Device and method for detecting at least one vital sign

Abstract

A device (20) for detecting at least one vital sign is provided, the device (20) comprising a plurality of emitters (21) that are each configured to emit electromagnetic radiation, a plurality of photodetectors (22) that are each configured to detect electromagnetic radiation from a detection region (23) and to provide sensor data, and a calculation unit (24), wherein the calculation unit (24) is configured to determine at least one vital sign, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region (23) on the measurement of the vital sign. Furthermore, a method for detecting at least one vital sign and a method for training a machine learning algorithm are provided.

Description

[0001] 2024P01359WO October 22 , 2025

[0002] P2024 , 0916 WO N

[0003] - 1 -

[0004] Description

[0005] DEVICE AND METHOD FOR DETECTING AT LEAST ONE VITAL S IGN

[0006] Vital signs such as heart rate and blood pressure can be detected using a wearable device . This has the advantage that the vital signs can be measured continuously . However, movements of the wearable device relative to the skin of the user or movements within the body of the user have an impact on the measurement of the vital signs . These so-called motion artefacts can lead to wrong measurements of the vital signs .

[0007] It is an obj ective to provide a device for detecting at least one vital sign with an improved accuracy . It is further an obj ective to provide a method for detecting at least one vital sign with an improved accuracy . It is further an obj ective to provide a method for training a machine learning algorithm that enables to detect at least one vital sign with an improved accuracy .

[0008] These obj ectives are achieved by the subj ect matter of the independent claims . Further developments and embodiments are described in dependent claims .

[0009] According to at least one embodiment of the device for detecting at least one vital sign, the device comprises a plurality of emitters that are each configured to emit electromagnetic radiation . The device can comprise at least ten emitters , at least 100 emitters or at least 1000 emitters . Thus , the device can comprise an array of emitters . Each emitter can comprise a light-emitting diode ( LED) . 2024P01359WO October 22 , 2025

[0010] P2024 , 0916 WO N

[0011] - 2 -

[0012] According to at least one embodiment of the device , the device comprises a plurality of photodetectors that are each configured to detect electromagnetic radiation from a detection region and to provide sensor data . The device can comprise at least ten photodetectors , at least 100 photodetectors or at least 1000 photodetectors . Thus , the device can comprise an array of photodetectors . Each photodetector can be configured to detect electromagnetic radiation that comes from the detection region . The detection region can be a volume . The detection region can be arranged adj acent to the device . The detection region can be the region that the photodetectors are sensitive to . This means , the detection region can be the region from which the photodetectors can detect electromagnetic radiation . Each photodetector can be configured to provide sensor data . The sensor data can be based on electromagnetic radiation detected by the respective photodetector . Each photodetector can comprise a polari zation filter and / or a spectral filter .

[0013] According to at least one embodiment of the device , the device comprises a calculation unit , wherein the calculation unit is configured to determine at least one vital sign, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region on the measurement of the vital sign . Motion in this case does not include the motion of blood or electrons . The calculation unit can comprise a processor or a chip . It is also possible that the calculation unit comprises a connection, such as a network connection, to a processor . The calculation unit can be configured to receive the sensor data . The calculation unit can be configured to determine at least one vital sign from 2024P01359WO October 22 , 2025

[0014] P2024 , 0916 WO N

[0015] 3 the sensor data . The calculation unit can be configured to determine at least one vital sign from the sensor data of at least some of the photodetectors . A vital sign can be for example blood pressure , heart rate , respiratory rate or blood oxygen concentration .

[0016] For determining the vital sign, the machine learning algorithm is employed . Motion of for example tissue within the detection region has an impact on the sensor data . In most cases , a vital sign can only be measured correctly i f there is no motion in the detection region . Therefore , the machine learning algorithm has been trained to determine at least one vital sign in such a way that the impact of motion within the detection region of the measurement of the vital sign is reduced . In an ideal case , the impact of motion within the detection region can be completely removed for determining the vital sign . However, in most cases the impact of motion within the detection region can be signi ficantly reduced but not removed completely . The machine learning algorithm can be trained to entangle the contribution of motion within the detection area and the vital sign from the sensor data . In this way, the vital sign can be determined with a reduced impact of motion within the detection area .

[0017] Since the device comprises a plurality of emitters and a plurality of photodetectors , the machine learning algorithm can base the measurement of the vital sign on sensor data of only some of the photodetectors i f there is motion in the detection region . The photodetectors are distributed over a certain area . Therefore , in most cases motion is only present in a part of the detection region . The machine learning algorithm can thus be provided with sensor data coming from parts of the detection region where there is no motion and 2024P01359WO October 22 , 2025

[0018] P2024 , 0916 WO N

[0019] - 4 - with sensor data coming from parts of the detection region where there is motion .

[0020] The large number of emitters and photodetectors also enables that the device combines several functionalities . For this purpose , the photodetectors or at least some of the photodetectors can comprise a polari zation filter and / or a spectral filter .

[0021] According to at least one embodiment of the device for detecting at least one vital sign, the device comprises a plurality of emitters that are each configured to emit electromagnetic radiation, a plurality of photodetectors that are each configured to detect electromagnetic radiation from a detection region and to provide sensor data, and a calculation unit , wherein the calculation unit is configured to determine at least one vital sign, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region on the measurement of the vital sign .

[0022] The device can be employed for a measurement of at least one vital sign with a reduced impact of motion within the detection region on the measurement . For this purpose , the machine learning algorithm is employed to retrieve the vital sign from the sensor data . Since the device comprises a plurality of emitters , a large region can be illuminated . Therefore , the detection region from which reflected electromagnetic radiation can be detected by the photodetectors is also larger . Moreover, the device comprises a plurality of photodetectors such that electromagnetic radiation from the larger detection region can be detected . 2024P01359WO October 22 , 2025

[0023] P2024 , 0916 WO N

[0024] - 5 -

[0025] This setup enables to reduce motion artefacts . I f the detection region is larger, it is more likely that motion does not occur in the whole detection region but only in parts of the detection region . A plurality of emitters and photodetectors leads to redundancies in the device which increases the probability that there are parts of the detection region where there is no motion . In this way, the accuracy of determining at least one vital sign is increased .

[0026] According to at least one embodiment of the device , the device comprises a display, and the emitters and the photodetectors are arranged at a side of the device facing away from the display . The device can be a wearable device . The emitters and the photodetectors can thus face the skin of the user . The display can be arranged at the opposite side in comparison to the emitters and the photodetectors . Thus , the display can be viewed by the user . Incorporating the emitters and the photodetectors in a device with a display has the advantage that the device is used and worn by the user anyways and can furthermore be employed for determining the at least one vital sign .

[0027] According to at least one embodiment of the device , the device comprises a plurality of pixels , wherein each pixel comprises three of the emitters and one of the photodetectors . The device can comprise at least ten pixels , at least 100 pixels or at least 1000 pixels . The pixels can all have the same setup . The device can comprise an array of pixels . The three emitters of each pixel can be a first emitter, a second emitter and a third emitter . This can mean, that each pixel comprises one first emitter, one second emitter and one third emitter . The first emitter can be configured to emit electromagnetic radiation of a first 2024P01359WO October 22 , 2025

[0028] P2024 , 0916 WO N

[0029] 6 wavelength . The second emitter can be configured to emit electromagnetic radiation of a second wavelength . The third emitter can be configured to emit electromagnetic radiation of a third wavelength . The first wavelength, the second wavelength and the third wavelength can all be di f ferent from each other . This setup with the plurality of pixels of the device has the advantage that each pixel can either be employed for the emission of electromagnetic radiation or the detection of electromagnetic radiation . Therefore , the arrangement of emitters and photodetectors within the device is configurable . It is possible to choose which pixels are employed as emitters and which pixels are employed as photodetectors . This enables a high degree of flexibility for the device .

[0030] According to at least one embodiment of the device , at least some of the emitters are configured to emit electromagnetic radiation of a first wavelength and at least some of the emitters are configured to emit electromagnetic radiation of a second wavelength, wherein the first wavelength is di f ferent from the second wavelength . It is also possible that at least some of the emitters are configured to emit electromagnetic radiation of a third wavelength, wherein the first wavelength, the second wavelength and the third wavelength are all di f ferent from each other . Employing emitters of di f ferent wavelengths has the advantage that di f ferent penetration depths of the electromagnetic radiation are enabled . This enlarges the detection region which makes it more likely that there are parts within the detection region where there is no motion during the measurement of the vital sign . With this , the accuracy of determining the at least one vital sign is improved . 2024P01359WO October 22 , 2025

[0031] P2024 , 0916 WO N

[0032] 7

[0033] According to at least one embodiment of the device , the si ze of each of the emitters is larger than the si ze of each of the photodetectors . This can mean, that the resolution of the photodetectors is higher than the resolution of the emitters . In this way, the probability is increased that at least some of the photodetectors detect sensor data from regions without motion which improves the accuracy of determining the at least one vital sign .

[0034] According to at least one embodiment of the device , the photodetectors are arranged along a grid . The photodetectors can be arranged along the lines of a grid . The grid can have a square structure . This can mean, that the lines of the grids form squares . The photodetectors can be arranged around the squares . This enables , that the photodetectors are arranged over a large area and that the detection region is large enough so that motion only occurs in parts of the detection region but not in the whole detection region . This enables the machine learning algorithm to retrieve the vital sign from sensor data coming from a part of the detection region without motion .

[0035] According to at least one embodiment of the device , the emitters and the photodetectors are arranged in a two- dimensional array . The emitters and the photodetectors can be arranged alternately in the two-dimensional array . The array of emitters and photodetectors can have a lateral extent along at least one direction of at least 1 cm and 4 cm at most , wherein the lateral extent is given in a plane that extends perpendicular to the main plane of extension of the device . The array of emitters and photodetectors can, for example , have a si ze of 2 x 2 cm2. This arrangement enables a 2024P01359WO October 22 , 2025

[0036] P2024 , 0916 WO N

[0037] - 8 - large detection region which makes it possible to reduce motion artefacts .

[0038] According to at least one embodiment of the device , the calculation unit is configured to average sensor data . The calculation unit can be configured to average sensor data coming from di f ferent parts of the detection region . In this way, sensor data coming from di f ferent parts of the detection region can be employed to determine the vital sign . This improves the accuracy of determining the vital sign .

[0039] According to at least one embodiment of the device , at least some emitters and photodetectors are arranged spaced apart from each other such that the space between neighboring emitters and / or photodetectors is larger than the si ze of the emitters or photodetectors . This can mean, that the emitters and photodetectors are sparsely distributed . The area over which the emitters and photodetectors are distributed can be larger than the total area of the emitters and photodetectors . The space between neighboring emitters can be larger than the si ze of one emitter or larger than the si ze of one photodetector . The space between neighboring photodetectors can be larger than the si ze of one emitter or larger than the si ze of one photodetector . The space between an emitter and a photodetector that are arranged adj acent to each other can be larger than the si ze of one emitter or larger than the si ze of one photodetector . This distribution of the emitters and the photodetectors enables a large detection region . With a large detection region, the accuracy of determining the vital sign can be improved .

[0040] According to at least one embodiment of the device , each emitter has a lateral extent of at least 10 pm and 4 mm at 2024P01359WO October 22 , 2025

[0041] P2024 , 0916 WO N

[0042] 9 most , wherein the lateral extent is given in a plane that extends perpendicular to a main plane of extension of the device . This can mean, that each emitter has a lateral extent along at least one direction of at least 10 pm and 4 mm at most . Each emitter can have a lateral extent of at least 10 pm and 1 mm at most . Each emitter can have a lateral extent of at least 1 mm and 4 mm at most . With the si ze of the emitter of at least 10 pm and 4 mm at most a high resolution of the emitters and a large detection region are enabled .

[0043] According to at least one embodiment of the device , each photodetector has a lateral extent of at least 5 pm and 200 pm at most , wherein the lateral extent is given in a plane that extends perpendicular to a main plane of extension of the device . This can mean, that each photodetector has a lateral extent along at least one direction of at least 5 pm and 200 pm at most . Each photodetector can have a lateral extent of at least 5 pm and 100 pm at most . Each photodetector can have a lateral extent of at least 10 pm and 200 pm at most . With the si ze of the photodetectors of at least 5 pm and 200 pm at most a high resolution of the photodetectors and a large detection region are enabled .

[0044] Furthermore , a method for detecting at least one vital sign is provided . The device for detecting at least one vital sign can preferably be employed for the method described herein . This means all features disclosed for the method for detecting at least one vital sign are also disclosed for the device for detecting at least one vital sign and vice-versa .

[0045] According to at least one embodiment of the method for detecting at least one vital sign, the method comprises 2024P01359WO October 22 , 2025

[0046] P2024 , 0916 WO N

[0047] - 10 - emitting electromagnetic radiation by a plurality of emitters of a device .

[0048] According to at least one embodiment of the method for detecting at least one vital sign, the method comprises detecting electromagnetic radiation by a plurality of photodetectors of the device from a detection region . Each of the photodetectors can be configured to detect electromagnetic radiation coming from a part of the detection region . Thus , di f ferent photodetectors detect electromagnetic radiation from di f ferent parts of the detection region . In this way, electromagnetic radiation from the whole detection region can be detected .

[0049] According to at least one embodiment of the method for detecting at least one vital sign, the method comprises providing sensor data by the photodetectors . This can mean, that each photodetector provides sensor data .

[0050] According to at least one embodiment of the method for detecting at least one vital sign, the method comprises determining at least one vital sign, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region on the measurement of the vital sign .

[0051] According to at least one embodiment of the method for detecting at least one vital sign, the method comprises emitting electromagnetic radiation by a plurality of emitters of a device , detecting electromagnetic radiation by a plurality of photodetectors of the device from a detection region, providing sensor data by the photodetectors , and 2024P01359WO October 22 , 2025

[0052] P2024 , 0916 WO N

[0053] - 11 - determining at least one vital sign, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region on the measurement of the vital sign .

[0054] The method for detecting at least one vital sign has the same advantages as described for the device for detecting at least one vital sign .

[0055] According to at least one embodiment of the method for detecting at least one vital sign, the device comprises a plurality of pixels , wherein each pixel comprises three of the emitters and one of the photodetectors , and for some of the pixels only the emitters are operated and for some of the pixels only the photodetector is operated . This means , some pixels are employed as emitters and some of the pixels are employed as photodetectors . Since all the pixels comprise emitters and a photodetector, it can be chosen freely which pixels are operated as emitters and which pixels are operated as photodetectors . This increases the flexibility . Increased flexibility can for example be employed to optimi ze the reduction of motion artefacts .

[0056] Furthermore , a method for training a machine learning algorithm is provided . The trained machine learning algorithm can preferably be employed for the method and device for detecting at least one vital sign described herein . This means all features disclosed for the method and device for detecting at least one vital sign are also disclosed for the method for training a machine learning algorithm and vice- versa . 2024P01359WO October 22 , 2025

[0057] P2024 , 0916 WO N

[0058] 12

[0059] According to at least one embodiment of the method for training a machine learning algorithm, the method comprises providing first sensor data to the machine learning algorithm, wherein the first sensor data is detected by a plurality of photodetectors in situations where there is no motion in a detection region from which the photodetectors are configured to detect electromagnetic radiation . The photodetectors can be photodetectors of the device described herein . That there is no motion in the detection region can mean, that there is no relative movement between the device and the user and that there is no motion within the detection region of for example tissue , muscles or filaments .

[0060] According to at least one embodiment of the method for training a machine learning algorithm, the method comprises providing second sensor data to the machine learning algorithm, wherein the second sensor data is detected by the plurality of photodetectors in situations where there is motion in the detection region . That there is motion in the detection region can mean, that there is a relative movement between the device and the user or that there is motion within the detection region of example tissue , muscles or filaments .

[0061] According to at least one embodiment of the method for training a machine learning algorithm, the method comprises providing the machine learning algorithm with the information that there is no motion in the detection region during the detection of the first sensor data und that there is motion in the detection region during the detection of the second sensor data . The machine learning algorithm can further be provided with the information that motion artefacts should be reduced or removed . Thus , the machine learning algorithm is 2024P01359WO October 22 , 2025

[0062] P2024 , 0916 WO N

[0063] 13 provided with sensor data with motion artefacts and with sensor data without motion artefacts . The machine learning algorithm can then be trained to provide at least one vital sign that has a low probability of being impacted by motion artefacts .

[0064] The machine learning algorithm can comprise a convolutional neural network . The machine learning algorithm can be trained using supervised learning . The machine learning algorithm can be based on a computer vision approach and / or an integration approach . For the computer vision approach the machine learning algorithm is provided with sensor data from all photodetectors of the device . In contrast , for the integration approach the machine learning algorithm is provided with averaged sensor data, where the sensor data is averaged for several photodetectors . Thus , the computer vision approach leads to a higher resolution and an improved reduction of motion artefacts in comparison to the integration approach . The integration approach can be wav2vec based .

[0065] The machine learning algorithm can be executed in the calculation unit , for example on a chip ( so-called edge Al ) . Alternatively or additionally, the machine learning algorithm can be executed on a server or a cloud server that is connected with the calculation unit via a network connection .

[0066] According to at least one embodiment of the method for training a machine learning algorithm, the method comprises providing first sensor data to the machine learning algorithm, wherein the first sensor data is detected by a plurality of photodetectors in situations where there is no motion in a detection region from which the photodetectors 2024P01359WO October 22 , 2025

[0067] P2024 , 0916 WO N

[0068] 14 are configured to detect electromagnetic radiation, providing second sensor data to the machine learning algorithm, wherein the second sensor data is detected by a plurality of photodetectors in situations where there is motion in the detection region, and providing the machine learning algorithm with the information that there is no motion in the detection region during the detection of the first sensor data und that there is motion in the detection region during the detection of the second sensor data .

[0069] The method for training a machine learning algorithm advantageously enables to train a machine learning algorithm that can be employed in the device and method for detecting at least one vital sign described herein .

[0070] Furthermore , a computer program product is provided which comprises instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method for training a machine learning algorithm described herein .

[0071] According to at least one embodiment of the device for detecting at least one vital sign, the device comprises at least ten emitters that are each configured to emit electromagnetic radiation .

[0072] According to at least one embodiment of the device , the device comprises at least ten photodetectors that are each configured to detect electromagnetic radiation from a detection region and to provide sensor data .

[0073] According to at least one embodiment of the device , the device comprises a display . 2024P01359WO October 22 , 2025

[0074] P2024 , 0916 WO N

[0075] - 15 -

[0076] According to at least one embodiment of the device , the emitters and the photodetectors are arranged at a side of the device that faces away from the display .

[0077] According to at least one embodiment of the device , the device is a wearable device . A wearable device is for example a watch, a smartphone , a mobile device , ear plugs , glasses or a tablet computer . The wearable device can be worn by a user . The wearable device can be configured to be fixed to a user and / or the wearable device can be configured to be held by a user .

[0078] According to at least one embodiment of the device for detecting at least one vital sign, the device comprises at least ten emitters that are each configured to emit electromagnetic radiation, at least ten photodetectors that are each configured to detect electromagnetic radiation from a detection region and to provide sensor data, and a front side , wherein the emitters and the photodetectors are arranged at a side of the device that faces away from the front side , and the device is a wearable device . At the front side , a display can be arranged . The front side can be a side that faces away from the user when worn by a user . The side of the device where the emitters and the photodetectors are arranged can face a part of the user when worn by the user . The side of the device where the emitters and the photodetectors are arranged can be in direct contact with the user when worn by the user .

[0079] The high number of emitters and photodetectors enables that sensor signals are detected from a large detection region . This enables to reduce the impact of motion on the 2024P01359WO October 22 , 2025

[0080] P2024 , 0916 WO N

[0081] - 16 - measurement of the vital sign as motion might occur only in parts of the detection region .

[0082] The following description of figures may further illustrate and explain exemplary embodiments . Components that are functionally identical or have an identical ef fect are denoted by identical references . Identical or ef fectively identical components might be described only with respect to the figures where they occur first . Their description is not necessarily repeated in successive figures .

[0083] Figure 1 shows an exemplary embodiment of the device for detecting at least one vital sign .

[0084] Figures 2 and 3 show further exemplary embodiments of the device for detecting at least one vital sign .

[0085] Figure 4 shows a part of an exemplary embodiment of the device for detecting at least one vital sign .

[0086] With figure 5 an exemplary embodiment of the method for detecting at least one vital sign is described .

[0087] With figure 6 an exemplary embodiment of the method for training a machine learning algorithm is described .

[0088] Figure 7 shows another exemplary embodiment of the device for detecting at least one vital sign .

[0089] Figure 1 shows an exemplary embodiment of the device 20 for detecting at least one vital sign . The device 20 is a wearable device , in this case , a watch . The device 20 comprises a display 25 . The display 25 is arranged at a side 2024P01359WO October 22 , 2025

[0090] P2024 , 0916 WO N

[0091] 17 of the device 20 that faces away from the arm of the user when worn by the user . The device 20 further comprises a plurality of emitters 21 that are each configured to emit electromagnetic radiation . The device 20 further comprises a plurality of photodetectors 22 that are each configured to detect electromagnetic radiation from a detection region 23 and to provide sensor data . The detection region 23 is shown in figure 4 . The emitters 21 and the photodetectors 22 are arranged at a side of the device 20 that faces away from the display 25 . Thus , the emitters 21 and the photodetectors 22 face the skin of the user when the device 20 is worn .

[0092] The device 20 further comprises a calculation unit 24 . The calculation unit 24 is configured to determine at least one vital sign, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region 23 on the measurement of the vital sign . The calculation unit 24 can be configured to average sensor data . The calculation unit 24 can be arranged next to the emitters 21 and photodetectors 22 as it is shown in figure 1 . It is also possible that the calculation unit 24 is arranged within the device 20 .

[0093] The device 20 comprises a plurality of pixels 26 , wherein each pixel 26 comprises three of the emitters 21 and one of the photodetectors 22 . As an example , nine pixels 26 are shown in figure 1 . However, the device 20 can comprise more than nine pixels 26 . The pixels 26 are arranged in a two- dimensional array .

[0094] At least some of the emitters 21 are configured to emit electromagnetic radiation of a first wavelength and at least some of the emitters 21 are configured to emit 2024P01359WO October 22 , 2025

[0095] P2024 , 0916 WO N

[0096] 18 electromagnetic radiation of a second wavelength, wherein the first wavelength is di f ferent from the second wavelength . The three emitters 21 of each pixel 26 can be a first emitter 27 , a second emitter 28 and a third emitter 29 . This can mean, that each pixel 26 comprises one first emitter 27 , one second emitter 28 and one third emitter 29 . The first emitters 27 can be configured to emit electromagnetic radiation of a first wavelength . The second emitters 28 can be configured to emit electromagnetic radiation of a second wavelength . The third emitters 29 can be configured to emit electromagnetic radiation of a third wavelength . The first wavelength, the second wavelength and the third wavelength can all be di f ferent from each other .

[0097] Each emitter 21 has a lateral extent of at least 10 pm and 4 mm at most , wherein the lateral extent is given in a plane that extends perpendicular to the main plane of extension of the device 20 . Each photodetector 22 has a lateral extent of at least 5 pm and 100 pm at most , wherein the lateral extent is given in a plane that extends perpendicular to the main plane of extension of the device 20 .

[0098] This setup with the plurality of pixels 26 of the device 20 has the advantage that each pixel 26 can either be employed for the emission of electromagnetic radiation or the detection of electromagnetic radiation . Therefore , the arrangement of emitters 21 and photodetectors 22 within the device 20 is configurable . It is possible to choose which pixels 26 are employed as emitters 21 and which pixels 26 are employed as photodetectors 22 . This enables a high degree of flexibility for the device 20 . 2024P01359WO October 22 , 2025

[0099] P2024 , 0916 WO N

[0100] - 19 -

[0101] Figure 2 shows a further exemplary embodiment of the device 20 for detecting at least one vital sign . The only di f ference to the embodiment shown in figure 1 is the arrangement and the si ze of the emitters 21 , photodetectors 22 and the calculation unit 24 . The photodetectors 22 are arranged along lines of a grid . The emitters 21 are arranged on the squares formed by the grid of the photodetectors 22 . As an example , nine emitters 21 are shown in figure 2 . However, the device 20 can comprise more than nine emitters 21 . The photodetectors 22 are not shown separately . A plurality of photodetectors 22 is arranged along the lines of the grid between the emitters 21 . The photodetectors 22 are signi ficantly smaller than the emitters 21 . Therefore , the photodetectors 22 are not shown separately . Thus , the si ze of each of the emitters 21 is larger than the si ze of each of the photodetectors 22 . Each emitter 21 has a lateral extent of at least 1 mm and 4 mm at most , wherein the lateral extent is given in a plane that extends perpendicular to the main plane of extension of the device 20 . Each photodetector 22 has a lateral extent of at least 5 pm and 100 pm at most , wherein the lateral extent is given in a plane that extends perpendicular to the main plane of extension of the device 20 .

[0102] The calculation unit 24 is arranged within the device 20 .

[0103] At least some of the emitters 21 are configured to emit electromagnetic radiation of a first wavelength and at least some of the emitters 21 are configured to emit electromagnetic radiation of a second wavelength, wherein the first wavelength is di f ferent from the second wavelength . 2024P01359WO October 22 , 2025

[0104] P2024 , 0916 WO N

[0105] 20

[0106] The setup of figure 2 has the advantage that a high resolution of the photodetectors 22 is achieved . At the same time , the emitters 21 and the photodetectors 22 are distributed over a large area such that the detection region 23 is large .

[0107] Figure 3 shows a further exemplary embodiment of the device 20 for detecting at least one vital sign . The only di f ference to the device 20 of figure 2 is the arrangement and the si ze of the emitters 21 and the photodetectors 22 . The emitters 21 and the photodetectors 22 are distributed more sparsely in the setup of figure 3 in comparison to the setup of figure 2 . The emitters 21 are arranged in the center of the device 20 and the photodetectors 22 are arranged at discrete locations around the emitters 21 . At least some emitters 21 and photodetectors 22 are arranged spaced apart from each other such that the space between neighboring emitters 21 and / or photodetectors 22 is larger than the si ze of the emitters 21 or photodetectors 22 .

[0108] Each emitter 21 has a lateral extent of at least 10 pm and 1 mm at most , wherein the lateral extent is given in a plane that extends perpendicular to the main plane of extension of the device 20 . Each photodetector 22 has a lateral extent of at least 10 pm and 200 pm at most , wherein the lateral extent is given in a plane that extends perpendicular to the main plane of extension of the device 20 .

[0109] With this setup, less emitters 21 and photodetectors 22 are required than for the setup of figure 2 . This reduces complexity and cost . At the same time , a large detection region 23 is formed due to the sparse distribution of the emitters 21 and photodetectors 22 . 2024P01359WO October 22 , 2025

[0110] P2024 , 0916 WO N

[0111] - 21 -

[0112] Figure 4 shows a part of an exemplary embodiment of the device 20 for detecting at least one vital sign . With figure 4 , the principle of detecting at least one vital sign is shown . As an example , three emitters 21 and three photodetectors 22 of the device 20 are shown . The device 20 is placed on the skin 30 of the user . Thus , the emitters 21 and the photodetectors 22 are in direct contact with the skin 30 of the user . The three emitters 21 are a first emitter 27 , a second emitter 28 and a third emitter 29 . The first emitter 27 is configured to emit electromagnetic radiation of a first wavelength . The second emitter 28 is configured to emit electromagnetic radiation of a second wavelength . The third emitter 29 is configured to emit electromagnetic radiation of a third wavelength . The first wavelength, the second wavelength and the third wavelength are all di f ferent from each other . Therefore , the electromagnetic radiation emitted by the three emitters 21 has di f ferent penetration depths into the skin 30 of the user . Therefore , the three photodetectors 22 are arranged at di f ferent distances to the emitters 21 in order to detect electromagnetic radiation reflected at di f ferent depths . In this way, the absorption of the blood can be determined which enables to determine at least one vital sign as for example the heart rate .

[0113] Figure 4 shows that the distance between the emitter 21 and the photodetector 22 , and the wavelength of the emitted electromagnetic radiation determines the detection depth of the sensor signal . By employing di f ferent emission wavelengths and di f ferent distances between the emitters 21 and photodetectors 22 , the detection region 23 can be enlarged . The detection region 23 is the region within the skin 30 of the user from which reflected electromagnetic radiation can be detected by the photodetectors 22 . 2024P01359WO October 22 , 2025

[0114] P2024 , 0916 WO N

[0115] 22

[0116] With figure 5 an exemplary embodiment of the method for detecting at least one vital sign is described . In a first step S I electromagnetic radiation is emitted by the plurality of emitters 21 of the device 20 . In a second step S2 electromagnetic radiation is detected by the plurality of photodetectors 22 of the device 20 from the detection region 23 . In a third step S3 sensor data is provided by the photodetectors 22 . In a fourth step S4 at least one vital sign is determined, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region 23 on the measurement of the vital sign .

[0117] The device 20 can comprise a plurality of pixels 26 , wherein each pixel 26 comprises three of the emitters 21 and one of the photodetectors 22 . This setup is shown in figure 1 . For some of the pixels 26 only the emitters 21 are operated and for some of the pixels 26 only the photodetector 22 is operated . This means , some of the pixels 26 are employed as emitters 21 and some of the pixels 26 are employed as photodetectors 22 . Since all the pixels 26 comprise emitters 21 and a photodetector 22 , it can be chosen freely which pixels 26 are operated as emitters 21 and which pixels 26 are operated as photodetectors 22 . This increases the flexibility . Increased flexibility can for example be employed to optimi ze the reduction of motion artefacts .

[0118] The method described with figure 5 can however also be carried out for the setups shown in figures 2 and 3 .

[0119] With figure 6 an exemplary embodiment of the method for training a machine learning algorithm is described . In a 2024P01359WO October 22 , 2025

[0120] P2024 , 0916 WO N

[0121] 23 first step S I first sensor data is provided to the machine learning algorithm, wherein the first sensor data is detected by the plurality of photodetectors 22 in situations where there is no motion in the detection region 23 from which the photodetectors 22 are configured to detect electromagnetic radiation . In a second step S2 second sensor data is provided to the machine learning algorithm, wherein the second sensor data is detected by the plurality of photodetectors 22 in situations where there is motion in the detection region 23 . In a third step S3 the machine learning algorithm is provided with the information that there is no motion in the detection region 23 during the detection of the first sensor data und that there is motion in the detection region 23 during the detection of the second sensor data .

[0122] Figure 7 shows another exemplary embodiment of the device 20 for detecting at least one vital sign . The device 20 comprises at least ten emitters 21 that are each configured to emit electromagnetic radiation, at least ten photodetectors 22 that are each configured to detect electromagnetic radiation from the detection region 23 and to provide sensor data, and a front side 31 , wherein the emitters 21 and the photodetectors 22 are arranged at a side of the device 20 that faces away from the front side 31 , and the device 20 is a wearable device 20 . At the front side 31 , for example a display 25 can be arranged . The emitters 21 and the photodetectors 22 are not explicitly shown in figure 7 as they can be arranged as shown in figure 1 , figure 2 or figure 3 .

[0123] It will be appreciated that the disclosure is not limited to the disclosed embodiments and to what has been particularly shown and described hereinabove . Rather, features recited in 2024P01359WO October 22, 2025

[0124] P2024, 0916 WO N

[0125] - 24 - separate dependent claims or in the description may advantageously be combined. Furthermore, the scope of the disclosure includes those variations and modifications, which will be apparent to those skilled in the art. The term "comprising", insofar it was used in the claims or in the description, does not exclude other elements or steps of a corresponding feature or procedure. In case that the terms "a" or "an" were used in conjunction with features, they do not exclude a plurality of such features. Moreover, any reference signs in the claims should not be construed as limiting the scope.

[0126] This patent application claims priority from German patent application 10 2024 135 633.2, the disclosure content of which is hereby included by reference.

[0127] 2024P01359WO October 22 , 2025

[0128] P2024 , 0916 WO N

[0129] - 25 -

[0130] References

[0131] 20 device

[0132] 21 emitter

[0133] 22 photodetector

[0134] 23 detection region

[0135] 24 calculation unit

[0136] 25 display

[0137] 26 pixel

[0138] 27 first emitter

[0139] 28 second emitter

[0140] 29 third emitter

[0141] 30 skin

[0142] 31 front side

Claims

2024P01359WO October 22, 2025P2024, 0916 WO N- 26 -Claims1. Device (20) for detecting at least one vital sign, the device (20) comprising:- a plurality of emitters (21) that are each configured to emit electromagnetic radiation,- a plurality of photodetectors (22) that are each configured to detect electromagnetic radiation from a detection region (23) and to provide sensor data, and- a calculation unit (24) , wherein- the calculation unit (24) is configured to determine at least one vital sign, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region (23) on the measurement of the vital sign.

2. Device (20) according to claim 1, wherein the device (20) comprises a display (25) , and the emitters (21) and the photodetectors (22) are arranged at a side of the device (20) facing away from the display (25) .

3. Device (20) according to one of the preceding claims, wherein the device (20) comprises a plurality of pixels (26) , wherein each pixel (26) comprises three of the emitters (21) and one of the photodetectors (22) .

4. Device (20) according to one of the preceding claims, wherein at least some of the emitters (21) are configured to emit electromagnetic radiation of a first wavelength and at least some of the emitters (21) are configured to emit electromagnetic radiation of a second wavelength, wherein the first wavelength is different from the second wavelength.2024P01359WO October 22, 2025P2024, 0916 WO N- 27 -5. Device (20) according to one of the preceding claims, wherein the size of each of the emitters (21) is larger than the size of each of the photodetectors (22) .

6. Device (20) according to one of the preceding claims, wherein the photodetectors (22) are arranged along a grid.

7. Device (20) according to one of the preceding claims, wherein the emitters (21) and the photodetectors (22) are arranged in a two-dimensional array.

8. Device (20) according to one of the preceding claims, wherein the calculation unit (24) is configured to average sensor data.

9. Device (20) according to one of the preceding claims, wherein at least some emitters (21) and photodetectors (22) are arranged spaced apart from each other such that the space between neighboring emitters (21) and / or photodetectors (22) is larger than the size of the emitters (21) or photodetectors (22) .

10. Device (20) according to one of the preceding claims, wherein each emitter (21) has a lateral extent of at least 10 pm and 4 mm at most, wherein the lateral extent is given in a plane that extends perpendicular to a main plane of extension of the device (20) .

11. Device (20) according to one of the preceding claims, wherein each photodetector (22) has a lateral extent of at least 5 pm and 200 pm at most, wherein the lateral extent is given in a plane that extends perpendicular to a main plane of extension of the device (20) .2024P01359WO October 22, 2025P2024, 0916 WO N2812. Method for detecting at least one vital sign, the method comprising :- emitting electromagnetic radiation by a plurality of emitters (21) of a device (20) ,- detecting electromagnetic radiation by a plurality of photodetectors (22) of the device (20) from a detection region (23) ,- providing sensor data by the photodetectors (22) , and- determining at least one vital sign, wherein a machine learning algorithm is used to retrieve the at least one vital sign from the sensor data, wherein the machine learning algorithm has been trained to reduce the impact of motion in the detection region (23) on the measurement of the vital sign .

13. Method according to claim 12, wherein- the device (20) comprises a plurality of pixels (26) , wherein each pixel (26) comprises three of the emitters (21) and one of the photodetectors (22) , and- for some of the pixels (26) only the emitters (21) are operated and for some of the pixels (26) only the photodetector (22) is operated.

14. Method for training a machine learning algorithm, the method comprising:- providing first sensor data to the machine learning algorithm, wherein the first sensor data is detected by a plurality of photodetectors (22) in situations where there is no motion in a detection region (23) from which the photodetectors (22) are configured to detect electromagnetic radiation,- providing second sensor data to the machine learning algorithm, wherein the second sensor data is detected by the2024P01359WO October 22, 2025P2024, 0916 WO N- 29 - plurality of photodetectors (22) in situations where there is motion in the detection region (23) , and- providing the machine learning algorithm with the information that there is no motion in the detection region (23) during the detection of the first sensor data und that there is motion in the detection region (23) during the detection of the second sensor data.

15. Computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method of claim 14.

Images