WO2021006790A1

WO2021006790A1 - Torso-mounted accelerometer signal reconstruction

Info

Publication number: WO2021006790A1
Application number: PCT/SE2020/050619
Authority: WO
Inventors: Mohammed El-Beltagy
Original assignee: Wememove Ab
Priority date: 2019-07-10
Filing date: 2020-06-15
Publication date: 2021-01-14
Also published as: SE1950879A1; EP4061215A1

Abstract

The present disclosure relates to a device and method for torso-mounted accelerometer signal reconstruction. In an aspect a method of acquiring recorded motion data of an individual (100) is provided. The method comprises acquiring (S101) motion data of the individual (100) recorded with a first torso-attached accelerometer (101), acquiring (S102) motion data of the individual (100) recorded with a second accelerometer (103) attached to a different part of the individual (100) than the torso, and determining (S103) a mapping function configured to map the motion data recorded with the second accelerometer (103) to the motion data recorded with the first accelerometer (101) for subsequently reconstructing torso-recorded motion data from motion data being recorded with the second accelerometer (103) and processed by the determined mapping function.

Description

TORSO-MOUNTED ACCELEROMETER SIGNAL RECONSTRUCTION TECHNICAL FIELD

[ooi] The present disclosure relates to a device and method for torso-mounted accelerometer signal reconstruction.

BACKGROUND

[002] Torso-mounted accelerometers are used in sports for motion detection and analysis with aim to improve movement pattern of a wearer of the accelerometer during exercising such as for instance running or cross-country skiing. Motion data is recorded and subsequently processed to provide an animation of the wearer’s motion pattern during the exercise for review by the wearer, possibly with feedback to the wearer on improving actions to be taken, such as a proposed change of stride length, an instruction to run in a more upright manner, to more aggressively pivot the arms back and forth, etc.

[003] There is a good rational to have a torso-mounted accelerometer for motion detection and analysis. Torso-mounted accelerometers tend to sit quite closely to the centre or mass of the wearer’s body, and hence can capture a great deal of motion nuances.

[004] However, torso-mounted accelerometers are rather inconvenient as they typically must be fastened and adjusted with a strap over the wearer’s chest and maintained in that position throughout the exercise.

SUMMARY

[005] One object is to solve, or at least mitigate, this problem in the art and thus to provide an improved method of acquiring recorded motion signals of an individual.

[006] This object is attained in a first aspect by a method of acquiring recorded motion data of an individual. The method comprises acquiring motion data of the individual recorded with a first torso-attached accelerometer, acquiring motion data of the individual recorded with a second accelerometer attached to a different part of the individual than the torso, and determining a mapping function configured to map the motion data recorded with the second accelerometer to the motion data recorded with the first accelerometer for subsequently reconstructing torso-recorded motion data from motion data being recorded with the second accelerometer and processed by the determined mapping function.

[007] This object is attained in a second aspect by a device configured to acquire recorded motion data of an individual. The device comprises a processing unit and a memory, said memory containing instructions executable by the processing unit, whereby the device is operative to acquire motion data of the individual recorded with a first torso-attached accelerometer, acquire motion data of the individual recorded with a second accelerometer attached to a different part of the individual than the torso, and to determine a mapping function configured to map the motion data recorded with the second accelerometer to the motion data recorded with the first accelerometer for subsequently reconstructing torso-recorded motion data from motion data being recorded with the second accelerometer and processed by the determined mapping function.

[008] This object is attained in a third aspect by a method of reconstructing motion data of an individual. The method comprises acquiring motion data of the individual recorded with a second accelerometer attached to a different part of the individual than a torso, and processing the acquired motion data in a mapping function having been previously determined by mapping the motion data recorded with the second

accelerometer (103) to motion data of the individual (too) recorded with a first torso- attached accelerometer (101), thereby reconstructing (S203) torso-recorded motion data.

[009] This object is attained in a fourth aspect by a device configured to reconstruct motion data of an individual. The device comprises a processing unit and a memory, the memory containing instructions executable by the processing unit, whereby the device is operative to acquire motion data of the individual recorded with a second accelerometer attached to a different part of the individual than a torso, and to process the acquired motion data in a mapping function having been previously determined by mapping the motion data recorded with the second accelerometer to motion data of the individual recorded with a first torso-attached accelerometer, thereby reconstructing torso- recorded motion data.

[0010] In an embodiment, when reconstructing torso-recorded motion data, the mapping function used is that determined using the method of the first aspect. [oon] A solution to the previously described problem in the art is proposed by allowing placement of the accelerometer elsewhere on the wearer’s body, somewhere more convenient, such as in a holder fastened around an upper part of the wearer’s arm or around the wrist, or in a pair of in-ear headphones or over-ear headphones to be fastened to the wearer’s head. It is also envisaged that the accelerometer is implemented in for instance a smart watch of in smart phone placed in a holder fastened around an upper part of the wearer’s arm or around the wrist.

[0012] However, a problem then arises that the accelerometer no longer is placed at the centre of mass of the wearer’s body, which e.g. has the consequence that algorithms developed in this particular technical field no longer will produce reliable motion analysis data as they assume torso-based motion signals for processing, while motion signals originating from other locations of the body will differ from the torso-based motion signals.

[0013] This is solved in that signals recorded by an accelerometer at another mounted location (AML) will be adapted and matched to signals of a torso-mounted (TM) accelerometer by utilizing a mapping function, after which process TM

accelerometer signals may be reconstructed from recorded AML accelerometer signals having been processed by said mapping function. Advantageously, the reconstructed TM accelerometer signals may be processed in the already available torso-based motion detection and analysis algorithms.

[0014] In an embodiment, the mapping function is determined such that an error between the reconstructed torso-recorded motion data and the corresponding motion data recorded with the first accelerometer is minimized.

[0015] In an embodiment, a user profile is acquired of the individual for which the motion data is acquired, the user profile being associated with the determined mapping function.

[0016] In an embodiment, the user profile comprises information including at least one of weight, height, sex, chest width, placement of the second accelerometer.

[0017] In an embodiment, the mapping function is determined based on motion data of a plurality individuals having a similar user profile. [0018] In an embodiment, a mapping function having been previously determined for a first individual is used to reconstruct torso-recorded motion data of a second individual.

[0019] In an embodiment, a request to use a determined mapping function is received, the request comprising a user profile of the requesting individual and motion data of the requesting individual recorded with a second accelerometer attached to a different part of the individual than a torso, and the motion data of the requesting individual is processed using a mapping function associated with a user profile best matching the user profile of the requesting individual to reconstruct torso-recorded motion data of the requesting individual.

[0020] In an embodiment, a motion pattern of the individual is detected based on the reconstructed torso-recorded motion data.

[0021] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise herein. All references to "a/an/the element, apparatus, component, means, step, etc." are to be interpreted openly as referring to at least one instance of the element, apparatus, component, means, step, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not have to be performed in the exact order disclosed, unless explicitly stated.

BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Aspects and embodiments are now described, by way of example, with refer ence to the accompanying drawings, in which:

[0023] Figure 1 illustrates a user wearing a torso-mounted accelerometer;

[0024] Figure 2 illustrates a user wearing a head-mounted accelerometer;

[0025] Figures 3a-c illustrate motion data recorded in three dimensions X, Y and Z by a torso-mounted accelerometer and a headphone-mounted accelerometer, respectively, according to an embodiment:

[0026] Figure 4 illustrates recording of motion data of a wearer, deriving of a mapping function, and reconstruction of torso-based motion data according to an embodiment; [0027] Figure 5 illustrates recorded and reconstructed accelerometer motion signals according to an embodiment;

[0028] Figure 6 illustrates a user requesting to use a previously determined mapping function associated with a user profile matching her own user profile according to an embodiment; and

[0029] Figure 7 illustrates a device configured to acquire recorded motion data of an individual according to an embodiment.

DETAILED DESCRIPTION

[0030] The aspects of the present disclosure will now be described more fully hereinafter with reference to the accompanying drawings, in which certain

embodiments are shown.

[0031] These aspects may, however, be embodied in many different forms and should not be construed as limiting; rather, these embodiments are provided by way of example so that this disclosure will be thorough and complete, and to fully convey the scope of all aspects to those skilled in the art. Like numbers refer to like elements throughout the description.

[0032] With reference to Figure 1, as previously mentioned, a torso-mounted accelerometer 101 has the advantage of being placed close to the centre of mass of the body of the wearer too, and hence can capture a great deal of motion nuances. The torso is widely considered to be the best location for placement of an accelerometer and available algorithms utilized for the motion detection and analysis are typically adapted for processing torso-based motion data. However, the torso is not the most convenient position for wearing the accelerometer 101, and a strap 102 is required to fasten the accelerometer 101 over the wearer’s chest.

[0033] A solution is proposed to this problem by allowing placement of the accelerometer elsewhere on the wearer’s body, somewhere more convenient, such as in a holder fastened around an upper part of the wearer’s arm or around the wrist, or in a pair of in-ear headphones or over-ear headphones to be fastened to the wearer’s head. It is also envisaged that the accelerometer is implemented in for instance a smart phone placed in a holder fastened around an upper part of the wearer’s arm or around the wrist. [0034] Figure 2 illustrates an accelerometer 103 being part of an in-ear headphone attached to the wearer too.

[0035] However, the problem then arises that the accelerometer 103 no longer is placed at the centre of mass of the wearer’s body, which e.g. has the consequence that algorithms developed in this particular technical field no longer will produce reliable motion analysis data as they assume torso-based motion signals for processing, while motion signals originating from other locations of the body will differ from the torso- based motion signals.

[0036] Hence, data representing motion key point indicators (KPIs) such as for example vertical oscillation (VO), cadence, ground contact time (GCT), etc., recorded for instance by a headphone-mounted accelerometer 103 will not match data recorded by a torso-mounted accelerometer 101, and the existing torso-based algorithms utilized for the motion detection and analysis will thus not produce accurate results if being supplied with motion data recorded by a headphone-mounted accelerometer. That is, KPIs computed using torso-based algorithms for processing torso-recorded motion signals will not match KPIs computed using available torso-based algorithms for processing motion signals recorded by accelerometers located at a part of the body of the wearer too different from the torso.

[0037] An embodiment solves this problem in that signals recorded by an

accelerometer 103 at another mounted location (AML) will be adapted and matched to signals of a torso-mounted (TM) accelerometer 101 by utilizing a mapping function, after which process TM accelerometer signals may be reconstructed from recorded AML accelerometer signals having been processed by said mapping function.

Advantageously, the reconstructed TM accelerometer signals may be processed in the already available torso-based motion detection and analysis algorithms. In the following,“motion signals” or“motion data” will denote signals produced by the accelerometers 101, 103 from which motion of the wearer too is detected and possibly even reconstructed by means of animation to be presented on a suitable display for review by the wearer.

[0038] Such signals are illustrated in the following with reference to Figures 3a-c.

[0039] Thus, during a derivation phase, TM motion signals will be recorded by a torso-mounted accelerometer 101, while AML motion signals will be recorded by for instance a headphone-mounted accelerometer 103, both being attached to the wearer too.

[0040] Figures 3a-c illustrate accelerometer data recorded in all three dimensions X, Y and Z, respectively, for the TM accelerometer and the AML accelerometer. The TM motion signals are illustrated with continuous lines, while the AML motion signals are illustrated with continuous lines. There is typically a non-linear relationship between the TM motion signals and the AML motion signals.

[0041] Figure 4 thus illustrates recording of motion data of the wearer too using the TM accelerometer 101 in step Slot and recording of motion data of the wearer too using the AML accelerometer 103 in step S102 during the derivation phase. In a next step S103, the AML accelerometer data is adjusted such that it matches the TM

accelerometer data by using an appropriately elaborated mapping function M.

Optionally, a user profile of the wearer too may be acquired and associated with the mapping function M is shown in step S104. This will be discussed further hereinbelow.

[0042] Hence, from the acquired AML accelerometer motion data and the acquired TM accelerometer motion data, a mathematical mapping function M is derived, thereby mapping the two sets of motion data to each other such that during a subsequent TM accelerometer motion data reconstruction phase, AML accelerometer motion data being recorded in step S201 by the AML accelerometer 103 and processed in step S202 by the mapping function M will result in (hypothetically recorded) TM accelerometer motion data being reconstructed:

TM accelerometer motion data = M(AML accelerometer motion data)

In other words, TM accelerometer motion data may subsequently be reconstructed in step S203 from AML accelerometer motion data having been recorded in step S201 and processed by the determined mapping function M in step S202.

[0043] Advantageously, once the mapping function M has been determined for the wearer too in the derivation phase using motion data of the TM accelerometer 101 and the AML accelerometer 103, TM accelerometer motion data can be reconstructed by processing recorded AML accelerometer motion data in the mapping function M.

[0044] Hence, the wearer too will only initially - during the derivation phase - have to wear the TM accelerometer 101. Once the mapping function M has been determined, the signals of the AML accelerometer 103 can be utilized to reconstruct the TM accelerometer signals and the TM accelerometer tot is thus advantageously no longer needed to obtain motion analysis information.

[0045] In an embodiment, the composite three-dimensional motion signal of the AML accelerometer 103 is denoted a(t) = a_x(t), a_y(t), a_z(t).

[0046] Similarly, the composite three-dimensional motion signal of the TM

accelerometer 101 is denoted r(t) = r_x(t), r_y(t), r_z(t).

[0047] When determining the mapping function M, the aim is to find an accurate predictor M which results in the following mapping:

[0048] That is, at any given time t, it should be possible to reconstruct the TM accelerometer motion signal by creating an estimate f (t) of the TM accelerometer motion signal based on the AML accelerometer motion signal over a time window defined by a lower bound and an upper bound t - di_¾ and t - 8_Ub, respectively.

[0049] During the previously described derivation phase, where the mapping function M is determined, the mapping function M may have to be fine-tuned - i.e. calibrated - such that an error between the estimates f (t) of the TM accelerometer motion signals (produced by processing the AML accelerometer motion signals a(t) with the mapping function M) and the actually measured values r(t) of the TM accelerometer motion signals is minimized. As is understood, during the derivation phase the respective accelerometer motion signals r(t) and a(t) are concurrently measured.

[0050] Hence, in this embodiment, the mapping function M is determined such that the error between the reconstructed value f(t)- i.e. the estimated value - of the TM accelerometer motion signal at a given point in time t, which is based on the AML accelerometer motion signal a(t) processed by the mapping function M, and the measured value r(t) of the TM accelerometer signal at said given point in time is minimized.

[0051] Different methods of determining M is envisaged, such as the utilization of convolutional neural networks, recurrent neural networks, neural differential equations, etc.

[0052] Each one of these approaches represents a rich class of functions that can be utilized to embody spatio-temporal mapping provided by M. The selection of the particular approach to be utilized depends on a number of factors, such as for instance desired level of accuracy required for the mapping or computational load required to reconstruct the TM motion data.

[0053] As an example, a recurrent neural network, such as a so called“long short term memory” network may be used to derive a highly accurate mapping model M, but it may computationally be too demanding for e.g. smart watch deployment; in which case a convolutional neural network may be considered being less accurate but on the other hand requiring less computational power.

[0054] Figure 5 illustrates in an upper view a motion signal recorded by a chest- mounted accelerometer 101 and a motion signal recorded by an ear-mounted

accelerometer 103 over a short span of time (for the Y dimension). The motion signal of the chest-mounted accelerometer 101 is illustrated with a continuous line, while the motion signal of the ear-mounted accelerometer 103 is illustrated with a dotted line.

[0055] In a lower view, the motion signal recorded by a chest-mounted

accelerometer 101 (continuous line) is shown together with the chest-recorded motion signal reconstructed/estimated from the motion signal recorded by the ear-mounted accelerometer 103 (dotted line) and processed by the mapping function M. As can be concluded, it is possible to reconstruct a chest-recorded motion signal which is quite accurate as compared to the actual chest-recorded motion signal.

[0056] As an example, the accuracy of the reconstruction may be determined by computing a so-called mean-squared error (MSE), wherein the mapping function M may be adjusted such that the reconstructed chest-recorded motion signals corresponds to the actually recorded chest signals in such a manner that the MSE is small (or even minimized).

[0057] Now, after the derivation process is finished, and the function M has been derived, for instance by a smart phone or computer of the wearer or even by a cloud server to which accelerometer data is transferred, the wearer no longer needs to use the TM accelerometer but can henceforth use her headphone-mounted accelerometer to record motion data while the algorithms utilized for motion detection and analysis - possibly hosted on a computer or smart-phone to which the AML accelerometer data is supplied after exercise (or in real-time) and on a display of which an animation may be presented - are the available algorithms utilized to process TM motion data. It may further be envisaged that the derivation process is performed by one device, such as a cloud server, while the reconstruction process is performed by another, e.g. a smart phone to which the determined mapping function M has been transferred from the cloud server.

[0058] It may be envisaged that each individual user will undergo the calibration process for best and most accurate end result. However, it could alternatively be envisaged in an embodiment that a generic model M is derived using AML and TM acceleration motion data of one or more users and that future users can adopt the derived model M with good result.

[0059] In a further embodiment, for each determined mapping function M, a corresponding user profile is stored containing information such as weight, height, sex, chest width, etc., wherein a new user can select a function M based on the closest matching user profile.

[0060] In yet an alternative embodiment, TM and AML accelerometer data of a great number of users having an identical or similar user profile is acquired to determine a corresponding mapping function M. That would have the advantage that a great amount of data originating from users having identical, or at least similar, user profiles will be used to derive a mathematical mapping model M.

[0061] With reference to Figure 6, a user too having for instance an earphone- mounted accelerometer 103 and a smart phone 104 may download a selected mapping function M from a server 105 located in the cloud 106 to a motion analysis app on the phone 104. The user too may selected a mapping function M which best matches her individual user profile. Alternatively, the user too may supply her user profile to the cloud server 105 along with motion data recorded by the AML accelerometer 103 for the user too to the cloud server 105, which subsequently reconstructs TM motion data based on the received user profile and the AML accelerometer motion data and possibly returns the reconstructed TM motion data to the smart phone 104 for presentation to the user too.

[0062] Thus, upon determining a mapping function M based on the recorded TM accelerometer motion data and the recorded AML accelerometer motion data as described in step S103 in Figure 4, the cloud server 105 will also in step 104 acquire a user profile of the user for which the function M is derived. Hence, a specific user profile may specify:

length: 162 cm,

weight: 55 kg,

sex: female, and

AML accelerometer placement: ear.

[0063] Now, the user profile is stored at the server 105 as profilei and the corresponding determined mapping function is stored as Ml.

[0064] For a second user, a second specific user profile profile2 is stored along with the corresponding determined mapping function M2. For a third user, a third specific user profile profiles is stored along with the corresponding determined mapping function M3, and so on. Hence, the cloud server 105 may contain a great number of determined mapping functions and associated user profiles.

[0065] Again, with reference to Figure 6, a user 100 wishing to acquire a suitable mapping function may thus make such a request to the server 105 via her smart phone 104 and include her user profile with the request. Assuming that the user 100 provides identical or similar information as that contained in the above exemplified user profile profilei, the server 105 will return the associated mapping function Ml.

[0066] Alternatively, as previously discussed, the server 105 may itself use the mapping function Ml for reconstruction of TM accelerometer motion data given that the user too also provides her AML accelerometer motion data on which the

reconstruction is to be based. The server 105 may thus subsequently supply the smart phone 104 with the reconstructed of TM accelerometer motion data, or possibly even an animation illustrating the motion pattern of the requesting individual too.

[0067] The mapping function Ml is likely to be able to accurately reconstruct TM accelerometer motion signals from the motion signals recorded by the ear-mounted accelerometer 103 given that the user too has the same user profile as the user associated with profilei. Thus, there is advantageously no need for the user too to wear a torso-mounted accelerometer to derive a mapping function, but an already determined mapping function Ml can be used with good accuracy.

[0068] Figure 7 illustrates a device 105, e.g. a cloud server, according to an

embodiment. The steps of the method performed by the device 105, being embodied e.g. in the form of a computer, server, smart phone, etc., of recording motion data of an individual according to embodiments are in practice performed by a processing unit 120 embodied in the form of one or more microprocessors arranged to execute a computer program 121 downloaded to a suitable storage volatile medium 122 associated with the microprocessor, such as a Random Access Memory (RAM), or a non-volatile storage medium such as a Flash memory or a hard disk drive. The processing unit 120 is arranged to cause the device 105 to carry out the method according to embodiments when the appropriate computer program 121 comprising computer-executable instructions is downloaded to the storage medium 122 and executed by the processing unit 120. The storage medium 122 may also be a computer program product comprising the computer program 121. Alternatively, the computer program 121 may be transferred to the storage medium 122 by means of a suitable computer program product, such as a Digital Versatile Disc (DVD) or a memory stick. As a further alternative, the computer program 121 may be downloaded to the storage medium 122 over a network. The processing unit 120 may alternatively be embodied in the form of a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a complex programmable logic device (CPLD), etc.

[0069] The aspects of the present disclosure have mainly been described above with reference to a few embodiments and examples thereof. However, as is readily

appreciated by a person skilled in the art, other embodiments than the ones disclosed above are equally possible within the scope of the disclosure, as defined by the appended patent claims.

[0070] Thus, while various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art. The various aspects and embodiments disclosed herein are for purposes of illustration and are not intended to be limiting, with the true scope and spirit being indicated by the following claims.

Claims

1. A method of acquiring recorded motion data of an individual (too) comprising: acquiring (Slot) motion data of the individual (too) recorded with a first torso- attached accelerometer (101);

acquiring (S102) motion data of the individual (too) recorded with a second accelerometer (103) attached to a different part of the individual (too) than the torso; determining (S103) a mapping function configured to map the motion data recorded with the second accelerometer (103) to the motion data recorded with the first accelerometer (101) for subsequently reconstructing torso-recorded motion data from motion data being recorded with the second accelerometer (103) and processed by the determined mapping function.

2. The method of claim 1, wherein the mapping function is determined such that an error between the reconstructed torso-recorded motion data and the corresponding motion data recorded with the first accelerometer (101) is minimized.

3. The method of claims 1 or 2, further comprising:

acquiring (S104) a user profile of the individual (too) for which the motion data is acquired, the user profile being associated with the determined mapping function.

4. The method of claim 3, the user profile comprising information including at least one of weight, height, sex, chest width, placement of the second accelerometer (103).

5. The method of any one of claims 3 or 4, wherein the mapping function is determined based on motion data of a plurality individuals having a similar user profile.

6. A method of reconstructing motion data of an individual (too) comprising:

acquiring (S201) motion data of the individual (too) recorded with a second accelerometer (103) attached to a different part of the individual (too) than a torso; processing (S202) the acquired motion data in a mapping function having been previously determined by mapping the motion data recorded with the second

7. The method of claim 6, wherein the mapping function is determined using the method of claim 1.

8. The method of claims 6 or 7, wherein a mapping function having been previously determined for a first individual is used to reconstruct torso-recorded motion data of a second individual.

9. The method of claim 8, further comprising;

receiving a request to use a determined mapping function, the request comprising a user profile of the requesting individual and motion data of the requesting individual (too) recorded with a second accelerometer (103) attached to a different part of the individual (too) than a torso; and

processing the motion data of the requesting individual using a mapping function associated with a user profile best matching the user profile of the requesting individual (too) to reconstruct torso-recorded motion data of the requesting individual.

10. The method of any one of claims 6-9, further comprising:

detecting a motion pattern of the individual based on the reconstructed torso- recorded motion data.

11. A device (105) configured to acquire recorded motion data of an individual (too), the device (105) comprising a processing unit (120) and a memory (122), said memory containing instructions (121) executable by said processing unit (120), whereby the device (105) is operative to:

acquire motion data of the individual (too) recorded with a first torso-attached accelerometer (101);

acquire motion data of the individual (too) recorded with a second accelerometer (103) attached to a different part of the individual (too) than the torso;

determine a mapping function configured to map the motion data recorded with the second accelerometer (103) to the motion data recorded with the first accelerometer (101) for subsequently reconstructing torso-recorded motion data from motion data being recorded with the second accelerometer (103) and processed by the determined mapping function.

12. The device (105) of claim 11 being operative to determine the mapping function such that an error between the reconstructed torso-recorded motion data and the corresponding motion data recorded with the first accelerometer (101) is minimized.

13. The device (105) of claims 11 or 12, further being operative to:

acquire a user profile of the individual (too) for which the motion data is acquired, the user profile being associated with the determined mapping function.

14. The device (105) of claim 3, the user profile comprising information including at least one of weight, height, sex, chest width, placement of the second accelerometer

(103).

15. The device (105) of any one of claims 13 or 14, being operative to determine the mapping function based on motion data of a plurality individuals having a similar user profile.

16. A device (105) configured to reconstruct motion data of an individual (too), the device (105) comprising a processing unit (120) and a memory (122), said memory containing instructions (121) executable by said processing unit (120), whereby the device (105) is operative to:

acquire motion data of the individual (too) recorded with a second accelerometer (103) attached to a different part of the individual (too) than a torso;

process the acquired motion data in a mapping function having been previously determined by mapping the motion data recorded with the second accelerometer (103) to motion data of the individual (too) recorded with a first torso-attached accelerometer (101), thereby reconstructing (S203) torso-recorded motion data.

17. The device (105) of claim 16, further being configured to determine the mapping function using the method of claim 1.

18. The device (105) of claims 16 or 17, being configured to use a mapping function having been previously determined for a first individual to reconstruct torso-recorded motion data of a second individual.

19. The device (105) of claim 18, further being operative to:

receive a request to use a determined mapping function, the request comprising a user profile of the requesting individual and motion data of the requesting individual (too) recorded with a second accelerometer (103) attached to a different part of the individual (too) than a torso; and

process the motion data of the requesting individual using a mapping function associated with a user profile best matching the user profile of the requesting individual (too) to reconstruct torso-recorded motion data of the requesting individual.

20. The device (105) of any one of claims 16-19, further being operative to:

detect a motion pattern of the individual based on the reconstructed torso- recorded motion data.

21. A computer program (121) comprising computer-executable instructions for causing a device (105) to perform steps recited in any one of claims 1-10 when the computer-executable instructions are executed on a processing unit (120) included in the device (105).

22. A computer program product comprising a computer readable medium (122), the computer readable medium having the computer program (121) according to claim 21 embodied thereon.