Wearable Sensor for monitoring joint

kinematics

Project Report Submitted in Partial Fulfilment of the Requirements

for the Degree of

BACHELOR OF TECHNOLOGY
by

Siddharth Anand Srivastav 19095092

Kushagra Yadav 19095058
BVV Pavan Kalyan 19095024

Under the guidance of

Dr. M. Thotappan

Department of Electronics Engineering

INDIAN INSTITUTE OF TECHNOLOGY (BHU) VARANASI
Varanasi 221005, India
November 2022
 DECLARATION

We hereby declare that the work reported in the project titled “Wearable sensor
for monitoring joint kinematics” submitted by Siddharth Anand Srivastav
(19095092), Kushagra Yadav (19095058), BVV Pavan Kalyan (19095024)
for the partial fulfillment of B.Tech degree at the Department of Electronics Engineer-
ing, Indian Institute of Technology BHU, Varanasi, is record of my work carried out
under the supervision of Dr. M. Thotappan. I further declare that the work reported
in this project has not been submitted and will not be submitted, either in part or
in full, for the award of any other degree in this institute or any other institute or
university.

Siddharth Anand Srivastav (19095092)

Kushagra Yadav (19095058)
BVV Pavan Kalyan (19095024)
(B.Tech)
Department of Electronics Engineering
IIT BHU, Varanasi
November 2022

2
 INDIAN INSTITUTE OF TECHNOLOGY BHU, VARANASI
DEPARTMENT OF ELECTRONICS ENGINEERING

CERTIFICATE BY CANDIDATE

This is to certify that the work incorporated in the project report entitled âWearable
sensor for monitoring joint kinematicsâ is a record of work carried out by Siddharth
Anand Srivastav (19095092),Kushagra Yadav(19095058), BVV Pavan Kalyan(19095024),
under my guidance and supervision for the award of B.Tech in Electronics Engineering
in Indian Institute of Technology (BHU), Varanasi.

Mentored By: Dr. M. Thotappan

Department of Electronics Engineering
IIT BHU- 221005

3
 INDIAN INSTITUTE OF TECHNOLOGY BHU, VARANASI
DEPARTMENT OF ELECTRONICS ENGINEERING

CERTIFICATE BY SUPERVISOR

It is certified that project entitled Wearable sensor for monitoring joint kine-
matics was carried under my supervision for partial fulfillment of B.Tech degree in
Electronics Engineering. It is further certified that the above candidate has car-
ried out the project work under my guidance during the academic session 2022-2023
at the Department of Electronics Engineering, Indian Institute of Technology BHU,
Varanasi.

Dr. M. Thotappan
Department of Electronics Engineering
IIT BHU- 221005

4
Contents

1 Introduction 2

2 Operating Principle 4

3 Design 7

4 Simulation and Results 10

5 Selection of Optimal Operation Frequency 13

6 Conclusion 18

1
Chapter 1

Introduction

Monitoring elbow joint kinematics after a medical procedure (tennis el-

bow, broken elbow surgery, etc.) is critical for maximizing/accelerating
rehabilitation, and preventing future injuries. The most common tech-
nologies used to date for monitoring elbow joint kinematics include
3D/2D motion capturing cameras and goniometers.

Specifically, optical camera-based motion capture labs and their mark-

erless versions are accurate but restricted to contrived environments.
Inertial Measurement Units (IMUs) can work outside laboratory set-
tings but suffer from inherent drift. Electromagnetics-based techniques,
such as radars, backscattering and Wi-Fi, are typically used to classify
activities instead of monitoring motion as a function of time and require
bulky set-ups that restrict them to contrived settings. Time-of-flight
sensors that utilize ultra-wideband radio or ultrasonic transceivers are
restricted by line-of-sight issues. Bending sensors that are worn directly
on the joint restrict natural motion and have limited cycles of use.

2
 Unfortunately, these technologies are not portable, implying that
they cannot capture elbow kinematics in the individualâs daily environ-
ment. Wearable joint monitoring sensors have recently been proposed,
however they raise concerns as to their obtrusiveness and maximum
number of flexions they can withstand. Therefore, there is a large
discrepancy between the capabilities of current technologies versus the
capabilities required in order to efficiently and effectively monitor elbow
joint kinematics in real-time and while the patient is performing daily
tasks (e.g., exercising, running, or simply walking or sitting).

To address these shortcomings in the state-of-the-art, we report a

technology that relies on wearable radio frequency (RF) loop sensors to
monitor joint (e.g., elbow/knee) kinematics in the individual’s natural
environment. The project seeks to develop a high-accuracy, lightweight,
flexible, robust, and low-cost wearable joint sensor that is unobtrusively
embroidered into a shirt.

3
Chapter 2

Operating Principle

Two mutually coupled coils (transmitter (Tx) and receiver (Rx)) of un-
equal radii are placed symmetrically across the elbow joint, as shown
in Fig. 1. When the joint flexes (i.e., bending of the limbs about the
joint, with the center of rotation marked as C in Fig. 1), the Tx and
Rx coils get misaligned with respect to each other. This changes the
transmission coefficient (|S21 |) between the two coils, which in turn can
be used to monitor joint flexion.
In this study, our ultimate aim is to monitor flexion angle (θf ), which is
defined as the angle formed between the axes of two limbs at the center
of rotation (C) per Fig. 1. It is important to note that coils have to
be, preferably, placed symmetrically across the joint to obtain a consis-
tently increasing trend of |S21 | with increasing θf , In fact, asymmetric
placement can also be used to monitor joint flexion; however, beyond
a certain limit, the increasing trend can flip after a particular flexion
angle and can cause ambiguity in measurement.

4
Fig. 1: Simulation set-up depicting wearable TC loop sensor upon a
cylindrical limb that is straight and bent at flexion angle θf

The working of the proposed wearable joint monitoring sensor is

based on the principles of magnetic flux .We propose to make the sys-
tem using two coils as shown in fig above .When the receiving and
transmitting coils are perfectly parallel i.e when there is no bending of
the joint ,an RF signal is fed through the transmitter,then the magnetic
flux captured at the receiver is maximum .In contrast ,if any bending
of the joint takes place it will reduce the magnetic flux captured at the
receiver and reducing the voltage at the receiving coil,this decrease can
be experimentally measured in order to determine the joint flexion.

In principle, the system can be realized using three modes of op-

eration, viz., 1) electrically large loops (circumference ∼ λ , where λ
represents wavelength); 2) electrically small loops (circumference < 0.1λ
) or 3) in between both modes of operation (0.1λ<circumference<λ ).
Extensive frequency studies similar to those indicate that electrically

5
small loops (ESLAs) operating at 34 MHz provide optimal performance
in terms of: 1) received power levels; 2) resolution (smallest detectable
angle); and 3) robustness to changes in tissue properties.

The operating principle can be better understood using Faraday′ s

law of induction

where BT x is the transmitted magnetic flux density vector, nRx is

the receiver area vector, and VRx is the voltage induced on the re-
ceiver. Changes in flexion/rotation angle correspond to changes in the
angle between nRx and BT x . This gets reflected in VRx and eventually
captured in the transmission coefficient |S21 |. That is, transmission co-
efficient becomes a function of flexion angle and, hence, can be used to
monitor joint flexion and rotation.

6
Chapter 3

Design

The designed wearable joint monitoring sensor can be decomposed into

three components in order to better analyze the findings ,a sensing in-
terface, a wireless interface and a power interface.

SENSING INTERFACE
The sensing interface consists of two coils serving as a transmitter and
receiver located at either end of the wearable sleeve. A low-power RF
signal is fed into the transmitter that generates a magnetic field, and
the flux captured at the receiver generates a voltage on the receiver
coil. The ISM band of 433 MHz was selected in order to avoid EM
interference with other wireless transmissions and to ensure good cou-
pling between the coils as attributed to its inherent long wavelength.
Next, an experimental prototype can be devised as shown in Fig.

7
WIRELESS INTERFACE

The BITalino sensor platform can be used in order to enable wireless

connectivity with the wearable joint monitoring sensor. The platform
includes a portable, lightweight Bluetooth interface that can be used
to wirelessly transmit the voltage readings captured at the receiver coil
to a laptop or smartphone. Further integration is still being conducted
to fully incorporate the BITalino Bluetooth IC into the experimental
setup of the sensing interface

POWER INTERFACE

A critical design requirement of the wearable joint monitoring sensor

was flexibility and unobtrusive integration to a fabric sleeve. Conven-
tional batteries are composed of bulky, rigid components; therefore,

8
electrochemical fabrics can be designed in order to serve as the power
supply of the wearable sensor. The designed electrochemical fabrics
must consist of solid silver and zinc metals deposited onto a flexible,
adhered onto the fabric with a polyvinylidene fluoride (PVDF) binding
solution. Once the electrochemical fabric was placed in a conducting
fluid (sweat, saline solution, etc.), the free ions in the conducting fluid
induced a charge across the metal deposits on the fabric. Furthermore,
conducting E-threads can be used to wire several metal deposits to-
gether: providing scalable voltage from a flexible power supply that
could be used to power the wearable joint monitoring sensor. Cur-
rently, research is still being conducted to fully integrate the designed
electrochemical fabric batteries with the wearable sensor.

9
Chapter 4

Simulation and Results

The employed simulation set-up is shown in Fig. 2. As a proof-of-

concept, a homogeneous cylindrical model of the human arm is consid-
ered. The upper arm and forearm are modeled as cylinders (3.9 cm in
radius, 25 cm in length), while the elbow joint is modeled as a sphere
(3.9 cm in radius). The tissue-simulating material is 2/3 muscle, as
frequently used in the literature to represent the average human body
propertie. Two single-turn coils, Coil 1 (Tx per Fig. 1) and Coil 2 (Rx
per Fig. 1) are wrapped around the cylindrical arm model, placed sym-
metrically with respect to the joint at a gap g12 between the two. Both
coils exhibit a radius of 4 cm, are simulated with 30 AWG (0.254 mm-
diameter) copper material, and are attached to a lumped capacitor to
make them resonant. Optionally, a second receiving coil may be added
to improve the system robustness and performance (Coil 3 in Fig. 2),.
All simulations are carried out using the frequency-domain solver of the
CST Microwave Studio and are based on the Finite Integral Technique.
Tetrahedral mesh is used to discretize the computation domain. The
below graph depicts the trend of S21 obtained with respect to flexion

10
angle at the resonance frequency, as obtained by simulation.

11
 Notably, there is a one-to-one correlation between |S21 | and θf , con-
firming the feasibility of monitoring joint flexion via the proposed ap-
proach.

12
Chapter 5

Selection of Optimal Operation

Frequency

As mentioned earlier, the same coils may operate in different modes

(antenna, inductive, or combination of the two) depending on the se-
lected operation frequency. Different modes have different advantages
and disadvantages for monitoring joint flexion, implying a trade-off for
optimal frequency selection. In particular, operating frequency is iden-
tified as having significant effect on mainly three parameters, i.e.,
a) power requirements,
b) flexion angle resolution, and
c) tissue dielectric property variations.
A thorough study is hereafter presented that explores five representa-
tive frequencies (24, 34, 70, 230, and 927 MHz) with respect to the
aforementioned parameters. To do so, the simulation set-up of Fig. 2
is considered (Coils 1, 2 only). Specifically, at frequencies below 120
MHz, coils of radius 4 cm operate in the inductive mode (circumference
< 0.1Λ). The 927 MHz frequency corresponds to the self-resonance of

13
the coils where they behave as loop antennas, while 230 MHz corre-
sponds to an example frequency in the combined mode of operation.

A. Effect on Power Requirements

Low transmit power is essential for minimizing battery requirements

and ensuring conformance to safety standards. Use of resonant coils
improves transmission to a great extent, while optimal frequency se-
lection may further boost transmission efficiency. To demonstrate the
latter, Fig. below plots |S21 | as a function of flexion angle (θf ) at various
operating frequencies. As seen, the antenna mode (927 MHz) does not
show a one-to-one correlation between |S21 | and θf , unlike the induc-
tive and combined modes. Besides, the transmission coefficient, |S21 |,
exhibits the lowest values for this mode making it the least efficient in
terms of transmit power. With the above in mind, the antenna mode is
considered unsuitable for the intended application. While in the induc-
tive and combined modes, |S21 | decreases as the operating frequency
is reduced. This can be attributed to changes in impedance matching
at different frequencies. Notably, this trend is not linear on the dB
scale, implying that transmission efficiency significantly reduces as we
move down in frequency. That is, higher frequencies (in the inductive
or combined mode) are preferred so as to enable higher received power
levels or, equivalently, lower power requirements on the transmitter end.

14
 B. Effect on Flexion Angle Resolution
As would be expected, high angular resolution is desired for the de-
signed system, implying that the range of |S21 | values corresponding to
the extreme 0◦ and 130◦ flexion angles should be as large as possible.
To better understand how this parameter is affected by the operating
frequency, Fig. below plots the aforementioned |S21 | range (i.e., |S21 | at
|S21 | = 0◦ subtracted from |S21 | at |S21 | = 130◦ ) for operating frequen-
cies in the inductive and combined mode. As seen, lower frequencies (in
the inductive or combined mode) are preferred so as to enable higher

15
angular resolution.

C. Tissue Dielectric Property Variations

Human tissue properties (permittivity, ϵr , and loss tangent, tanδ)

vary from person to person and even from time to time for the same
individual. The proposed system must be ideally insensitive to such
changes in the underlying tissues.

The reason is that coils are inductively coupled via magnetic fields
and are not radiating. Since human tissue has a relative permeability
of µr∼1, magnetic coupling is not affected by the presence or absence
of tissue.

For optimal frequency selection, the effect of all three aforementioned

parameters needs to be taken into account. To ensure tolerance to tissue
dielectric property variations as well as high angular resolution, oper-

16
ation deep in the inductive mode is preferred. On the other hand, for
reduced power requirements on the transmit side, higher frequencies in
the inductive mode or even the combined mode are preferred. With this
trade-off in mind, we herewith select the 34 MHz operating frequency.

17
Chapter 6

Conclusion

There is continued work being done in order to further refine an exper-

imental prototype of the wearable sensor. Specifically, research is being
done to develop accurate and efficient post-processing models in order
to increase the precision of the detected flexion angles. Additionally,
the need for a flexible, unobtrusive power supply led to the invention
of a fully flexible electrochemical fabric. The electrochemical fabrics
are capable of producing scalable DC power, enough to unobtrusively
power the wearable sensor.

Contrary to state-of-the-art technologies used to monitor joint flex-

ion, the reported approach:
a) can be seamlessly integrated into garments for real-world monitor-
ing,
b) does not suffer from integration errors,
c) does not require line-of-sight, and
d) does not obstruct natural joint movement.
Also, cost of the constituent sensor materials (e-threads or copper wire,
connectors, and capacitors) is very low as compared to state-of-the-art

18
IMUs or optical cameras. Even more importantly, the reported ap-
proach is extremely robust to changes in human tissue dielectric prop-
erties as verified via simulations in canonical models, and brings forward
multiple degrees of freedom to optimize the design, per application re-
quirements.

19
Bibliography

[1] A. Salarian , ”Ambulatory monitoring of physical activities in pa-

tients with Parkinson’s disease”

[2] V. Bonnet , ”Monitoring of hip and knee joint angles using a single
inertial measurement unit during lower limb rehabilitation”

[3] H. Wagner, ”Upper-body kinematics in team-handball throw tennis

serve and volleyball spike”

[4] B. C. Heiderscheit, ”Reliability and between-limb symmetry of

joint kinetics and kinematics during a countermovement jump”

[5] M. El-Gohary and J. McNames, ”Shoulder and elbow joint angle

tracking with inertial sensors”

20