GB2613869A - Sensing apparatus and method of manufacture - Google Patents

Abstract

Headphones 100 comprise a deformable earpad 102. The ear pad 102 has electroencephalography sensors 106,107. One EEG sensor 106A sits above a user’s ear, another 106B below the ear. A reference EEG 107 is included. The earpad may have only two EEG sensors, with one acting both as active and reference sensor. The EEG sensors may have an interchangeable head. Also claimed: an EEG signal obtained from a user is denoised based on, e.g., a user’s heart rate, motion, sweat, or skin conductance. Also claimed: an EEG sensor with a resiliently deformable support scaffold coated in a conductive layer, the scaffold providing bristles for contacting a user’s scalp (Fig. 4A). Glasses or headphones may have these EEG sensors. Also claimed: making EEG sensors by forming a resiliently deformable support scaffold from silicone or rubber; and brushing a layer of electrically conductive coating onto the support scaffold.

Description

Sensing apparatus and method of manufacture

Field of the invention

The present disclosure relates to a sensing apparatus and method of manufacture, in 5 particular to an electroencephalography sensor and method of manufacture.

Background

Sensing the emotions of an individual can be desirable for many reasons. One such reason may be to determine their likes or dislikes, such as music that the individual enjoys listening to or not. However, a conventional technique to sense emotions may involve performing image recognition on a user's face, but this is cumbersome and typically requires a camera trained on the individual's face which isn't mobile or comfortable for users, and so consequently has a limited range of usage times. Another conventional technique involves sensing voice, however, the limitations with voice are that it only works while talking which isn't always accessible, as well also not being as accurate. Another conventional technique involves large neural imaging arrays, but these aren't portable and are obtrusive and don't fit into user's existing habits. Another conventional technique involves monitoring heart rate, but this is low fidelity.

US 2020/0060571 describes a device for measuring and/or stimulating a brain activity, preferably an EEG device, comprising means for transmitting and/or detecting physiological signals produced by the brain of an individual, and a support for the transmission and/or detection means, wherein the support is configured to extend over the top of the individual's head, the support comprising means for removably attaching to an accessory intended to be worn by the individual, on his or her head, such as an audio headset, the support being configured such that, when the device is worn by the individual, the means for transmitting and/or detecting physiological signals are held in substantially close contact with the individual's head by the accessory. -2 -

WO 2020/120865 describes an ionic conductive polymeric composition defined by the following general formula: (PH)x + (SOH)y + z(MCI); in which: -PH represents a polymer containing protic functions; -SOH represents a plasticizing polyol with a molecular mass of not less than 75 g/mol and not greater than 250 g/mol, in the form of discrete molecules; -MCI represents sodium or potassium chloride (M= Na or K); -0.3 5)dy 5 3, x representing the amount by weight of the polymer PH, and y the amount by weight of the polyol SOH; -0.5% 5 z 5 15%, z representing the percentage by weight of MCI relative to the polyol SOH. Said polymeric composition may be used particularly as conductive material in electrodes for measuring electrophysiological signals.

US 2021/0219896 describes an electrode for recording a physiological electrical signal of a living being made from electrically conductive material and including a body having a front surface and a rear surface for electrical contact, the electrode further including a rigid mounting element attached to the body, at an edge of the rear electrical contact surface and designed so as to produce a removable mounting of the electrode on a recording sensor.

Bleichner et al., "Concealed, Unobtrusive Ear-Centered EEG Acquisition: cEEGrids for Transparent EEG", frontiers in Human Neuroscience, 7 April 2017, doi: 10.3389/fnhum.2017.00163 describes a "transparent" EEG technique which aims for motion tolerant data acquisition and describes ear-centred EEG solutions, and shows that miniature electrodes placed in and around the human ear are a feasible solution and are sensitive enough to pick up electrical signals stemming from various brain and non-brain sources.

Summary of the invention

Aspects of the invention are as set out in the independent claims and optional features are set out in the dependent claims. Aspects of the invention may be provided in conjunction with each other and features of one aspect may be applied to other aspects.

In a first aspect there is provided an apparatus for obtaining physiological electrical -3 -signals of a living being such as a human. The apparatus comprises headphones or a headphone insert comprising at least one deformable ear pad configured in use to be biased against the user's head around one of the user's ears. The deformable ear pad comprises at least three electroencephalography sensors configured to receive electrical signals from the user, wherein in use at least one of the electroencephalography sensors is located above the user's ear, and at least one of the electroencephalography sensors is located below the user's ear, and wherein a third electroencephalography sensor is configured to act as a reference.

In some examples, however, it will be understood that the deformable ear pad may comprise only two electroencephalography sensors, with one sensor being configured to act as both an active sensor and a reference sensor.

Advantageously, embodiments of the claims provide a smaller and more portable means to obtain emotion data from users, for example via electroencephalography. Advantageously, the noise removal methods of embodiments of the disclosure enable on-the-go classifications as well as a wider range of emotions with higher accuracy.

The electroencephalography sensor may be conductive or capacitive. The electroencephalography sensors may be resiliently deformable. The electroencephalography sensors may be at least partially embedded into the at least one ear pad.

In some examples the apparatus, such as headphones, comprises a pair of ear pads, one ear pad of the pair of ear pads comprising at least three electroencephalography sensors, and one ear pad of the pair of ear pads comprising at least two electroencephalography sensors. For example, where there are three electroencephalography sensors one of these may act as a reference sensor, with the 30 other two acting as active sensors. -4 -

The electroencephalography sensors may comprise an anchor portion and a head portion, for example an interchangeable head portion. The anchor portion may be submerged or embedded within an ear pad and the head portion may be exposed from 5 the ear pad. In some examples the anchor portion may be conductive. For example, the anchor portion may be a conductive mount for the interchangeable head portion. For example, the anchor portion may comprise a mating portion, for example a male mating portion, and the head may comprise a corresponding mating portion, for example a corresponding female mating portion, adapted to mate with the male mating portion of 10 the anchor. In this way advantageously different head portions may be used for different users. This may be particularly advantageous due to different users' hairstyles requiring different head portions to contact the underlying skin/scalp.

Each head portion may comprise a plurality of bristles configured to contact a user's scalp through their hair. For example, each head portion may comprise at least three bristles, and preferably between five and fifteen bristles. In some examples each bristle may be located on a respective corner or edge of the head portion. Advantageously this may improve the degrees of freedom of each bristle thereby improving contact of bristles with the skin/scalp. In some examples, the plurality of bristles vary in length. For example, the bristles may vary in length from one head portion to another or may vary in length within each head portion.

In some examples each of the electroencephalography sensors comprise a resiliently deformable support scaffold coated in an electrically conductive coating. For example, each interchangeable head portion of the electroencephalography sensors may comprise a resiliently deformable support scaffold coated in an electrically conductive coating such as an electrically conductive polymer. The coating may be a layer at least 0.5 mm thick, preferably between 0.8 -1.0 mm thick. However, in some examples each electroencephalography sensor may be made entirely from a conductive polymer such 30 as the conductive polymer that may provide the electrically conductive coating. The polymer or conductive coating may comprise a blend of graphite, silicone or rubber, and -5 -petroleum ether (naphtha). The graphite, silicone/rubber and petroleum ether may be in a blend, for example in the ratio of 1:1:1. In some examples the polymer of conductive coating may consist solely of graphite, silicone and petroleum ether. In some examples the polymer of conductive coating may comprise additives, such as carbon nanotubes and/or graphene.

In some examples the apparatus may comprise processing circuitry. The processing circuitry may perform some local processing of the physiological signals. In some examples the processing circuitry comprises a wireless interface, and the processing circuitry is configured to receive sensor signals from the electroencephalography sensors and send them (optionally with a degree of local processing) to a remote device via the wireless interface (e.g., for further processing by the remote device). The processing circuitry may further comprise an amplifier for amplifying the signals before they are transmitted via the wireless interface.

In some examples the processing circuitry is configured to perform active signal acquisition using bipolar sensors. However, in some examples the sensors may be operated in a unipolar mode of operation and the signals processed (for example via local or remote processing circuitry) to provide bipolar signals. Measuring signals in a unipolar mode of operation may advantageously allow greater flexibility in terms of what is measured, how the signals are processed, and what the signals are measured against (i.e., what is used as a reference).

In some examples the apparatus further comprises other sensors configured to obtain an indication of a parameter(s) of a user, such as heart rate, motion, sweat, skin conductance. For example, in some examples the apparatus further comprises a heart rate sensor such as a photoplethysmography, PPG, sensor. The PPG sensor may be either embedded in the ear pad or configured to obtain a signal from a user from inside a region bounded by the earpad, for example a region barely touching the earlobe. The processing circuitry (either locally or remotely) may be configured to remove artifacts from the physiological signals (e.g. electroencephalography signals) using other signals -6 -providing an indication of a parameter(s). For example, the processing circuitry (either locally or remotely) may be configured to perform independent component analysis (ICA) to remove noise. For example, the apparatus may comprise a PPG sensor and/or an inertial measurement unit and wherein the processing circuitry is configured to remove motion artefact noise and/or heart rate noise caused by the user and/or muscle movement. In some examples the processing circuitry may comprise an adaptive or digital filter for removing noise. For example, the adaptive or digital filter may be configured to reduce mean square error of a signal.

In some examples the processing circuitry local to the apparatus may be described as having an analogue front end. For example, the processing circuitry local to the apparatus/device may be configured to obtain electroencephalography signals, as well as other signals providing an indication of a parameter(s) of the user in an analogue format. In some examples the local processing circuitry may be configured to convert these analogue signals into a digital format before sending via the wireless interface to a remote device. However, in other examples the apparatus may send the analogue signals via the wireless interface to the remote device.

In another aspect there is provided a computer-implemented method of removing noise from an electroencephalography, EEG, signal. The method comprises receiving signals indicative of electroencephalography signals from a sensor coupled to a user and receiving an indication of a parameter (such as hear rate, motion, sweat) of the user, and processing the electroencephalography signal to denoise the electroencephalography based on the indication of a parameter of the user. Processing the electroencephalography signal to denoise the electroencephalography based on the indication of a parameter of the user comprise performing independent component analysis based on the indication of a parameter of the user. The denoising may be performed locally to the user or at a remote device.

In another aspect there is provided an electroencephalography sensor comprising a resiliently deformable support scaffold coated in a conductive layer. The resiliently -7 -deformable support scaffold may provide a plurality of bristles for contacting a user's scalp. The conductive layer may comprise a blend of graphite or graphene, silicone and petroleum ether. For example, the conductive layer may comprise graphene and/or carbon nanotubes. In some examples the graphite or graphene, silicone and petroleum ether are in the ratio of 1:1:1.

The conductive layer may be at least 0.5mm thick, preferably between 0.8 and 1.0mm thick.

The resiliently deformable support scaffold may comprise at least one of silicone and rubber.

It will be understood that the electroencephalography sensors described above may be incorporated into many other items worn by a user to determine emotions of a user. 15 Accordingly, in other aspects there may be provided headphones, cycling helmet and/or glasses comprising the electroencephalography sensors.

In another aspect there is provided a method of manufacturing electroencephalography sensors. The method comprises forming a resiliently deformable support scaffold from silicone or rubber and brushing a layer of electrically conductive coating onto the support scaffold. In examples where graphite is used, brushing the electrically conductive coating onto the support scaffold advantageously aligns the graphite in the coating to improve conductive of the coating.

However, in some examples (such as when the electrically conductive coating comprises graphene and/or carbon nanotubes) the electrically conductive coating may be applied by dipping the support scaffold in the electrically conductive coating.

Forming the resiliently deformable support scaffold from silicone or rubber may comprise 30 mixing 2-part mix of silicone or rubber and catalyst hardener in a ratio of 1:1. -8 -

Brushing a layer of electrically conductive coating onto the support scaffold comprises brushing to a layer with a thickness of at least 0.5 mm thick, preferably to a thickness of 0.8-1mm. In some examples the method may comprise allowing the brushed electrically conductive layer to cure after brushing for at least 3 hours, preferably 3 to 5 hours.

The conductive layer may comprise a blend of graphite or graphene, silicone and petroleum ether. In some examples the graphite may be in the form of carbon nanotubes.

In another aspect there is provided a kit comprising a pair of ear pads and a plurality of interchangeable electroencephalograph sensor heads to be used with the ear pads. Each of the plurality of interchangeable heads may have bristles of a different length and/or configuration (e.g. shape and/or arrangement/layout) to each other.

Drawings Embodiments of the disclosure will now be described, by way of example only, with reference to the accompanying drawings, in which: Fig. 1 shows a perspective view of a pair of example headphones comprising ear pads comprising electroencephalography sensors; Fig. 2 shows a side view of the example headphones of Fig. 1; Fig. 3A shows a perspective view of an example electroencephalography sensor coupled to a portion of the headphone of Figs. 1 and 2; Fig. 38 shows a perspective of the example electroencephalography sensor of Fig. 3A embedded into the ear pad; Fig. 4A shows a perspective of an example electroencephalography sensor; Fig. 48 shows a side view of the example electroencephalography sensor of Fig. 4A; Fig. 5A shows a side view of another example electroencephalography sensor; Fig. 58 shows an end view of the example electroencephalography sensor of Fig. 5A; Fig. 50 shows a plan view of the example electroencephalography sensor of Figs. 5A and 58; -9 -Fig. 5D shows a perspective view of the example electroencephalography sensor of Figs. 5A to 50; Fig. 6 shows an example process flow chart of a method of manufacturing an example electroencephalography sensor such as the example electroencephalography sensor of 5 any of Figs. 3A to 50; Fig. 7 shows an example process flow chart of a method of removing noise from an electroencephalography signal; and Fig. 8 shows a schematic of example processing circuitry that may be used, for example, with the headset of Figs. 1 and 2

Specific description

Fig. 1 shows a perspective view of a pair of example headphones comprising ear pads comprising electroencephalography sensors and Fig. 2 shows a side view of the example headphones of Fig. 1. The headphones shown in Figs. 1 and 2 comprise a left earpiece and a right earpiece coupled by a headband 105. The left earpiece comprises a left can 109 coupled to the headband 105 via a first hanger 110 and comprises a left ear pad 101. The right earpiece comprises a right can 108 coupled to headband 105 via a second hanger 112 and comprises a second ear pad 102. The second ear pad 102 comprises first and second active encephalography sensors 106A, 106B and a reference sensor 107. The first ear pad 101 comprises only first and second active encephalography sensors 104A, 104B. Each ear pad 101, 102 is deformable and is adapted, in use, to surround a user's ear. In some examples the ear pads 101, 102 a removable and replaceable/interchangeable e.g., based on the size of the user's head/ears to provide a better fit. On the left ear pad 101 the first active encephalography sensor 104A is located so that in use it is above the user's ear, and the second active encephalography sensor 104B is located so that in use it is below the user's ear. On the right ear pad 102 the first active encephalography sensor 106A is located so that in use it is above the user's ear, and the second active encephalography sensor 106B is located so that in use it is below the user's ear. On the right ear pad 102 the reference sensor 107 is located in use to be just in front of the user's ear but could equally be located elsewhere such as just in front of the user's ear. In some examples instead of being part -10 -of the ear pads 101, 102 the reference sensor 107 could be located elsewhere so long as it is in a position configured to contact the user's skin/scalp e.g., as part of the headband 105. The exact placements varying between individuals may be compensated for via a calibration process explained in more detail below.

In the example shown, each of the sensors 104A, 104B, 106A, 106B and 107 are the same (as in they have the same form factor and profile) but in other examples the sensor may differ. For example, the sensors may have different head portions and bristles (as described in more detail below) to improve contact with the patient's skin if they have a lot of hair near their ears. For example, sensors located near the top of the user's ear may have longer bristles to protrude through hair compared to sensors located near the bottom of the user's ear.

Fig. 3A shows a perspective view of an example electroencephalography sensor such as any of the sensors 104A, 104B, 106A, 106B, 107 coupled to a portion of the headphone (in this example being part of the can 108, 109) of Figs. 1 and 2, and Fig. 3B shows a perspective of the example electroencephalography sensor of Fig. 3A embedded into an ear pad 102. In the example shown active sensor 106A comprises a head portion 308 and an anchor portion 310. The head portion 308 comprises a plurality of conductive and resiliently deformable bristles 350 (in this example four) for contacting the user's skin. Providing resiliently deformable bristles 350 in this manner advantageously improves contact with the user's scalp without causing pain when worn by the user for a sustained period. The bristles 350 of the head portion 308 of the sensor 106A are coupled to circuitry in the ear can 108 by a series of low-resistance copper components.

These components are also designed to retain the flexible electrode to the headphone ear pad 102 for reliable placement and contact to the user's head. To ensure the best contact to the soft electrode provided by the bristles 350, a length of copper wire 318 is wound around the anchor portion 310 of the electrode 4-6 times and compressed so that the anchor portion 310 is deformed slightly. This reduces resistance between the connection. As the headphone ear pad 102 is flexible and moves to form a decent seal around the user's head, the electrodes needed to comply to that movement to form a decent connection to the user's skin. As the active electrodes 104A, 1043, 106A, 1063 need to be hard-mounted and fixed in-place inside the headphones 100, a flexible and reliable method of connecting the electrode to the active sensors is required. For this, a 25mm length of copper ribbon 317 (-6mm in width) is used. One end of the ribbon 317 is soldered directly to the copper wire 318 on the electrode. The other end is left to make contact to the active electrode plate 315 that is coupled to the ear can 108 during final assembly. The anchor portion 310 of the electrode 106A, together with its copper wire 318 and copper ribbon 317 is then inserted through a hole in the ear pad 102. This leaves the head portion 308 of the electrode protruding out, where a copper retaining pin 316 (a 20mm length of copper wire) is then inserted to ensure the electrode is locked in place. The copper ribbon 317 is then pressed against the active electrode plate 315 and secured in place once the ear pad 102 is press fitted to the headphones 100.

Fig. 4A shows a perspective of an example electroencephalography sensor such as the sensor 106A described above with reference to Figs. 3A and 33. Fig. 43 shows a side view of the example electroencephalography sensor of Fig. 4A. In the example shown in Figs. 4A and 4B, it can be seen that the sensor 106A comprises a head portion 308 comprises four bristles 350. The four bristles 350 are mounted on respective corners of a flange 355 the separates the head portion 308 from the anchor portion 310. Each of the four bristles 350 are tapered and terminate in a rounded end configured to make contact with a user's skin. Advantageously providing each bristle 350 on a respective corner of the flange 355 improves the flexibility and degrees of freedom of each Bristol 350, thereby improving contact with a user's skin and also improving comfort for the user. The anchor portion 310 in these examples is provided by a tapered stem portion.

In the example shown, the anchor portion is 15 mm long, the flange 355 3 mm long/thick, and each bristle is 5 mm long.

In other examples the electroencephalography sensors may take other forms. for example, Fig. 5A shows a side view of another example electroencephalography sensor, 30 Fig. 53 shows an end view of the example electroencephalography sensor of Fig. 5A, Fig. 50 shows a plan view of the example electroencephalography sensor of Figs. 5A -12 -and 53 and Fig. 50 shows a perspective view of the example electroencephalography sensor of Figs. 5A to 50. As shown in Figs. 5A to 5D, like with the example shown in Figs. 3A to 4B and as described above, the electroencephalography sensor 500 of Figs. 5A to 5D comprises a head portion 508 and an anchor portion 510. The head portion comprises a plurality of bristles 550 On this example 11, all of an identical shape and size) mounted to a flange portion 555 of the head portion 508. The flange portion 555 of the head portion is gently curved, for example to follow the curvature of an ear pad for use with headphones, such as the ear pad of Figs. 1 and 2 described above. In this example the anchor portion 510, however, is conductive, and may for example be configured to be mounted directly to the ear can 108, 109 of the headphones and/or the headband 105. In the example show the head portion 508 comprises a female receiving portion (a v-shaped groove) for receiving a corresponding v-shaped male protrusion from the anchor portion 510. In this way the head portion 508 may be detachably coupled to the anchor portion 510 and/or different head portions 508 (with, for example different arrangements, sizes and/or configurations of bristles) coupled to the anchor portion 510.

Preferably the minimum number of bristles 350, 550 is three in order to ensure that they end up cutting through hair, we found that too few bristles can end up aligning and resting on hair but having a multi prong approach with bristles on different corners of a shape with at least three sides has the required degrees of freedom to ensure scalp contact on different people and configurations. There may be a range of bristles 350, 550 with varying lengths to cut through hair. At the moment the most universally functional solution is five bristles 350, 550 with 5mm length. But with shorter hair there may be shorter bristles 350, 550 and longer hair longer bristles 350, 550. In both circumstances there may be a larger number of bristles 350, for example between five and fifteen bristles 350. The sensors 104A, 104B, 106A, 106B, 107 may be modular to allow for interchanging sensors/head portions 308, 508 to combat the variety of hairstyles. The main rule is that the sensor material needs to touch the scalp and the bristles 350, 550 are designed to be as short and as comfy as possible whilst still maintaining good contact.

-13 -In the examples shown the electroencephalography sensors 106A, 500 are made from a flexible and resiliently deformable support scaffold with an electrically conductive coating. In the case of the example shown in Figs. 5A to 5D, only the head portion 508 may be made from a flexible and resiliently deformable support scaffold with an electrically 5 conductive coating as the base portion 510 may be made from a conductive material, for example a metal such as copper. The support scaffold may comprise silicone or rubber. For example, the support scaffold may consist of silicone or rubber. In some examples there may not be a support scaffold, however, and the entire electroencephalography sensor may be made from the same electrically conductive material as the coating which 10 may be a polymer. The electrically conductive coating may comprise, for example, a silicone and graphite mix. This may be made in two stages.

Stage 1: The material is mixed and needs to be molded into a suitable shape to acquire data from the bristles. The issue we had to overcome that in order for the material to dry and "cure" the process requires far too long for useable manufacturing with standard injection molding. To address this a soft flexible scaffold can be rubber or silicon etc as a base that is easy to mold and cures rapidly in order to create the shape needed.

The silicone core (support scaffold) is created by mixing 2-part mix of silicone and catalyst hardener. When mixed to a ratio of 1:1, a liquid is created which turns into silicone rubber once cured for -24 hours.

This liquid is then poured into our custom designed 2-part mold, 3D printed from SLS 25 Nylon for easy removal once the silicone has cured. Once cured, the mold is broken to reveal the final form of the electrode core. Any excess silicone is trimmed if needed.

Stage 2: To make the electrode conductive, a conductive coating comprised of silicone, graphite 30 and Naptha solvent is applied to the silicone core (support scaffold). Precisely mixing these elements by weight (see sample number 5 in the table below) produces a -14 -conductive paste which is then brushed onto the silicone core (support scaffold). It must be brushed, and not applied any other way (e.g., dipped or sprayed) as this brushing action helps to align the graphite particles such that they physically contact each other, ensuring maximum overall conductivity of the electrode. However, it should be noted that in some examples another conductive additive may be used instead of the graphite, for example carbon nanotubes, in which case brushing may not be needed.

To ensure best conductivity, the paste should be applied evenly and in generous amounts, to achieve a layer of conductive coating that's at least 0.5mm and preferably between 0.8-1mm in thickness. After allowing the conductive paste to cure for three to five hours, the coating thickness is measured with digital measurement calipers and a second coating added if below the 0.8-1.0mm threshold.

The composition of the electrically conductive coating is, in equal proportions, graphite 15 powder, petroleum ether (naphtha) and silicone.

Method 1 Combine the graphite powder and petroleum ether and stir to ensure the graphite powder is evenly distributed throughout the mixture.

2 Add the silicone and stir to ensure that it is dissolved fully -the viscosity of the silicone will "spread" throughout the mixture as this occurs.

3 While the mixture is liquid, apply it to the surface of the sensors as needed as evenly as possible.

4 Allow the mixture to fully set before using.

In some examples the composition may comprise graphene additives and a revised silicone blend to both increase the durability and manufacturability of the compound and make it more conductive.

Example of one additive is graphene powder: * Bulk Density -0.24 gfcc -15 - * Thickness -5-10 nm * Diameter -5- 10 micron * Carbon Purity -> 99% * Surface Area -150 sq. Meter per gram Example of another additive is a graphene nanotube formulation: TUBALL MATRIX 602, a graphene nanotube-formulated super-concentrate based on crosslinking carrier for liquid silicone rubbers (LSR), room temperature vulcanised rubber (RTV) and high consistency rubbers (HCR), was specifically developed to improve nanotube usability by providing a fine dispersion of nanotubes in the host matrix while maintaining softness. TUBALLTm MATRIX is a line of concentrates based on TUBALLTm single wall carbon nanotubes (SWCNT) produced by OCSiAl. TUBALLTm is a unique SWCNT additive that provides electrical conductivity at low dosages not achievable with any standard conductive additive.

Table 1 Properties of example compositions for conductive coating Sample number Silicone type Silicone (9) Graphite (9) Naptha (9) Resistance Comments (20k) #1 Two part 10 2 0 Infinite Not enough graphite.

#2 Two part 10 3 0 Infinite Not enough graphite.

#3 Two part 10 4 0 Infinite Not enough graphite.

#4 Two part 10 5 0 3-4 High graphite content required, but results in a very thick mix.

#5 Tube 5 5 5 0.2-0.5 Naptha helps to evenly distribute graphite, but evaporates which causes significant shrinkage in the molded part.

#6 Tube 4 1 0 3-5 Very variable between different probe locations Moulding tests Conductive thread Two part 0.05 Molding conductive thread with silicone is very difficult. Sewing thread into a protruding shape -16 - #3 Clean mouldings Two part 10 4 0 Infinite #5 Squished mouldings Tube 5 5 5 0.2-0.5 Highly conductive but molding has air bubbles. Can be solved with SLA mould and using syringe to inject material.

#6 Small moulding Tube 4 1 0 3-5 Very variable between different probe locations Pogo pins Two part 1 0 Metal pogo pins are not comfortable and would require a much harder silicone to securely hold them in place Fig. 6 shows an example process flow chart of a method of manufacturing an example electroencephalography sensor such as the example electroencephalography sensor of any of Figs. 3A to 5D. The method comprises the steps of forming 610 a resiliently deformable support scaffold from silicone or rubber and applying (e.g., by brushing) 620 a layer of electrically conductive coating onto the support scaffold.

Fig. 7 shows an example process flow chart of a method 700 of removing noise from an electroencephalography signal. Receiving 710 signals indicative of electroencephalography signals from a sensor coupled to a user, receiving 720 an indication of a parameter of the user, and processing 730 the electroencephalography signal to denoise the electroencephalography signal based on the indication of a parameter of the user. Optionally the method comprises performing independent component analysis to denoise the electroencephalography signal based on the indication of a parameter of the user.

The electroencephalography sensors described herein are configured to monitor neural activity in real time and may use a combination of sensors and denoising algorithms to remove noise and provide a high-fidelity signal. Machine learning algorithms may be applied to the signals to classify them into emotional states to identify the user's emotions. This may then be used, for example, to enhance music -17 -suggestion, provide data and analytics to users and enable enhanced content experiences via an api platform with partners.

The advanced denoising method may enable users to monitor emotions in real time from a small form factor with relatively few points of measure with high fidelity. Headphones are portable and embed into users existing habits. The form factor of headphones means that capturing electroencephalography data doesn't face the same issues as voice or facial expression recognition.

In some examples, devices comprising the electroencephalography sensors (such as headphones, glasses, helmets etc.) may comprise processing circuitry. The processing circuitry may perform some local processing of the physiological signals. In some examples the processing circuitry comprises a wireless interface, and the processing circuitry is configured to receive sensor signals from the electroencephalography sensors and send them (optionally with a degree of local processing) to a remote device via the wireless interface (e.g., for further processing by the remote device). The processing circuitry may further comprise an amplifier for amplifying the signals before they are transmitted via the wireless interface.

The processing circuitry of the headphones 100 described above may be configured to obtain electroencephalography data from the electroencephalography sensors in what is commonly referred to as a monopolar or bipolar mode. If a bipolar mode is used the potential difference between two sensors is measured at all times. In the example shown in Figs. 1 and 2, five sensors are used, two pairs for signal acquisition and one sensor for reference. The sensors are composed of the sensing material (described in the table above as material #5) which is directly connected to a gold-plated sensing board with a unity gain amplifier in it. In some examples up to four pairs of sensors may be used together with one reference. Active sensors may be needed to make it possible to use dry electrodes as they have enough gain to not lose any signal.

-18 -The signal amplification is controlled by a ADS1299 chip in a logic board forming part of the local processing circuitry. The chip performs analogue-to-digital conversion at 250 Hertz (Samples per Second) and has a programmable gain between 1, 2, 4, 6, 8, 12, or 24. In the future this can be substituted with what are commonly referred to as DSP chips, higher performance chipsets with the ability to perform digital operations at almost instantaneous speeds. This will enable: * The device to be operated with dynamic gain: modify the amplification of the system depending on inputs from other sensors to make sure that the least amount of noise is picked up, * The data flow to be controlled by automatically turning off/on when the user is wearing the device. The device-on recognition can be performed by calculating parameters of the incoming signal on the fly. This can be achieved with a DSP chip, for example.

In the example shown the processing circuitry does not perform any on-device filtering, although in some examples baseline filtering may be performed on-device using combinations of amplifiers.

In some examples the apparatus further comprises other sensors configured to obtain an indication of a parameter(s) of a user, such as heart rate, motion, sweat, skin conductance. For example, in some examples the apparatus further comprises a heart rate sensor such as a photoplethysmography, PPG, sensor. The PPG sensor may be either embedded in the ear pad or configured to obtain a signal from a user from inside a region bounded by the earpad, for example a region barely touching the earlobe. The processing circuitry (either locally or remotely) may be configured to remove artefacts from the physiological signals (e.g., electroencephalography signals) using other signals providing an indication of a parameter(s). For example, the processing circuitry (either locally or remotely) may be configured to perform independent component analysis (ICA) to remove noise. For example, the apparatus may comprise a PPG sensor and/or an inertial measurement unit and wherein the processing circuitry is configured to remove motion artefact noise and/or heart rate noise caused by the user and/or muscle -19 -movement. As described in more detail below, the processing circuitry may comprise one or more adaptive or digital filters for removing noise, for example the adaptive or digital filter may be configured to reduce mean square error of a signal.

For example, a PPG sensor may be placed either on the headphones ear pad or in the inside just barely touching the earlobe. The device may record PPG signal (such as heart rate, HR) which may be used for denoising the EEG data, useful in removing the HR artefact present in the signal using ICA methods (as described in more detail below) and/or a feature of the classification models (i.e., used to classify a user's mood), where the HR is a physiological parameter which contains information about a user's mood. Sensors could be placed on either side of the headphones, but preferably they are touching a skin portion where there is good blood flow.

In some examples the sensor may comprise an Inertial Measurement Unit. This may be a 6-axis or 9-axis sensor that measures any of: orientation, velocity, and gravitational forces by combining an accelerometer, gyroscope, and/or magnetometer into one. This may be used to measure the movement of the head (e.g., headphones, glasses) while the user is wearing them. This may be used for removal of motion artefact noise, which may be caused by movements of the user and/or muscle movement. By recording the movement continuously, components of the EEG signal correlated with movement may be isolated and hence removed. This technique may require an IMU sensor together with ICA (described below).

In some examples a microphone sensor may be used. This may be placed, for example, 25 inside the headphone housing, for example in one of the left or right cans 108, 109. The recording from this microphone may be used to eliminate jaw movement artefacts.

Another type of noise intrinsically present in EEG recordings is the blinking artefact. In order to remove this, in calibration of data, users may be taken through an exercise 30 where they blink to a predefined "beep" sound which has a certain pattern. Since this pattern is known, it can be used to analyse bits of the signal where we a blink is known to -20 -have occurred and algorithms such as a machine learning algorithm can be used to learn to identify them.

In signal processing, independent component analysis (ICA) is a computational method for separating a multivariate signal into additive subcomponents. This is done by assuming that the subcomponents are, potentially, non-Gaussian signals and that they are statistically independent from each other. In practise, given a collection of n signals, where n is a natural number (n > 0), ICA will attempt to extract n components that it deems to be most independent.

The assumption that a signal is composed of different components which can be "statistically independent" is very important when processing EEG signals, as the noise signals present in the data can be assumed to be uncorrelated (a person's movement is fairly random in a general setting hence so will the noise be). Hence, ICA can be used (for example by processing circuitry) to split the signal into different and uncorrelated components. Once the components are identified, it can be determined which of the components are noise and which contain information (data) about emotions. Since ICA does not give any indication on what each component means with respect to the signal, a method may be used to identify the noise components from the data components.

In order to remove more predictable sources of noise, such as motion and heart rate, adaptive filters may be used. Adaptive filters are digital filters whose coefficients change with an objective to make the filter converge to an optimal state. The optimization criterion is a cost function, which is most commonly the mean square of the error signal between the output of the adaptive filter and the desired signal. As the filter adapts its coefficients, the mean square error (MSE) converges to its minimal value. At this state, the filter is adapted and the coefficients have converged to a solution. The filter output is then said to match very closely to the desired signal. When the input data characteristics change, sometimes called filter environment, the filter adapts to the new environment by generating a new set of coefficients for the new data. -21 -

In order to adapt these filters to the present application, a filter may be designed such that it minimizes the error between our input signal and one of the forms of noise (motion artefact data, heart rate, etc...). What the filter should ideally then output is a prediction of the isolated noise component which can be then removed from the data via a negative operation. This method may be designed to run in cascade for all predictable forms of noise, yielding ultimately a signal which is clean of those components in question.

Fig. 8 shows a schematic of example processing circuitry that may be used, for example, with the headset of Figs. 1 and 2. The headset 100 comprises speakers 855 coupled to 10 a Bluetooth BM83 audio module 853. The BT audio module 853 is coupled to an antenna 851 which may be operable to communicate with a corresponding antenna 801 of a smartphone 800. The headset 100 also comprises sensing pads (electroencephalography sensors) 106A, 106B, as well as optionally other sensors to obtain an indication of at least one other physiological parameter (such as heart rate, movement, blinking, jaw movement etc.). The sensing pads 106A, 160B are coupled to an application microcontroller unit (MCU) 857 which is operable to communicate with the BT audio module 853 via UART. The smartphone 800 is configured to communicate with the headset 100 via a wireless interface On this example Bluetooth) and control operation of the headset 100. Moreover, the smartphone 800 may be configured to act as a remote device to process signals received from the sensing pads 106A, 106B (and any additional optional sensors). However, in other examples the smartphone 800 may merely act as an intermediary and forward (e.g. via a wireless telecommunications network such as 3G, 4G or 5G) any signals received from the sensing pads 106A, 106B (and any additional optional sensors) to a remote device/server, such as the cloud, for processing.

The application MCU 857 is the MCU running the main application, hosting the business logic of connection management, signal acquisition and processing from the sensing pads 16A, 106B and user interface operating on the smartphone 800 The BM83 is a Bluetooth EDR/Smart module that contains a radio and a small -22 -microcontroller that manages the low-level details of the radio link. It also has an analogue audio interface to the speakers 855 and optional microphones (not shown).

The MCU 857 and the BM83 853 communicate using a command/event protocol over 5 UART, in which the MCU 857 is generally the master and the 3M83 853 is the slave.

The radio link between the BM83 853 and the smartphone 800 takes advantage of both Bluetooth EDR and Bluetooth Smart (a.k.a BLE) for different purposes. Bluetooth EDR is used for different types of audio and media control, specifically the headset 100 uses: * A2DP for high quality audio * AVRCP for media control (play/stop/volume...) * HFP for voice calls and call management Bluetooth Smart is employed purely for the management of the sensing part, the transfer of data from them and other ancillary functions. A custom BLE service providing a serial-15 line emulation is used as a transport layer.

On first power-up, the MCU 857 manages the interaction between the BM83 853 and the external device On the example above, a smartphone 800) to advertise until the smartphone 800 initiates a Bluetooth pairing procedure. Once the standard pairing process is completed, a Pairing Record is stored in the BM83 853, and the encryption keys stored in it can be used to establish an encrypted connection with the same device until either device erases them.

As described above, there may need to be a calibration process when sensors are first worn by a user for example to calibrate the algorithms to fit the user's brain pattern. This is needed so the features extracted from the users can be normalized by this calibration and subsequently will be in the same domain as the rest of the features allowing for inference to be performed. Calibration may be performed in a number of different ways and a user might sit through one or multiple of them. For example, different calibration techniques may comprise the user: * Sitting still and focussing on a black point in the middle of a white screen for 30-60 -23 -seconds.

* Sitting still and going through an emotion evocation experience using a video/game and respond to a questionnaire after to assess mood state.

It will be appreciated from the discussion above that the embodiments shown in the Figures are merely exemplary, and include features which may be generalised, removed or replaced as described herein and as set out in the claims. In the context of the present disclosure other examples and variations of the apparatus and methods described herein will be apparent to a person of skill in the art.

Claims (25)

1. -24 -CLAIMS: 1. An apparatus for obtaining physiological electrical signals of a living being, the apparatus comprising headphones or a headphone insert comprising at least one 5 deformable ear pad configured in use to be biased against the user's head around one of the user's ears, wherein the deformable ear pad comprises at least three electroencephalography sensors configured to receive electrical signals from the user, wherein in use at least one of the electroencephalography sensors is located above the user's ear, and at least one of the electroencephalography sensors is located below the 10 user's ear, and wherein a third electroencephalography sensor is configured to act as a reference.
2. 2. The apparatus of claim 1 wherein the electroencephalography sensors are at least one of conductive or capacitive.
3. 3. The apparatus of claim 1 or 2 wherein the electroencephalography sensors are resiliently deformable.
4. 4. The apparatus of any of the previous claims wherein the electroencephalography sensors are at least partially embedded into the at least one ear pad.
5. 5. The apparatus of any of the previous claims comprising a pair of ear pads, one ear pad of the pair of ear pads comprising at least three electroencephalography sensors, and one ear pad of the pair of ear pads comprising at least two 25 electroencephalography sensors.
6. 6. The apparatus of any of the previous claims wherein the electroencephalography sensors comprise an anchor portion and an interchangeable head portion.
7. 7. The apparatus of claim 6 wherein the head portion comprises a plurality of bristles configured to contact a user's scalp through their hair.
8. -25 - 8. The apparatus of claim 6 or 7 wherein the head portion comprises at least 3 bristles, each bristle on a respective corner or edge of the head portion.
9. 9. The apparatus of claim 7 or 8 wherein the plurality of bristles vary in length.
10. 10. The apparatus of any of claims 6 to 9 wherein the interchangeable head portion of the electroencephalography sensors comprise a resiliently deformable support scaffold coated in an electrically conductive coating.
11. 11. The apparatus of claim 10 wherein the conductive coating comprises a blend of graphite, silicone and petroleum ether.
12. 12. The apparatus of claim 11 wherein the graphite, silicone and petroleum ether are 15 a blend in the ratio of 1:1:1.
13. 13. The apparatus of any of the previous claims further comprising processing circuitry comprising a wireless interface, wherein the processing circuitry is configured to receive sensor signals from the electroencephalography sensors and send them to a 20 remote device via the wireless interface.
14. 14. The apparatus of claim 13 further comprising an amplifier for amplifying the signals before they are transmitted via the wireless interface.
15. 15. The apparatus of claim 13 or 14 wherein the processing circuitry is configured to perform active signal acquisition using bipolar sensors.
16. 16. The apparatus of any of the previous claims further comprising a photoplethysmography, PPG, sensor and wherein the processing circuitry is configured 30 to remove head rate artifact using signals from the PPG sensor.-26 -
17. 17. The apparatus of any of claims 13 to 16 further comprising an inertial measurement unit and wherein the processing circuitry is configured to remove motion artefact noise caused by the user and/or muscle movement
18. 18. A computer-implemented method of removing noise from an electroencephalography, EEG, signal, the method comprising: receiving signals indicative of electroencephalography signals from a sensor coupled to a user; receiving an indication of a parameter of the user; and processing the electroencephalography signal to denoise the electroencephalography signal based on the indication of a parameter of the user.
19. 19. An electroencephalography sensor comprising a resiliently deformable support scaffold coated in a conductive layer, wherein the resiliently deformable support scaffold 15 provides a plurality of bristles for contacting a user's scalp.
20. 20. The electroencephalography sensor of claim 19 wherein the conductive layer comprises a blend of graphite or graphene, silicone and petroleum ether, and wherein the resiliently deformable support scaffold comprises at least one of silicone and rubber..
21. 21. The electroencephalography sensor of any of claims 18 to 20 wherein the conductive layer is at least 0.5mm thick.
22. 22. Headphones comprising the electroencephalography sensor of any of claims 18 25 to 21.
23. 23. Glasses comprising the electroencephalography sensor of any of claims 18 to 21.
24. 24. A method of manufacturing electroencephalography sensors, the method 30 comprising: forming a resiliently deformable support scaffold from silicone or rubber; -27 -brushing a layer of electrically conductive coating onto the support scaffold.
25. 25. The method of claim 24 wherein brushing a layer of electrically conductive coating onto the support scaffold comprises brushing to a layer with a thickness of at least 0.5 5 mm in thickness, wherein the conductive layer comprises a blend of graphite or graphene, silicone and petroleum ether.