alternatives have been described above. As those skilled in the art will appreciate, a number of modifications and further alternatives can be made to the above embodiments whilst still benefiting from the inventions embodied therein. It will therefore be understood that the invention is not limited to the described embodiments and encompasses modifications apparent to those skilled in the art lying within the scope of the claims appended hereto.

The invention has been described above by way of implementation in a hair styling device for straightening hair chair straighteners') which employ flat hair styling heaters 6. However, it could alternatively be implemented in any form of hair styling device, such as (but not limited to) crimpers, curlers or heated brushes. The heaters 6 may define a heating surface that is flat, curved, ridged or in the shape of a barrel. The hair styling device may have two arms like the device illustrated in Figure 1 or it may be a single armed device. The heaters described above may also be used in hair dryers or in combination devices that use conductive heating and air to dry and style the user's hair (such as those described in the applicant's earlier PCT application WO 2021/019239).

In embodiments where air is used, the heaters 6 may be perforated so that air passes through the heater and is warmed by the heater as the air passes through.

In the above embodiments, Metal Oxide Semiconductor Field Effect Transistor (MOSFET) switches were used to control powering and sensing of the heater electrodes. As those skilled in the art will appreciate, other switches could be used instead. For example, Field Effect Transistors (FETs) could be used, such as Gallium Nitride FETs or bipolar junction transistors (BJTs).

In the above embodiments, a DC power source was used to provide electrical power for heating the heater electrodes 64. This DC power source will typically be one or more batteries, although DC supplies that derive their power from a mains power AC signal may be used. Thicker or more dielectric layers are typically used between the heater electrodes 64 and the hair contacting surface of the hair styler when AC power is used to heat the heaters.

In the above-described examples the hair styling device 10 may comprise a single heater 6, or may alternatively comprise two or more heaters 6.

Throughout the description and claims of this specification, the words "comprise" and "contain" and variations of the words, forexample "comprising" and "containing", means "including but not limited to", and is not intended to (and does not) exclude other components, integers or steps.

The expressions "to dry hair", "drying hair" or "decrease a moisture level of hair" and the like, as used in the present disclosure, can refer both to the removal of "unbound" water that exists on the outside of hair when wet, or the removal of "bound" water, which exists inside individual hairs, and which can be interacted with when heat styling hair.

The "bound" water need not necessarily be removed when drying hair, although removal of some bound water may occur during a drying or styling process.

Various other modifications will be apparent to those skilled in the art and will not be described in further detail here.

Claims

CLAIMS1. Apparatus for drying or styling hair, the apparatus comprising: a heater comprising a hair contacting surface for heating hair of a user that 5 contacts the hair contacting surface by conduction, the heater comprising a plurality of independently controllable heater electrodes that define a plurality of independently controllable heating zones of the hair contacting surface; a memory that stores a trained classifier configured to classify the hair of the user of the apparatus into a class among a plurality of reference hair classes; and a controller coupled to the memory and configured to control the apparatus to: determine a power demand of the apparatus, based on a temperature of the plurality of independently controllable heating zones or based on the power output needed to maintain each heating zone of the plurality of heating zones at a target temperature; determine a hair tress size which has been loaded by the user into the apparatus, by determining, based on a drop of the temperature of one or more heating zones of the plurality of independently controllable heating zones or based on an increase in the power output needed to maintain the one or more heating zones at the target temperature, whether hair is engaged with the one or more heating zones, and, if it is, by determining an amount of hair which is engaged with the one or more heating zones, the hair tress size being determined based on the amount of hair engaged with the one or more heating zones across all of the one or more heating zones; use the stored trained classifier to determine the type of hair of the user by classifying the hair into a reference hair class, based on the determined power demand and the determined hair tress size; cause the apparatus to control the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones which are engaged with the hair, based on the determined hair type.
2. The apparatus of claim 1, further comprising: communication circuitry such that the apparatus is configured to receive user input information from an external processing device, and wherein the user input information is input on a user interface associated with the user and configured to enable the user to input the input information.
3. The apparatus according to claim 2 wherein the external processing device is part of a mobile phone or a smart watch and is configured to run an application, remote 10 from the apparatus.
4. The apparatus according to claim 2 or 3, wherein the controller is further configured to control the apparatus to use the stored trained classifier such that the plurality of reference hair classes into which the hair of the user is to be classified is narrowed down to a subset of the plurality of reference hair classes, based on the received user input information.
5. The apparatus according to claim 4, wherein the user input information comprises at least one of: an indication by the user of the hair type they believe they have; a result from the user answering a questionnaire about their hair and/or the environment.
6. The apparatus according to any preceding claim, wherein the controller is configured 25 to: determine a grip pressure based on data associated with a pressure sensor configured to measure a pressure between movable arms of the apparatus; and cause the apparatus to use the stored trained classifier to determine the type of hair of the user, further based on the grip pressure.
7. The apparatus according to any preceding claim, wherein the controller is configured to: determine a motion of the apparatus when the user is styling or drying their hair, based on data associated with one or more motion sensors of the apparatus; and cause the apparatus to use the stored trained classifier to determine the type of hair of the user, further based on the motion of the apparatus.
8. The apparatus according to claim 7, wherein the one or more motion sensors of the apparatus comprise any one or more of an accelerometer, a gyrometer, a magnetometer, an inclination sensor, and wherein the motion of the hair styling or drying appliance comprises: a rotation of the apparatus when styling or drying the hair; and/or a speed of movement of the apparatus when styling or drying the hair.
9. The apparatus according to any preceding claim, wherein the controller is configured to: process further sensor data to determine diagnostic information about the hair being styled or dried; and cause the apparatus to use the stored trained classifier to determine the type of hair of the user, further based on the determined diagnostic information.
10. The apparatus of claim 9, wherein the diagnostic information comprises at least one of: a level of moisture in the hair, a humidity ambient to the hair being styled or dried, a temperature ambient to the hair being styled or dried, a geographic location of the hair being styled or dried.
11. The apparatus according to any preceding claim, wherein the controller is configured to control the temperature or the power output of the one or more heating zones by capping the increase in the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones based on the determined type of hair, to avoid the hair to be burnt.
12. The apparatus according to any of claims 1 to 11, wherein the plurality of reference hair classes comprises nine reference classes, referred to as (1a, 1b, 1c, 2a, 2b, 2c, 3a, 3b, 3c), the class la corresponding to fine, straight blonde hair and the class 3c corresponding to thick, curly dark hair, and wherein the controller is configured to set the target temperature or the power output of the one or more heating zones of the plurality of controllable heating zones 5 which are engaged with the hair, based on the determined type of hair with reference to the nine reference classes.
13. The apparatus according to claim 12, wherein the controller is configured to set the target temperature or the power output of the one or more heating zones of the plurality 10 of controllable heating zones so as to enable a better drying or styling of the hair which is loaded in the apparatus.
14. The apparatus of any preceding claim, wherein the trained classifier comprises a machine learning algorithm comprising a convolutional neural network.
15. The apparatus according to any preceding claim, wherein the apparatus is a hair straightener, a hair dryer, a hot paddle brush, a hot round brush, a heater roller, or a hair curler.
16. A method for generating a classifier configured to classify hair of a user of apparatus for drying or styling hair, into a class among a plurality of reference hair classes, the method comprising: obtaining a plurality of annotated training streams of data; and training the classifier by applying a machine learning algorithm to the obtained 25 training streams of data, wherein the annotation indicates the class of hair associated with each training stream of data, and wherein the plurality of annotated training streams of data comprises, for each class of hair among the plurality of reference hair classes: a training stream of data corresponding to a power demand of the apparatus when styling or drying hair belonging to the class of hair; and a training stream of data corresponding to a hair tress size which is typically loaded in the apparatus when styling or drying hair belonging to the class of hair.
17. The method according to claim 16, wherein the classifier is trained to minimize a classification loss between the annotated class of hair associated with each training stream of data and a classification of the hair determined by the classifier.
18. The method of claim 17, wherein the classification loss comprises a similarity metric of 143-norm, p being an integer greater or equal to 1, such as an average absolute deviation or a least mean square distance.
19. The method of any of claims 16 to 18, wherein the plurality of training streams of data comprises, for each class of hair among the plurality of reference hair classes: a training stream of data corresponding to sensor data indicative of a grip pressure between movable arms of the apparatus when styling or drying hair belonging to the class of hair; and/or one or more training streams of data corresponding to sensor data indicative of a motion of the apparatus when the user is styling or drying hair belonging to the class of 20 hair, optionally wherein the motion of the hair styling or drying appliance comprises: a rotation of the apparatus when styling or drying the hair; and/or a speed of movement of the apparatus when styling or drying the hair; and/or a training stream of data corresponding to sensor data indicative of diagnostic 25 information about the hair being styled or dried.
20. The method according to any of claims 16 to 19, wherein the classifier is trained to classify the hair of the user into a subset of the plurality of reference hair classes, based on user input information.
21. The method according to claim 20, wherein the user input information comprises at least one of: an indication of a suspected use's hair type; a result from a questionnaire about their hair and/or the environment.
22. The method of any of claims 16 to 21, wherein the machine learning algorithm 5 comprises a convolutional neural network.
23. A method of producing apparatus for drying or styling hair, wherein the apparatus comprises a heater comprising a hair contacting surface for heating hair of a user that contacts the hair contacting surface by conduction; the heater comprising a plurality of independently controllable heater electrodes that define a plurality of independently controllable heating zones of the hair contacting surface, a memory and a controller coupled to the memory, the method comprising: obtaining a classifier generated by the method according to any one of claims 16 to 21 and storing the obtained classifier in the memory of the appliance, the controller being configured to control the apparatus to: use the stored trained classifier to determine the type of hair of the user by classifying the hair into a reference hair class; cause the apparatus to control the temperature or the power output of the one or more heating zones of the plurality of controllable heating zones which are engaged with the hair, based on the determined hair type.
24. The method of claim 23, wherein the storing comprises transmitting the generated classifier to the apparatus via a network, the apparatus receiving and storing the classifier, optionally wherein the classifier is generated, stored and/or transmitted in the form of one or more of: a data representation of the classifier; executable code for applying the classifier.
25. A computer program or a computer program product comprising instructions 30 which, when executed by a processor, enable the processor to perform the method according to any one of claims 16 to 24.