IEEE SENSORS JOURNAL, VOL. 30, NO.

8, JUN 15 2022 1

A Indoor Positioning System of Bluetooth AOA

Using Uniform Linear Array Based on Two-point
Position Principle
Wenzhao Shu, Shuai Wang

Abstract— In recent years, indoor positioning research has

received a lot of attention, and Bluetooth Low Energy
Angle of Arrival(AOA) positioning technology is one of the
hot technologies. This paper mainly regards the Bluetooth
AOA postiong system as the reasear object and proposes
1D-MUSIC algorithm combining the principle of Two-Point
Positioning. This algorithm can not only be applied to linear
antenna array, but also to rectangular antenna array and cir-
cular antenna array. This methed makes the best of the I/Q
data and improves the accuracy of position. By definitions,
we take two steps into experiment. First, the IQ samples
from Constant Tone Extension (CTE) can be obtained and
use the phase difference to get the angle.Then the posi-
tion of tag can be conclulated by the plane mathematical
relations. In this paper, we use a Uniform Linear Array to
solve the signal angle, Two-Point Positioning(TPP) is to
obtain the angle of two base stations, and uses the spatial
mathematical relationship between the two base stations
and the label to determine the position of a label. Finally, in
order to verify the feasibility of the positioning principle,
we conduct experiments in an open room, using three
development boards plus array antenna boards to build
a complete two-point positioning system. The numerical
values of the label positions of different places are collected, and the calculation error is within 0.5m to 1.5m by combining
the two-point positioning principle with the Multiple Signal Classification algorithm(MUSIC), and it is proved that the two-
point positioning is practically feasible.
Index Terms— Bluetooth low energy, angle of arrival, uniform linear array, IQ sample, MUSIC algorithm, two-point
positioning.

I. I NTRODUCTION advancement of technology, global satellite navigation systems

have been widely used in people’s lives. GPS is one of the

T HE indoor positioning technology can be divided into

indoor positioning and outdoor positioning according
to the positioning area. Outdoor positioning technology has
most famous global satellite navigation systems, other well-
known foreign navigation systems are Russia’s GLONASS,
the European Union’s GALILEO, China has also developed its
been developed for many years, and with the continuous own navigation system – BDS [1]. Although outdoor position-
ing technology has developed so well, most of human life is
This paragraph of the first footnote will contain the date on which you
submitted your paper for review. It will also contain support information, still mainly indoors, and indoor positioning is in great demand.
including sponsor and financial support acknowledgment. For example, The indoor positioning environment [2] is very complex, and
“This work was supported in part by the U.S. Department of Commerce the walls, the ground and various glasses will block and reflect
under Grant BS123456.”
The next few paragraphs should contain the authors’ current affilia- the signal. Therefore, it is not feasible to apply GNSS directly
tions, including current address and e-mail. For example, F. A. Author indoors. Although the indoor environment is complex and
is with the National Institute of Standards and Technology, Boulder, CO diverse, it is closely related to our lives. As a result, [27]
80305 USA (e-mail: [email protected]).
S. B. Author, Jr., was with Rice University, Houston, TX 77005 USA. [28]indoor positioning technology has also been developed
He is now with the Department of Physics, Colorado State University, very rapidly. In recent years, various wireless technologies
Fort Collins, CO 80523 USA (e-mail: [email protected]). have been applied in indoor positioning, including WIFI,
T. C. Author is with the Electrical Engineering Department, University
of Colorado, Boulder, CO 80309 USA, on leave from the National Re- RFID, Zigbee, UWB, BLE, and visible light [3]. Among these
search Institute for Metals, Tsukuba, Japan (e-mail: [email protected]). positioning technologies, the most interesting development
2 IEEE SENSORS JOURNAL, VOL. 30, NO. 8, JUN 15 2022

is Bluetooth low energy technology. Bluetooth low energy

technology is Bluetooth 4.0 version [5] after the only one, its
main feature is low power consumption, by using Bluetooth
sleep technology to make Bluetooth devices do not work when
in hibernation, waiting for work and then wake up, which can
greatly reduce energy loss. It is the use of this feature, the
general coin cell battery in the state of power supply can make
the Bluetooth device work for 3 to 5 years, this technology is
very convenient for the application of Bluetooth positioning
[4].
Bluetooth uses the ISM’s unrestricted 2.4GHZ band, where
signals such as WIFI also interfere with it [11]. The Bluetooth
channel length is about 80MHZ, according to the protocol, it
can be divided into 40 channels, each channel occupies 2MHZ.
According to the different channel functions, we can divide it
into broadcast channels and data channels, 37 data channels,
and 3 broadcast channels. Bluetooth AOA positioning adopts Fig. 1: BLE AOA postioning structure. The transmitter trans-
two modes of connection and connectionless [6], the connec- mits the signal at the same time as the receiver receives
tion mode is to exchange data information using the data chan- samples through the RF switch controlling antenna, and the
nel, in which the data exchange will use frequency hopping values sampled by the spatial spectrum algorithm are used to
technology, which can effectively reduce external interference. estimate the AOA.
Connectionless mode is a direct working broadcast state that
provides positioning capabilities for multiple labels.
There are two main principles of Bluetooth positioning, one this short introduction, the principle of the BLE AOA is
is to use the Received Signal Strength(RSS) [7], and the other introduced in Section II. Section III details the process of
is Direction of Arrival(DOA) [8]. The former is the use of the MUSIC algorithm. Section IV provides the TPP priciple
the received signal [23] in space as the path increases, while and giving solution expression of position. Section V presents
the signal strength continues to decline. According to this how to deal with IQ data and make position solution. Section
study, previous generations have summarized the mathematical VI shows the experiment of using boards to locate the la-
formula of this signal strength model. At the same time, we bel.Finally, Section VI makes a conclusion through this paper.
can also establish a database at different points in the space,
and use fingerprint data for data comparison to directly obtain II. AOA P RINCIPLE
the location information of unknown points in space. DOA is
Fig.1 represents the Bluetooth AOA system strcture. Blue-
to use different antennas to reach different antennas there will
tooth positioning mainly includes three parts: transmitter and
be a phase difference, through the spatial spectrum estimation
receiver, and antenna array with RF switch.
algorithm [9] to solve the two angles of the signal, one is
used to measure the pitch angle of the height, the other is to
measure the azimuth angle of the plane [22]. In this article, we
use real environment to conduct experiments, use IQ numerical
signal processing, and apply one-dimensional music algorithm
to truly apply AOA to the actual scene, which is not involved
by predecessors. At the same time, we also use two-point
positioning to provide a new exploration for building an indoor
positioning system.
In order to solve the position of the label, a localization
algorithm based on the principle of two-point localization is Fig. 2: CTE packet structure. The transmitter transmits the
proposed. Before applying the two-point calculation principle, BLE packet including CTE, which is the RF sinusoidal signal
we firstly use the MUSIC algorithm [10] to solve the azimuth in the end of packet. The CTE can be demoulded to IQ
of the tag. Through this, the system of positioning can be samples.
made by two process. First, the spatial coordinate system
of the respective positions of the tag and the base station
is constructed, and it is assumed that the two base stations
are just located on the two coordinate axes. Then, according A. CTE and IQ sample
to the trigonometric function, the relationship between the The packet structure of CTE [13] is shown in Fig.2. CTE
known azimuth and the coordinate is established by using the is short for Constant Tone Extensions, it is a set of exten-
trigonometric function expression, and the position of the label sion packets added to the Bluetooth broadcast for Bluetooth
is calculated by solving the matrix equation at last [12]. direction finding, the length of the packets varies from 16
The remainder of the paper is organized as follows: After µs to 160 µs. A CTE is a series of unwhitened all ones. The
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2022) 3

CTE package is mainly divided into three parts: the protection

period, the reference period and the switch-sample period
[14]. The first two period are fixed, the guard period takes
4 microseconds, and the reference period takes 8 microsec-
onds. The length of the switching sampling period completely
depends on the length of array antenna, and the sample
period can be divided into 1 microsecond and 2 microseconds
according to the regulations. The time of the switching period
and the sampling period is equal, and the sampling is switched
once. For example, assuming that the sampling antenna array
is 12 and the sampling period is 1 microsecond, the length of
the switching sampling period is 24 microseconds.
(a) Nordic antenna array of 12. (b) silab antenna array of 16

Fig. 4: Two types of antenna array. Both antenna arrays have

their RF Switch. Through Antenna Parttern, corresponding
antenna can be controlled and then sampling the IQ values.

B. Antenna Array
The IQ samples are switched by the antenna array and then
divided into two values according to quadrature and in-phase.
The antenna array is an array antenna group composed of
multiple single antennas. Acoording to the difference of shape,
the antenna arrays can be divided into linear array and plane
array. A linear array is a set of antennas arranged at equal
intervals. Because it is one-dimensional, it is generally used
for plane angle estimation calculation. The other is the plane
array. The common plane arrays are mainly divided into two
types, circular arrays and square arrays. Plane arrays are 2-
dimensional and can be used to estimate azimuth and elevation
angles. The Fig.4 shows the antenna boards of Nordic and
Fig. 3: The relation of original signal and IQ sample. The Silab [16], both of which are square array antenna boards.
original signal is devided into In-phase and Quadrature values. The difference is that Nordic has removed four antennas in
The amplitude and phase of signal is can be expressed by I and the middle, but from the calculation point of view, there is no
Q, the phase difference is conculate by two antenna’s phase. difference between the two.
The antenna board is controlled by a radio frequency switch,
CTE can be divided into AOA and AOD according to the and its logic is a gate circuit in the circuit, which can
type. According to the provisions of Bluetooth protocol 5.1, be combined by 0 and 1 binary to achieve the purpose of
AOA is selected by default in the configuration, and AOD is controlling multiple antennas. The Fig.4 shows the antenna
generally divided into two types: AOD 1 µs and AOD 2 µs. switching, which is controlled by four IO ports. The four IO
The CTE is decomposed into I (In-phase) and Q(Quadrature) ports can control 16 digital states according to different logical
signals after passing through the antenna array according to combinations. Each digital state corresponds to an antenna
Fig.3. Its mathematical relationship can be expressed by the switching switch. By Fig.5, the antenna switching control and
equation 1 and 2 below, the amplitude of the signal is the its lead the logic level of the pin can be displayed.
sum of the squares of the two signals, and the angle is the
tangent ratio of the Q-channel signal and the I-channel signal. C. AOA Conculation
Assuming that the expression of the signal is A sin(θ), and
IQ is the value of its two quadrature signals, the following AOA Conculation is defided by the priciple of phase differ-
formula can be obtained from the above [13]: ence(PD) [17]. Through, the signal’s phase of antena arrary
p can be conculated. Assuming that the signal is a plane wave
A = I 2 + Q2 (1) and the array antenna is a linear array, since the time it takes to
reach different antenna arrays is different, there will be a phase
Q difference between the two adjacent antennas. The specific
θ = arctan (2)
I representation can be seen in the figure below. First, the
In the above formulas, A represents the amplitude of the respective phases of antenna 1 and antenna 2 can be calculated
signal, θ represents the phase angle of the signal, whose unit according to the above. The signal reaches antenna 2 at a
is angle, and I and Q respectively represents the I/Q value of distance from antenna 1. The distance is calculated according
the two signals after sampling. to the electromagnetic wave theory. Then, according to the
4 IEEE SENSORS JOURNAL, VOL. 30, NO. 8, JUN 15 2022

Fig. 5: Logical level of different antennas. The one indicate

that the serial number of antenna is used. From bluetooth core
5.1, this RF switch is contrlled by Antenna Parttern.

triangle relationship, the angle and distance can be established

by using the sine function [18] to obtain the following formula:
φλ
D cos(θ) = (3)
2π
Finally the above formula is deformed, and the angle value Fig. 6: The model of AOA in URA.The model is made of 16
is solved by the inverse cosine function. Available: antennas, it has 4 rows and 4 columns.Every blue rectangle is
  on behalf of antenna, the antenna is switched and the phase
φλ
θ = arccos (4) difference can be conclulated.
2πD
Among the founula, θ is the angle of received signal, D is
the distance of two antennas, λ represents the wavelength of and N represents the number of columns of array elements
received signal, φ mean the antennas’s phase angle difference. on the y-axis. The reference point selects the origin, and
the distance of the first antenna [19] on the x-axis relative
III. AOA ALGORITHM to the first antenna is (k − 1)dx. So signal direction vector
In order to explain the AOA algorithm, we put forward the can be represented by ax (θ, ϕ) = e(−j2π(k−1)dx sin(θ) cos(ϕ)) ,
following assumptions: Similarly, for y axis, ay (θ, ϕ) = e(−j2π(k−1)dy sin(θ) sin(ϕ)) .
1. There are generally multiple signal sources, but the signal Above all, the next formula cab be ratiocinated:
source (ie, label) used in this paper is regarded as one. ax (θ, ϕ) = [1, · · ·, e(−j2π(M −1)dx sin(θ) cos(ϕ)) ]T (5)
2. The signal is a narrowband signal, and the distance D
between the antennas is less than the wavelength of the signal ay (θ, ϕ) = [1, · · ·, e(−j2π(N −1)dy sin(θ) sin(ϕ)) ]T (6)
λ
. The signal source is one, so the direction matrix of the array
2
3. There is no external interference, such as multipath and element and the direction vector of its array element are equal.
emission effects caused by objects such as walls and glass. Supposing that Ax and Ay are used to denote the orientation
The noise is additive Gaussian white noise. matrices in the x-direction and y-direction, respectively. The
Fig.6 represents the mathematical model of AOA estimation, following formula can be obtained [20]:
where the two angles θ and ϕ represent the pitch angle and Ax (θ, ϕ) = [1, · · ·, e(−j2π(M −1)dx sin(θ) cos(ϕ)) ]T (7)
the plane angle, respectively. dx and dy represent the spacing
(−j2π(N −1)dy sin(θ) sin(ϕ)) T
between the antennas in the x-axis and y-axis, and the small Ay (θ, ϕ) = [1, · · ·, e ] (8)
blue rectangles represent the antennas.
For the x-axis subarray, the received signal is:
A planar array is a multidimensionalization of a linear
array, so we decompose this model into linear arrays in both xx (t) = Ax s(t) + nx (t) (9)
directions. Since the x-direction and y-direction arrays are
For the y-axis subarray, same as x-axis:
similar, only one direction can be analyzed. Assuming that
the mathematical expression of the signal is A sin(φ), the
xy (t) = Ay s(t) + ny (t) (10)
actual calculation can be calculated by the IQ value. The
type of antenna array is Uniform Rectangle Array(URA) with The nx (t) and ny (t) represent separately the AWGN of x-
M ×N , M represents the number of array rows on the x-axis, axis and y-axis.
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2022) 5

The whole Output signal x(t) are expressed by xx (t) and

xy (t), as follow:

x(t) = [Ax ⊙ Ay ]s(t) + n(t) (11)

Among them, Ax ⊙ Ay represents the Khatri-Rao product
[21] of Ax and Ay . According to the formula, we get: Ax ⊙
Ay = Ax ⊗ Ay , the ⊗ represents the Kroecker product of two
matrix, Calculate the above equation directly:

Ax 1, · · ·, e(−j2π(N −1)dy sin(θ) cos(ϕ)) , · · ·

 

⊗ =  e(−j2πdx sin(θ) sin(ϕ)) e(−j2π(M −1)dx sin(θ) sin(ϕ)) , · · ·,

Ay e(−j2π((M −1)dx sin(θ) sin(ϕ)+(N −1)dy sin(θ) cos(ϕ)))
(12)
Ax ⊗ Ay is M N × 1 matrix. Through Equation (12), the
covariance matrix Rxx of the signal can be constructed. Ac-
cording to the spatial spectrum algorithm, the signal subspace
and the noise subspace are orthogonal. By decomposing the
covariance matrix and sorting according to the size of the
eigenvalues, the covariance matrix can be expressed as a signal
Fig. 7: The construction strpriciple of Two-Point positioning.
subspace and the noise subspace.
The two Base stations can get the angles from the same
Rxx = E[x(t) ∗ x(t)H ] (13) Tag, Then by trigonometric function the tag position can be
X X conculated.
Rxx = US US + UN UN (14)
S N
Where E means the mean of matrix, H represents the
For base station 2, the same expression can be gotten.
Conjugate transpose of matrix, US and UNPindicates
P separately
the signal and noise space eigenvector, and is on the y − y2
S N tanϕ = (17)
behalf of separately the signal and noise space eigenvalue. x − x2
1 Expand the above formulas, the left side to the coordinate
PM U SIC = H H A(θ, ϕ)
(15) value of the label can be isolated, and the right side is the
A(θ, ϕ) UN UN
value of the base station coordinate, which can be written as
Spectral function is expressed by Equation (15), then using the following matrix equation:
the space search method finds the maximum value of the
spectral peak, the corresponding angle value is the direction
    
1 − tan θ y y − x1 tan θ
of the target signal. For linear arrays, only one direction is = 1 (18)
1 − tan ϕ x y2 − x2 tan ϕ
considered, and the rest of the calculation is the same as
the two-dimensional algorithm, so the derivation will not be Then use the inverse of the matrix, the position of tag can be
carried out here. conculated. The Equation 19 is presentation of tag’s position
coordinate.
IV. T WO -P OIMT P OSITIONING     −1
y y − x1 tan θ 1 − tan θ
According to the previous analysis, we can use the AOA = 1 (19)
x y2 − x2 tan ϕ 1 − tan ϕ
algorithm to calculate the angle value. According to the above
When the base station is not in a special location, it is also
Fig.7, it is assumed that the angle between the first base station
possible to establish two equations based on the relationship
and the tag is θ, and its coordinates in the figure are (x1 , y1 ).
between the coordinates established by the tag and the base
Similarly, for another base station, its angle with the tag is
station, using quadrilateral calculations, and then using the
ϕ, its coordinates are (x2 , y2 )(suppose both angle values are
mathematical relationship to solve them. The rest will not be
not equal to 90 degree). Since the base station is installed, its
elaborated upon.
coordinates are known. Assuming that the coordinates of the
label to be found are (x, y), according to the trigonometric
function relationship [24], the expression of its coordinates V. DATA P ROCESSING AND P SITION C ONCULATION
and the known angle can be established. In order to simplify This chapter is to use the development board for exper-
the calculation, the base station is placed on the y-axis and iments. The development board we use is the Nrf52833dk
the x-axis respectively in this experiment. development board from Nordic in Fig.8. The antenna board
For base station 1, the relationship can be established adopts Nordic’s rectangular antenna board with 12 antennas.
according to the tangent value, as follows: This elaboration is mainly through two aspects of processing
y − y1 [25]. On the one hand, the original acquired IQ value is
tanθ = (16) normalized and digitally processed, and converted into signal
x − x1
6 IEEE SENSORS JOURNAL, VOL. 30, NO. 8, JUN 15 2022

TABLE I: The phase increasing algorithm

The phase increasing algorithm

def function phase-increasing[φref ]
count = 0;
for i in range(2, len(φref )):
if φref (i) ≤ φref (i−1):
count += 1
Fig. 8: Nordic nrf52833dk. The board is designed by Nordic φref (i) = φref (i−1)+count*2π
company to using Direction Finding, which has strong Tx
power and Rx sensibility. There is demo used to get the IQ else
sample by serial port. continue

amplitude and phase. The other is to use the phase to calculate

the formula, and it is discussed separately at this time. Here
the angle through the one-dimensional MUSIC algorithm, and
Fig.9, according to the relationship between different IQ values
then use the two-point positioning algorithm [26] to calculate
and zero, they can be classified, and the derived formula is as
the position.
follows:

Q


 arctan + π (I < 0)

 I

 Q
arctan (I > 0, Q ≥ 0)


I



φ(Q, I) = arctan + 2πQ
(I < 0, Q < 0) (20)

π
 I

 (I = 0, Q > 0)
2


 3π


 (I = 0, Q < 0)
2
When all IQ value phase values have been calculated, the
phase must also be corrected. Since the original signal is a
sinusoidal signal, the period is exactly 2µs. The calculated
value of the next antenna sampling should be larger than the
previous one. For this, we summarize a calculation algorithm
[15], as shown in the Table.I. This method is used to process
the above data to obtain the final phase difference. According
Fig. 9: The original IQ samples. The blue line and red line to the above calculation, 38 groups of all phase values that
show the IQ sample from the diffrent antenna, whose values increase according to the specified phase can be obtained.
change with antenna switch. The IQ sample are used to At the same time, the amplitude of the collected signal is
conculate the phase and amplitude. calculated.

B. Angle Conclusion and Position Solution

A. Data processing According to the formula, it can be known that the spacing
Based on the two point positioning, we just need to use a of the antenna elements is very important. Therefore, we use
linear array. According to the Fig.4, we will use the antennas the center of the reference antenna as the origin to establish a
11, 12, 1, 2. The antenna 12 is abbreviated as A12 , and the coordinate system. According to the antenna technical manual
others are similar. A11 is the reference antenna, A12 , A1 , A2 of nordic company, it can be known that the antenna spacing
are all on the sampling antenna and the CTE length is 72 µs. D is 40 mm. It can be known that the x coordinates of
The figure below is the sampled IQ sample, according to the the antennas 11, 2, 1, and 2 are 0, 0.04, 0.08, 0.12. At the
software definition, the cycle period is 60 µs, and the number same time, the obtained phase and amplitude are respectively
of sampling cycles is 10 times. As mentioned above, the phase recorded as φ1 , · · · , φ38 , according to the number, and the
can be calculated with , but due to the angle limitation of the amplitude is Am1 , · · · , Am38 . The calculation is performed
tangent value, when the angle is located in the first quadrant by using the sampling antenna, the first reference antenna
and the third quadrant, the tangent value of the calculated samples are culled. Then, the extraction is performed in the
angle is the same, but the angle is different. At the same time, same antenna in a loop, and the phase and amplitude vectors
if the value of I is zero, it cannot be calculated directly with are constructed respectively.
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2022) 7

between the maximum and minimum angles is between 3 and

φall = [φ1 , φ2 , · · ·, φ37 , φ38 ] (21) 10 degrees, indicating that the algorithm can solve the azimuth
angle between the tag and the base station, and the accuracy
Amall = [Am1 , Am2 , · · ·, Am37 , Am38 ] (22) is still very efficient.
The first antenna sample starts from serial number 9, and the
numbers of loop antenna is 3. According to the same antenna
data we can get (for example the first antenna):
φ12 = [φ9 , φ12 , φ15 , φ18 , φ21 , φ24 , φ27 , φ30 , φ33 , φ36 ]T
(23)
Am9 , Am12 , Am15 , Am18 , , Am21 T
Am12 = [ ] (24)
, Am24 , Am27 , Am30 , Am33 , Am36
Using the amplitude and phase obtained above, we construct
the signal s(t) = Ae−j∗ϕ . Based on the first antenna, the
steering vector of the antenna can be gotten. Assuming that the
plane angle is θ, it can be known that its vector is e−j∗2π0.04 .
It can be seen that the output matrix of the first antenna is:
x1 = [Am9 e−jφ9 e−j0.04π , · · ·, Am36 e−jφ36 e−j0.04π ]T (25)

Am9 e−jφ9 e−j0.04π , · · ·, Am36 e−jφ36 e−j0.04π

 

x =  Am10 e−jφ10 e−j0.08π , · · ·, Am37 e−jφ37 e−j0.04π 

Am11 e−jφ11 e−j0.08π , · · ·, Am38 e−jφ38 e−j0.04π
(26)
In the same way, other antennas are processed, and the final
output matrix is a 10 × 3 matrix, which is then averaged,
and then singular value decomposition is performed. We go to
the noise vector-its last row of covariance matrix. Using the
spectral function, we assume that the angle varies from 0 to
180 degrees, and the direction angle can be solved by finding
the maximum value using extreme value search.

Fig. 11: The experiment model. This model are made of

two base stations and one tag. The origin is established on
the connection between the two base stations, changing the
position of tag. The position coordinate can be conculate by
TPP.

VI. E XPERIMENT R ESULT

As shown in the Fig.11 below, we conducted the experiment
in the center of a room with a length of 12 meters and a
width of 8 meters. We placed two base stations on the y-axis
and the x-axis respectively, with a distance of 0.4 m from
the origin. Both base stations have antennas, the ones without
Fig. 10: The angle between the true position and solution by antennas are tags. We set the label placement points as (0.52,
algorithm. The true angle are 42,58,106,138,169 degree, the 0.56), (1.60, 1.90), (2.34.2.69), and then use the implemented
angle of solution are 46,53,110,148,172 degree. The diffrence algorithm to position the same point, and the solution sampling
is not big. is 5 times, and the obtained data is averaged. The positions
obtained by solving are (0.43, 0.49), (1.29, 1.75), (2.08, 2.79).
Here, five positions are selected and their angles are solved The obtained data will be converted into the following Fig.12.
using the one-dimensional MUSIC algorithm. The above From the experiments conducted, the one-dimensional MU-
Fig.10 is its real position angle and the angle solved by the SIC algorithm and the two-point positioning algorithm can
algorithm. It can be seen from the algorithm that the difference effectively calculate the position of the label, and the error
8 IEEE SENSORS JOURNAL, VOL. 30, NO. 8, JUN 15 2022

range, whether in the x-coordinate or the y-coordinate, does very efficient and the calculation accuracy is also very high.
not exceed 2 m. In addition, the calculation of indoor positioning can be
At the same time, due to the accuracy of the one- achieved by using two-point positioning again, which was not
dimensional MUSIC algorithm calculation, the corresponding available before. The proposed algorithm can effectively solve
angle value is calculated, and then the position of the label the position problem of spatial tags. Two-point positioning is
can be calculated through two points. very reliable for calculating the plane position, and its accuracy
is high and the error is small. In a word, this article is of great
reference significance for the application of Bluetooth AOA
positioning.

R EFERENCES
[1] Z. Wang, Z. Yang, and T. Dong, “A review of wearable technologies for
elderly care that can accurately track indoor position, recognize physical
activities and monitor vital signs in real time,” Sensors, vol. 17, no. 2,pp.
341-346, 2017.
[2] L. Kajdocsi, A. Dörömbözi and J. Kovács, “Development of Bluetooth
Mesh Core Stack using OmNET++,” 2019 IEEE 17th International Sym-
posium on Intelligent Systems and Informatics (SISY), Subotica,Serbia,
2019, pp. 23-28.
[3] B. Friedlander, “A sensitivity analysis of the MUSIC algo-
rithm,”IEEETrans. Acoust., Speech Signal Process., vol. 38, no. 10, pp.
1740–1751,Oct. 1990.
[4] N. B. Suryavanshi, K. V. Reddy, V. R. Chandrika, “Direction Finding
Capability in Bluetooth 5.1 Standard,” in International Conference on
Ubiquitous Communications and Network Computing, Springer,Cham,
pp. 53-65, 2019.
[5] T. Wang, H. Zhao, and Y. Shen, “An efficient single-anchor localization
Fig. 12: The position between the true position and solution method using ultra-wide bandwidth systems,” Appl. Sci., vol. 10, no. 1,
by TPP. There are 3 group position estimated. The closer the Jan. 2020.
[6] Z. Liu, W. Dai, and M. Z. Win, “Mercury: An infrastructure-free
tag is to the base station, the position caculation are more system for network localization and navigation,” IEEE Trans. Mobile
accurate. Even so, the errors are no more than 2 m. Comput.,vol. 17, no. 5, pp. 1119–1133, May 2018.
[7] S. Mehryar et al., “Belief condensation filtering for rssi-based state esti-
mation in indoor localization,” in Proc. IEEE Int. Conf. Acoust.,Speech
and Signal Process., May 2019, pp. 8385–8389.
[8] P. Zand et al., “A high-accuracy phase-based ranging solution with Blue-
VII. CONCLUSION tooth Low Energy (BLE),” in Proc. IEEE Wireless Commun. Netw.Conf.,
As a new indoor positioning technology, Bluetooth AOA Marrakech, Morocco, Apr. 2019, pp. 1–8.
[9] S. Mazuelas, Y. Shen, and M. Z. Win, “Spatiotemporal information
has been recognized by the Bluetooth Alliance. Bluetooth coupling in network navigation,” IEEE Trans. Inf. Theory, vol. 64,no.
AOA positioning technology can achieve high positioning 12, pp. 7759–7779, Dec. 2018.
accuracy, low cost and low power consumption. Based on [10] N. H. Nguyen, K. Do˘ganc¸ay, and E. E. Kuruoglu, “An iteratively
reweighted instrumental-variable estimator for robust 3-D AOA local-
the Bluetooth positioning technology, this article proposes ization in impulsive noise,” IEEE Trans. Signal Process., vol. 67, no.18,
a system for calculating the position of the tag from the pp. 4795–4808, Sep. 2019.
improved MUSIC algorithm combined with the principle of [11] S. Monfared, T. Nguyen, L. Petrillo, P. De Doncker and F. Hor-
lin,“Experimental Demonstration of BLE Transmitter Positioning Based
two-point positioning. From the experimental results, the use on AOA Estimation,” 2018 IEEE 29th Annual International Sympo-
of two-point positioning to solve the tag position is very sium onPersonal, Indoor and Mobile Radio Communications (PIMRC),
efficient, and the time required for calculation is very less. Bologna,2018, pp. 856-859.
[12] Marco Cominelli, Paul Patras, and Francesco Gringoli. 2019. Dead on
Besides this, the most important thing is that its positioning Arrival: An Empirical Study of The Bluetooth 5.1 Positioning System.In
error is also very small, compared with other positioning Proceedings of the 13th International Workshop on Wireless Network
technologies, so this system is a Bluetooth indoor positioning Testbeds, Experimental Evaluation and Characterization (WiNTECH
’19).Association for Computing Machinery, New York, NY, USA,
system worthy of application. First, a principle about AOA is 13–20.
proposed. The structure of the CTE packet, the relationship [13] Mikael Q Kuisma. Whiteboard Web:I/Q Data for
between IQ sampling and the final signal are introduced in Dummies.Available:http://whiteboard.ping.se/SDR/IQ.
[14] M. Wooley. (2019). Bluetooth Direction Finding: A Tech-
detail, and the amplitude and phase of the calculated signal are nical Overview. Bluetooth Resources. [Online]. Available:
explained from the source. Then, we deduce the algorithm flow https://www.bluetooth.com/bluetooth-resources/bluetooth-direction-finding/
of MUSIC through the Bluetooth AOA calculation principle, [15] H. Ye, B. Yang, Z. Long and C. Dai, ”A Method of Indoor Positioning
by Signal Fitting and PDDA Algorithm Using BLE AOA Device,” in
construct the signal from the IQ value, then construct the IEEE Sensors Journal, vol. 22, no. 8, pp. 7877-7887, 15 April15, 2022.
antenna steering vector and solve the output signal, and finally [16] Q. Yan, J. Chen, G. Ottoy, and L. D. Strycker, “Robust AOA
decompose the eigenvalue of the output signal matrix, and use based acoustic source localization method with unreliable measure-
ments,”Signal Process., vol. 152, pp. 13–21, Nov. 2018.
the spectral function to solve the direction angle.Using two- [17] W. Xiong, C. Schindelhauer, H. C. So, D. J. Schott, and S. J.
point positioning to bring in the solved angle, the position Rupitsch,“Robust TDOA source localization based on Lagrange pro-
coordinates of the label can be solved. The whole system gramming neural network,” IEEE Signal Process. Lett., vol. 28, pp.
1090–1094,2021.
realizes the original DOA estimation from theory to practical [18] T.-K. Le and K. C. Ho, “Joint source and sensor localization by angles
application, and the proposed AOA calculation algorithm is of arrival,” IEEE Trans. Signal Process., vol. 68, pp. 6521–6534, 2020.
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2022) 9

[19] S. Wu, S. Zhang, K. Xu, and D. Huang, “Probability weighting

localization algorithm based on NLOS identification in wireless net-
work”,Wireless Commun. Mobile Comput., vol. 2019, Art. 2707469.
[20] A. Gabbrielli, W. Xiong, D. J. Schott, G. Fischer, J. Wendeberg,
F.H¨oflinger, L. M. Reindl, C. Schindelhauer, and S. J. Rupitsch,
“Anecho suppression delay estimator for angle of arrival ultrasonic
indoor localization,” IEEE Trans. Instrum. Meas., vol. 70, pp. 1–12,
2021.
[21] A. Alma’aitah, B. Alsaify, and R. Bani-Hani, “Three-dimensional em-
pirical AoA localization technique for indoor applications,” Sensors,vol.
19, no. 24, p. 5544, Dec. 2019.
[22] X. Qiu, B. Wang, J. Wang, and Y. Shen, “AOA-based BLE local-
ization with carrier frequency offset mitigation,” in Proc. IEEE Int.
Conf.Commun. Workshops (ICC Workshops), Jun. 2020, pp. 1–5.
[23] Z. Hajiakhondi-Meybodi, M. Salimibeni, K. N. Plataniotis, and A.
Mohammadi, “Bluetooth low energy-based angle of arrival estimation
via switch antenna array for indoor localization,” in Proc. IEEE 23rd
Int. Conf. Inf. Fusion (FUSION), Jul. 2020, pp. 1–6.
[24] M. Al-Sadoon et al., “A new low complexity angle of arrival algo-
rithm for 1D and 2D direction estimation in MIMO smart antenna
systems,”Sensors, vol. 17, no. 11, p. 2631, Nov. 2017.
[25] C. Huang, Y. Zhuang, H. Liu, J. Li, and W. Wang, “A performance
evaluation framework for direction finding using BLE AoA/AoD re-
ceivers,”IEEE Internet Things J., vol. 8, no. 5, pp. 3331–3345, Mar.
2021.
[26] P. Zand, J. Romme, J. Govers, F. Pasveer, and G. Dolmans, “A
highaccuracy phase-based ranging solution with Bluetooth low energy
(BLE),” in Proc. IEEE Wireless Commun. Netw. Conf. (WCNC),Apr.
2019, pp. 1–8.
[27] G. Avitabile, A. Florio and G. Coviello, “Angle of Arrival Estimation
through a Full-Hardware Approach for Adaptive Beamforming,” in IEEE
Transactions on Circuits and Systems II: Express Briefs.
[28] S. Monfared, A. Delepaut, M. Van Eeckhaute, P. De Doncker and
F.Horlin, “Iterative Localization Method Using AoA for IoT Sensor Net-
works,” 2019 IEEE 89th Vehicular Technology Conference (VTC2019-
Spring), Kuala Lumpur, Malaysia, 2019, pp. 1-6.