WO2018160141A1 - Appareil et procédé de localisation et/ou de suivi - Google Patents

Appareil et procédé de localisation et/ou de suivi Download PDF

Info

Publication number
WO2018160141A1
WO2018160141A1 PCT/SG2018/050101 SG2018050101W WO2018160141A1 WO 2018160141 A1 WO2018160141 A1 WO 2018160141A1 SG 2018050101 W SG2018050101 W SG 2018050101W WO 2018160141 A1 WO2018160141 A1 WO 2018160141A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
parameters
path
estimator
tof
Prior art date
Application number
PCT/SG2018/050101
Other languages
English (en)
Inventor
Yaxiong XIE
Mo Li
Original Assignee
Nanyang Technological University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanyang Technological University filed Critical Nanyang Technological University
Publication of WO2018160141A1 publication Critical patent/WO2018160141A1/fr

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
    • G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
    • G01S5/0273—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves using multipath or indirect path propagation signals in position determination
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S3/00—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received
    • G01S3/02—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using radio waves
    • G01S3/74—Multi-channel systems specially adapted for direction-finding, i.e. having a single antenna system capable of giving simultaneous indications of the directions of different signals

Definitions

  • the present invention relates to an apparatus and method for localisation and/or tracking.
  • the present invention relates to estimation of parameters of indoor wireless signals for localisation and tracking.
  • Wireless localisation e.g. using radio frequency (RF) waves
  • RF radio frequency
  • Wireless sensing has been used for gesture recognition, human activity recognition and RF imaging. Both areas have a number of important applications including security (e.g. for detecting an intruder to private spaces), elderly care (e.g. generating an alert when an elderly person is behaving abnormally or has fallen down) and retail (e.g. analysing customer behavior in shops).
  • Wi-Fi based systems stand out as particularly promising due to the pervasive availability of Wi-Fi enabled devices.
  • Billions of Wi-Fi access points (APs) have already been deployed worldwide and almost all mobile devices like cell-phones, tablets, laptops and so on, are equipped with Wi-Fi chipsets.
  • APs Wi-Fi access points
  • Making use of existing Wi-Fi infrastructures saves the need for designing a dedicated hardware and hence facilitates real-world deployment.
  • wireless signals When transmitted to the physical environment, wireless signals travel along different paths between a transmitter (also referred to as an emitter or a sender) and a receiver. Along each path, the signals can be reflected by a number of objects (i.e. reflectors) located in the propagation environment and hence the signals from each path arrive at the receiver from different angles, at different times, which results in amplitude and phase variations of the received signal. This phenomenon is referred to as a multipath effect.
  • the multipath effect links the wireless transmitter, the physical environment and the receiver together.
  • Multipath parameters e.g. angle-of-arrival (AoA), angle-of- departure (AoD), time-of-flight (ToF), Doppler shift and amplitude
  • a line of sight (LoS) path directly connects the transmitter and the receiver.
  • a reflection path connects the reflectors with the transmitter/receiver pair and allows the localisation of the reflector once the parameters of the reflection path are accurately estimated.
  • a Doppler shift provides information about a radio velocity (i.e. motion) of the transmitter or the reflector with respect to the receiver. Accordingly, motion information together with location information enables applications like human- computer-interaction (e.g. gesture recognition), intruder detection, indoor floor map reconstruction, and so on.
  • an apparatus for localisation and/or tracking comprising:
  • a receiver for receiving a received signal, the received signal comprising a superposition of multiple signals derived from a wireless signal transmitted from a transmitter along a plurality of different paths; a computer processor; and a data storage device, the data storage device comprising instructions operative by the computer processor to separate the received signal into the multiple signals by: a) estimating initial path parameters comprising:
  • step (i) and (ii) repeating step (i) and (ii) using the re-estimated background noise as the initial background noise until the parameters for each signal path satisfy a convergence criteria;
  • embodiments of present invention provide an apparatus for localisation and/or tracking which is able to estimate initial path parameters and refine these parameters using an iterative technique so as to obtain more and more accurate results.
  • the apparatus can be used to resolve each individual signal path and to accurately estimate a number of parameters from the resolved signals.
  • the present apparatus overcomes fundamental accuracy limitations of the prior art such as those dictated by antenna count (for AoA methods) and frequency bandwidth (for ToF methods). Instead, finer resolution is achieved by jointly exploring more properties (i.e. signal dimensions) of the received signal. With multiple signal dimensions jointly estimated, two signals are resolvable if they are resolvable in any individual dimension, which increases the effective system resolution without requiring changing of the resolution in each individual dimension.
  • the estimated parameters will be used in wireless systems for localisation and/or tracking.
  • Particular applications include security (e.g. for detecting an intruder), elderly care (e.g. generating an alert when an elderly person is behaving abnormally or has fallen down) and retail (e.g. analysing customer behavior in shops).
  • the parameters may be multi-dimensional. In some embodiments, the parameters are at least 2-dimensional, at least 3-dimensional or at least 4-dimensional.
  • the parameters may be selected from the list comprising: Angle of Arrival (AoA), Angle of Departure (AoD), Time of Flight (ToF), Doppler Shift and Attenuation.
  • AoA Angle of Arrival
  • AoD Angle of Departure
  • ToF Time of Flight
  • Doppler Shift Attenuation.
  • the estimator tool may employ a maximal likelihood based method.
  • the estimator tool may be a multi-dimensional estimator comprising one or more of: an AoA estimator, an AoD estimator, a Doppler estimator, a ToF estimator and an amplitude estimator.
  • the estimator tool may estimate the parameters individually or jointly (i.e. together).
  • the estimator tool may jointly estimate the parameters for each signal path in a sequential manner.
  • the estimator tool may be flexible and may be configured to jointly estimate an arbitrary number of parameters.
  • the number (L) of signal paths may be derived from a number of peaks in an output signal from the estimator tool.
  • the parameters for each signal path may be derived from locations of peaks in an output signal of the estimator tool.
  • the initial background noise may be estimated by re-constructing each signal path with the estimated parameters and subtracting the re-constructed signal paths from the received signal.
  • known technologies from information theory may be employed to estimate the initial path parameters.
  • multiple one dimensional estimator tools may be employed to estimate multidimensional parameters.
  • a one dimensional multiple signal classification (MUSIC) estimator tool [reference 27] may be employed to estimate AoA and then the estimated AoA may be used to estimate the ToF using another MUSIC estimator tool and so on.
  • MUSIC multiple signal classification
  • Known two dimensional estimators such as SpotFi [reference 17] or Joint Angle and Delay Estimation (JADE) [reference 52] may be used as the estimator tool in some embodiments.
  • the receiver may comprise a plurality of antennas and the AoA estimator may comprise a beamformer configured to remove phase differences between signals of a same transmission timeslot arriving at different antennas in the receiver, summing resulting signals and scanning possible incidence angles to determine an angle of incidence generating a maximum beamformer output.
  • the AoD estimator may comprise a beamformer configured to apply a phase shift corresponding to possible departure angles to the sum of the resulting signals from the AoA estimator for each transmission time slot, to sum these signals, to determine an angle of departure generating a maximum beamformer output and to remove the applied phase shifts.
  • the Doppler estimator may comprise applying different Doppler shifts to an output from the AoD estimator and correlating this signal with the transmitted wireless signal to identify a Doppler shift producing a closest correlation and providing an output signal using the Doppler shift identified.
  • the ToF estimator may comprise correlating an output signal from the Doppler estimator with a delayed version of the transmitted wireless signal to determine the ToF.
  • the apparatus may apply successive interference cancellation by:
  • step (1) and (2) repeating step (1) and (2) using the residual signal as the received signal until the residual signal has a power below a predefined threshold.
  • the number (L) of signal paths may be estimated as the number of the reconstructed strongest signal paths.
  • the estimation of more than one parameter may comprise fixing all but one of the parameters and searching for an optimal value for the one parameter.
  • the optimum value may be determined using an expectation maximum (EM) model or a generalized expectation maximum (GEM) model.
  • EM expectation maximum
  • GEM generalized expectation maximum
  • the apparatus may be further configured to determine a direct signal path by identifying the signal path with a shortest ToF and/or a highest amplitude and/or a smallest AoA and AoD difference, and to calculate a difference between the ToF of the direct signal path and a ground-truth ToF to determine a synchronization offset and to correct the ToF measurements for all signal paths by the synchronization offset.
  • the apparatus may be further configured to correct for local oscillator offset.
  • the apparatus may be further configured to correct for carrier frequency offset.
  • the apparatus may be further configured for passive localisation and tracking and may further comprise a transmitter for transmitting the wireless signal along the plurality of different paths.
  • the receiver and the transmitter may be positioned on a ceiling to ensure a direct path between them exists.
  • the apparatus may be further configured to detect a reflection signal from a mobile object by identifying a signal with a non-zero Doppler shift and to use the AoA, AoD and ToF to track a motion of the mobile object.
  • the apparatus may be further configured for active localisation and tracking wherein a transmitter on the target transmits the wireless signal along the plurality of different paths.
  • the apparatus may be further configured to determine a direct signal path by identifying the signal path with a shortest ToF and/or a highest amplitude and/or a smallest AoA and AoD difference, and to localise the transmitter by using the ToF of the direct signal path to estimate a distance between the transmitter and the receiver, and using the AoA of the direct path signal to estimate an angle of the transmitter so as to localise the transmitter.
  • the predefined threshold may correspond to a dynamic range of the apparatus.
  • the convergence criteria may comprise determining when a difference of each parameter for each signal path between two successive iterations is smaller than a predefined convergence threshold.
  • the received signal may be constituted by a Wi-Fi signal.
  • a method for localisation and/or tracking comprising: receiving a received signal, the received signal comprising a superposition of multiple signals derived from a wireless signal transmitted from a transmitter along a plurality of different paths; separating the received signal into the multiple signals by: a) estimating initial path parameters by:
  • step (i) and (ii) repeating step (i) and (ii) using the re-estimated background noise as the initial background noise until the parameters for each signal path satisfy a convergence criteria;
  • MLE maximum likelihood estimation
  • EM expectation-maximising
  • EM may be applied if the observed data Y is incomplete and there is some unobservable data set X whose probability distribution function depends on ⁇ with the property that maximizing p(X
  • X is a variable that ⁇ X ⁇ Y is a Markov chain.
  • an EM algorithm can be used to solve it.
  • EM is not a specific algorithm towards solving any specific problem.
  • EM is a general type of algorithm for solving MLE problems. Reference 7 introduces this type of algorithm and demonstrates its convergence. EM provides a high-level guideline for solving such types of problems but does not give any detailed instructions to solve any specific problems like, e.g. estimating the parameters of indoor multipath signals as per embodiments of the present invention.
  • Reference 53 proposes a parameter estimation algorithm to estimate parameters using a sub-space based algorithm known as Joint Angle and Delay Estimation (JADE).
  • JADE Joint Angle and Delay Estimation
  • a sub-space based algorithm is different in design principle from a maximum likelihood estimation based algorithm.
  • Sub-space based algorithms project the received signal into signal sub-space and noise sub-space by performing Eigen analysis on a covariance matrix of the received signal. Parameters that can make the projected signal and noise sub-space orthogonal to each other provide the estimated results. Accordingly, a sub-spaced based algorithm is not an iterative algorithm.
  • the core operations in steps (a) and (b) of the present method are all iterative operations.
  • reference 53 is designed to merely estimate angle (i.e. AoA) and delay (i.e. ToF).
  • the present method is capable of jointly estimating multi-dimensional parameters, i.e., AoA, AoD, ToF, Doppler shift and signal strength for every multipath signal.
  • the present method is also flexible and can be easily tuned to estimate different combinations of parameters (e.g., 1 -Dimensional ToF, 2-Dimensional AoA + ToF, 3-Dimensional AoA + AoD + ToF, 4-Dimensional AoA + AoD + ToF + Doppler, etc.), to cope with different system conditions.
  • Interference cancellation is a basic signal processing technique to remove interference. If there are multiple interference sources, their impact can be removed successively, which is also referred to as successive interference cancellation (SIC). SIC is widely used to decode multiple colliding packets, as reference 1 1 does. It decodes the strongest packet, re-constructs the signal of the first packets using the decoded bits, cancels the re-constructed signal from the received signal and decodes the second strongest packet until no packet can be decoded.
  • SIC successive interference cancellation
  • Embodiments of the present invention apply SIC to separate the different signal paths, which is different from traditional usage of SIC for packet decoding where different packets are treated as interference to each other.
  • the present method treats the signal of the same packet but from different paths as interference to each other, which has not been studied previously and is different from traditional usage of SIC.
  • SIC may be performed in step (a) of the present method.
  • it involves reconstruction of the original signal from the strongest wireless signal path by using the estimated parameters, which is not part of traditional usage of SIC where existing methods use decoded packet data to reconstruct the signal (superimposed signal from multi-path) as a whole.
  • the reconstruction does not rely on the channel parameters from individual signal paths.
  • Step (a) only provides the initial estimates for the present iterative parameter estimation algorithm.
  • the parameter obtained from step (a) is not accurate enough since the estimation error in the strongest path will cause energy leakage into the residual signal, which impacts the estimation of the next signal as explained in more detail below.
  • step (b) is performed to use the output from step (a) and iteratively refine the path parameter estimation for more accurate results.
  • Step (b) is not part of SIC and is not involved in any sub-space based algorithms. It is therefore believed that the combination of steps proposed for the present method provides a new and advantageous approach to estimating parameters of indoor wireless signals.
  • Applications of embodiments of the invention include but are not limited to indoor Wi-Fi based localisation systems (i.e. for locating active mobile devices like mobile phones or passive localisation systems for locating a reflector (e.g. human body) not carrying an active wireless device.
  • Embodiments of the invention may be employed in a human-computer interface to support applications like gesture recognition, human activity recognition and so on.
  • Some embodiments of the invention may be employed in security systems to detect the intruders.
  • Figure 1 illustrates joint parameter estimation in a) ToF and AoA dimensions and b) ToF, AoA and Doppler shift dimensions in accordance with embodiments of the invention
  • Figure 2A shows a schematic diagram of an apparatus for localisation and/or tracking in accordance with embodiments of the invention
  • Figure 2B illustrates a method (referred to as mD-Track) for localisation and/or tracking in accordance with embodiments of the present invention
  • Figure 3 illustrates a four-dimensional estimator tool in accordance with embodiments of the invention
  • Figure 4 shows a signal structure for a Wi-Fi (multiple input and multiple output) Ml MO preamble, as may be employed in embodiments of the invention
  • Figure 5 shows an example output (z-function) of the estimator tool with input: (a) superimposed signal of a stronger signal with AoA and ToF [60.7°, 20.8ns] and a weaker signal with [73.4°, 28.1 ns]; (b) the residual signal after cancelling the stronger signal; (c) reconstructed stronger signal and (d) reconstructed weaker signal after three iterations;
  • Figure 7 shows (a) a histogram of absolute ToF of a shortest path and that of a relative ToF between two paths; (b) a cumulative distribution function (CDF) of a ToF error;
  • Figure 8 shows three transmitter-chains and three receiver-chains connected together with a power splitter/combiner in (a) one transmitter-chain is connected with two receiver-chains and the remaining two transmitter-chains are connected with the other receiver-chains in (b) and (c);
  • Figure 9 shows (a) a general experimental set up in an indoor meeting room; (b) a customer off the shelf (COTS) Wi-Fi access point (AP) with three antennas; and (c) a WARP (wireless open access research platform) with eight antennas;
  • COTS customer off the shelf
  • AP Wi-Fi access point
  • WARP wireless open access research platform
  • Figure 10 plots experimental results indicating the probability that two paths can be resolved for (a) MUSIC 1-Dimensional (1 D) estimate of ToF; (b) MUSIC 1-Dimensional (1 D) estimate of AoA; (c) SpotFi 2-Dimensional joint estimate of AoA+ToF; (d) mD- Track 2-Dimensional (2D) joint estimate of AoA+ToF; (e) mD-Track 3-Dimensional (3D) joint estimate of AoA+ToF+Doppler with Doppler unchanged; (f) mD-Track 4- Dimensional (4D) joint estimate of AoA+ToF+Doppler+AoD, with AoD and Doppler unchanged; (g) mD-Track 4-Dimensional (4D) joint estimate of AoA+ToF+Doppler+AoD, with ToF and Doppler unchanged; and (h) mD-Track 4- Dimensional (4D) joint estimate of AoA
  • Figure 11 shows graphs of a cumulative distribution function (CDF) of (a) AoA estimation error of a direct path signal; (b) AoA estimation error of reflection path signals; (c) ToF estimation error; and (d) AoD estimation error, for mD-Track, SpotFi, and MUSIC methods;
  • CDF cumulative distribution function
  • Figure 12 shows a graph of localisation error of mD-Track and SpotFi when using the WARP platform
  • Figure 13 shows a graph of localisation error of mD-Track when using a commercial Wi-Fi device
  • Figure 14 shows (a) an indoor floorplan map of a meeting room; (b) a localisation error of mD-Track 2D; and (c) a localisation error of mD-Track 3D, for a person located in the centre of the room;
  • Figure 15 shows a bar chart of localisation error of mDTrack 3D when tracking multiple targets separated by different distances
  • Figure 16 shows graphs of Doppler shifts introduced by different motions when (a) the transmitter moves; (b) a human walks; (c) a human stands still and waves his hand; (d) a human stands still and moves his finger;
  • Figure 17 shows a plot of estimated ToFs, AoAs and Doppler shifts resolved using mDTrack for three static paths in the environment and four mobile paths caused by four hands of two persons; and
  • Figure 18 shows graphs of (a) Number of iterations mD-Track takes to converge with different number of multipaths; (b) End-to-end execution time of mD-Track; and (c) End-to-end execution time of SpotFi and mD-Track with varied granularities (i.e. step sizes).
  • Embodiments of the present invention relate to a method and apparatus for localisation and/or tracking by estimating parameters of indoor wireless signals, which may be used in an active or passive tracking system and which is capable of jointly utilising information from many dimensions (i.e. signal parameters) to break the resolution limit of each individual dimension. Accordingly, embodiments of the invention relate to multi-dimensional tracking and an exemplary system is described in detail below and is referred to as mD-Track.
  • Embodiments of the invention apply a novel path separation algorithm to resolve multipath signals at a much higher resolution (e.g. 25 times higher) than in the prior art, isolating signals reflected off targets of interest.
  • a much higher resolution e.g. 25 times higher
  • Embodiments of the invention also utilise novel methods to refine the signal parameters to achieve real-time operation.
  • mD-Track has been implemented on both WARP and off-the-shelf commodity Wi-Fi hardware and its performance has been evaluated in different indoor environments.
  • the experimental results demonstrate simultaneous tracking of up to four human targets at a median location error of 47cm and a 3.4 times accuracy improvement and a 30 times computational complexity reduction for passive localisation of a single target compared to a state-of-the-art SpotFi system.
  • Passive motion tracking without any device carried by or attached to a target has been an exciting area of recent interest, with important applications including security surveillance [reference 25], elderly care [reference 2], and retail business [reference 30].
  • Wi-Fi based systems stand out as particularly promising due to the pervasive availability of Wi-Fi access points (APs).
  • APs Wi-Fi access points
  • Wi-Fi based systems rely on signal reflections off targets to extract essential motion and location information.
  • passive tracking is more challenging than localisation of active wireless transmitters, because reflected signals of interest are typically orders of magnitude weaker than for a direct-path signal, and are typically superimposed with a strong direct-path signal as well as other signals reflected from walls, furniture, and other nearby clutter. How to accurately resolve and identify the weak signal reflected off a target of interest becomes a major challenge for these systems.
  • Figure 1 illustrates the intuition behind the present idea.
  • three signals S1 , S2 and S3 are deemed to arrive at a wireless receiver simultaneously.
  • signals S1 and S2 are close in time (i.e. have a similar ToF) and therefore cannot be resolved easily by employing ToF.
  • S1 and S2 can be relatively easily resolved with AoA, since their AoA difference is large.
  • signals S1 and S3 cannot be easily separated with AoA alone, but are resolvable via ToF.
  • This concept extends to higher dimensions as shown in Figure 1 (b), where the receiver jointly estimates ToF, AoA, and Doppler shift.
  • signals S4 and S5 are close to each other in both AoA and ToF, but since S4 is a signal from a moving source, it exhibits a non-zero Doppler shift that separates it from S5, a reflected signal from a static object with zero Doppler shift. Accordingly, embodiments of the invention propose to improve signal resolvability by jointly exploiting information from more signal dimensions, without changing a resolution limit of any individual dimension.
  • the exemplary mD-Track is a passive Wi-Fi tracking system that fuses information from multiple signal dimensions, significantly improving resolvability without requiring a wider frequency bandwidth or a larger number of antennas.
  • To utilise multidimensional information one approach is to explore each signal dimension separately by performing multiple one-dimensional parameter estimations and then combining the results.
  • one-dimensional estimates often cannot achieve the required resolution granularity (as illustrated in Figure 1 (a)) and pairing path parameters from different dimensions is a challenging problem in its own right.
  • mD-Track jointly estimates multi-dimensional parameters at substantially the same time (e.g. jointly estimating the parameters for each signal path in a sequential manner) so the signal parameters of a same signal are always paired inherently and the resolvability is greatly improved since two signals become resolvable if they can be resolved in any one of the dimensions.
  • mDTrack With signals whose parameters are jointly estimated, mDTrack can passively localise and track the motions of multiple targets simultaneously with only a single transmitter-receiver pair. mD-Track can therefore resolve instances of multiple incoming signals that other super-resolution signal processing methods fail to separate. mD-Track has a high sensitivity to accurately detect even subtle finger movements, enabling it to carry out fine-grained tracking. In order to do so, however, and to make the above approach practical, mD-Track, overcomes the following challenges: a) Multipath signal separation and accurate parameter estimation
  • mD-Track may employ successive interference cancellation (SIC) [references 11 , 29], a technique originally proposed for data communication, to remove inter-path interference.
  • SIC successive interference cancellation
  • mD-Track employs iterative path parameter refinement during which the signals are iteratively re-estimated and more accurately reconstructed with refined parameters in multiple rounds of estimation, which can be proven to be an expectation maximization estimation algorithm that recalls the structure of a Turbo Decoder [reference 32]. b) Bounding computation for real time operation
  • the experimental evaluation begins with a sensitivity analysis that analyses each signal parameter, measuring the relative ability of different parameters to resolve signals. Head-to-head comparisons in parameter estimation and passive localisation demonstrate 3.5 times accuracy improvements over the state-of- the-art SpotFi system [reference 17]. Further experiments measure the effect of adding another signal dimension (Doppler shift) to mD-Track, showing approximately a 3 times accuracy improvement, in contrast with a marginal 20% improvement from doubling the frequency bandwidth. Also different from existing approaches, the number of signal paths mD-Track can resolve is no longer limited by the number of antennas. With just three antennas available on the COTS Wi-Fi AP, mD-Track is demonstrated to resolve more than 10 signal paths and to estimate the parameters of each signal path accurately. Finally, with the capability of separating reflected signals from multiple close-by targets, for the first time, mD-Track is able to achieve high accuracy device-free (i.e. passive) multi-target localisation with just a single transceiver pair.
  • Localisation or motion tracking of a target relies on separating superimposed signals and accurately estimating each signal's parameters. For example, active localisation relies on separating the direct path signal from all other signal paths, whereas passive localisation estimates a target's position by separating and estimating the parameters of the reflection paths from the target instead.
  • a wireless signal is characterised by multi-dimensional parameters, each parameter providing a piece of location or motion information about the target.
  • an apparatus 10 for localisation and/or tracking as illustrated in Figure 2A.
  • the apparatus 10 comprises a receiver 12 for receiving a received signal, the received signal comprising a superposition of multiple signals derived from a wireless signal transmitted from a transmitter 14 along a plurality of different paths (i.e. Direct Path, Path 1 , Path 2 etc.); a computer processor 16; and a data storage device 18, the data storage device 18 comprising instructions operative by the computer processor 16 to separate the received signal into the multiple signals by: a) estimating initial path parameters by:
  • step (i) and (ii) repeating step (i) and (ii) using the re-estimated background noise as the initial background noise until the parameters for each signal path satisfy a convergence criteria:
  • Figure 2A also shows the parameters that can be retrieved from a wireless signal transmitted by the transmitter 14 and reflected from a mobile human target 20 and a static object 22.
  • the parameters for each signal path comprise the time of flight ( ⁇ ), the angle of arrival ( ⁇ ), the angle of departure ( ⁇ ) and the Doppler shift ( ⁇ ).
  • ToF time of flight
  • the angle of arrival (AoA) ( ⁇ ) indicates the direction of the reflected signal as it arrives at the receiver 12, and the angle of departure (AoD) ( ⁇ ) indicates the direction of the signal sent from the transmitter 14.
  • the number of antennas at the receiver 12 and transmitter 14 determines the resolution of the AoA and AoD estimates, respectively.
  • Movement of the transmitter 14, receiver 12, or reflectors 20, 22 all introduce frequency shifts to a carrier frequency of the signal which is referred as Doppler shift ( ⁇ ).
  • Doppler shift provides motion information and can help to estimate a relative velocity of the target 20.
  • Doppler shift resolution is related to an observation interval such that the longer the interval, the finer the resolution.
  • a further parameter of each signal path is its complex attenuation (a) which relates to signal strength. For example, when a signal propagates in air from the transmitter 14 to the receiver 12, its signal strength will be attenuated by a.
  • Figure 2B illustrates a method 30 for localisation and/or tracking in accordance with embodiments of the invention.
  • the method 30 comprising: a step 32 of receiving a received signal, the received signal comprising a superposition of multiple signals derived from a wireless signal transmitted from a transmitter along a plurality of different paths; followed by separating the received signal into the multiple signals by: a) a step 34 of estimating initial path parameters by:
  • the transmitter 14 may comprise an array of N antennas and the receiver 12 may comprise an array of M antennas. Both arrays may be linear with a uniform spacing d between adjacent antennas.
  • An N-antenna transmitter array steering vector g( ⁇ p) characterises a phase relationship of the signal coming out of the N transmitting antennas while an M-antenna receiver array steering vector c(0) characterises a phase relationship of the signal arriving at the M receiving antennas.
  • t indicates time
  • T signals a transpose of a matrix
  • j indicates a complex number.
  • W(t) [w t (t), w 2 (t), ... , w M (t)] T is M-dimensional complex Gaussian noise capturing the background noise. It should be noted that s(t) and s(t; ⁇ ) are used interchangeably throughout this specification.
  • the estimator tool 40 is illustrated in Figure 3 and is designed to be factored into a number of different modules, each corresponding to one of the above parameters such that multiple modules are employed together to jointly estimate all of the parameters.
  • the estimator tool 40 is configured as a 4-dimensional sequential estimator for estimating the parameters AoA, AoD, Doppler shift and ToF of a wireless signal as it propagates along a single path from a transmitter array 42 to a receiver array 44.
  • the signal is illustrated as comprising two consecutive time slots t and the estimator tool 40 is shown to comprise an AoA estimator module 46, an AoD estimator module 48, a Doppler shift estimator module 50 and a ToF estimator module 52 as will be explained in more detail below.
  • the AoA estimator module 46 is a beamformer which selectively receives signals arriving at a specific AoA 0 in each time slot. It is implemented by multiplying the signal received at each antenna in the receiver array 44 with different complex weights and then combining them. For example, to configure the array to only receive signals from angle 0, weights are applied to remove via phase rotation function c(0) the phase differences at each antenna due to differences in the propagation lengths caused by the incidence angle 0, as depicted in Figure 3.
  • the signals from all the antennas align in time and are added constructively, so the strength of signal at AoA 0 is maximised. Scanning every possible incidence angle, the AoA that produces the maximum beamformer output is the AoA estimate.
  • the AoA estimator can be extended to 3-dimensions including both azimuth and elevation by employing a two- dimensional receiver antenna array.
  • a signal structure 60 of a Wi-Fi MIMO (multiple input, multiple output) preamble for spatial multiplexing is exploited to estimate the AoD using the AoD estimator module 48.
  • two Long Training Symbols (LTSs) are transmitted by two antennas (TX ant-1 and TX ant-2) in the transmitter array 42 in two time slots sequentially before respective payloads 1 and 2.
  • a spatial mapping matrix is multiplied to the LTS in Figure 4 when transmitted but is inverted at the receiver 12 and hence is ignored here.
  • the signal transmitted by the second antenna travels an additional distance to reach the receiver array 44 and hence introduces an extra phase shift.
  • AoA estimates of a single time slot transmission do not capture such an extra phase shift.
  • mD-Track analyses the transmissions from each of the antennas in the transmitter array 42 in successive time slots, as illustrated in Figure 3 for a two-antenna example.
  • the Doppler estimator module 50 estimates a shift in carrier frequency introduced to the signal by motion of the transmitter 14, receiver 12 or reflector 20. From Equation (1) it is noted that the Doppler shift causes a 2 yt phase shift to the received signal. This can be therefore be cancelled by multiplying the signal with e -j2 yt Different Doppler shifts are applied, calculating a correlation with the transmitted signal U(t), which is maximised when the applied Doppler shift matches the actual Doppler shift.
  • Embodiments of the present invention combine the above individual modules into a four-dimensional estimator z(0, ⁇ , ⁇ , ⁇ ) as shown in Figure 3 for a two-antenna transmitter 14 and receiver 12.
  • the first step is to estimate AoA for each of the two timeslots using the AoA estimator modules 46, before feeding the outputs of each AoA beamformer into the AoD estimator module 48.
  • the output of the AoD estimator module 48 is then passed to the Doppler estimator module 50 to estimate and remove any Doppler shift, before the signal is correlated with a delayed transmitter signal U(t--c) to estimate the ToF via the ToF estimator module 52.
  • the processing of the signal through the estimator tool 40 is sequential. However, the order of processing is carefully chosen: Doppler shift affects all antennas equally, and so the AoA and AoD estimator modules 46, 48, which rely on measuring phase differences across different antennas, are unimpaired by any arbitrary Doppler shifts. Furthermore, a parameter search, as described below, allows joint estimation of all of the parameters for each signal path. For the general four-dimensional estimator tool 40 shown in Figure 3, the output is as per Equation (6): where ⁇ (0, ⁇ , ⁇ , ⁇ ), referred to as the z-function, is calculated for all possible [ ⁇ , ⁇ , ⁇ , ⁇ ] combinations. The estimation of the parameters [ ⁇ , ⁇ , ⁇ , ⁇ ] is then obtained by Equation (7):
  • Equation (1) Combining Equation (1) and Equation (7) with Equation (2) then yields a, as per Equation (8):
  • the design is modular it has high flexibility and can be adapted to a three- dimensional estimator for a single-antenna transmitter 14 case, by removing the AoD estimator module 48 and the second AoA estimator module 46 for the second time slot as this is no longer required. It may also be adapted to a two-dimensional estimator or a one-dimensional estimator (e.g. a ToF estimator for a single antenna transceiver case).
  • V , 3 ⁇ 4 , ...uJ (10) for all L paths in Y(t). Note that L is also unknown and thus must be estimated.
  • Resolvability is the premise of accurate path estimation. Non-resolvable signals are merged and the estimated parameters of the combined signal lead to a non- existing path which deviates from the true paths.
  • the multi-dimensional estimator tool 40 described above improves the resolvability by higher dimensional parameters [0 h ⁇ , ⁇ , ⁇ , ⁇ 1 ] ⁇ of each path, the energies of nearby signal paths may affect the estimation accuracy (for example, a weaker signal may be overwhelmed by a nearby strong signal and thus not be detected).
  • a controlled experiment is described below to demonstrate such an effect, where two signals with different AoAs (60.7° and 73.4°) and ToFs (20.6ns and 28.1 ns) are created and superimposed.
  • the signal with ToF 20.6ns is 10dB stronger than the other signal.
  • the superimposed signal is input into a 2-dimensional estimator tool as per Figure 3 but configured for estimating only AoA and ToF and an output z-function is as depicted in Figure 5(a). It was expected that two peaks would appear in Figure 5(a). However, the weaker signal, in practice, is dominated by the stronger signal and cannot be detected from the z-function. Similar results have been obtained in previous studies when a 1-dimensional MUSIC or 2- dimensional SpotFi estimator is employed. A main reason for this is that those approaches identify the signal paths and estimate the parameters in a single round. No iterative refining process is included to accurately determine energy shares between different signal paths, as per embodiments of the present invention.
  • mD-Track employs an iterative path parameter refinement process during which the path signals are iteratively re-estimated and more accurately reconstructed from the refined parameters in multiple rounds of estimation.
  • a successive interference cancellation (SIC) technique is used as borrowed from communication theory [references 1 1 , 29] to remove the inter-path interference, and derive the initial path estimates as a starting point for the refinement process.
  • Figure 5(b) shows the residual signal after cancelling the stronger signal
  • Figure 5(c) shows a reconstructed stronger signal as explained below
  • Figure 5(d) shows a reconstructed weaker signal as explained below, after three iterations.
  • the peaks of the z-function appear at [63.5°; 22.5ns] in Figure 5(a), [85.9°; 30.5ns] in Figure 5(b), [61.2°; 21.3ns] in Figure 5(c), and [76.7°; 27.0ns] in Figure 5(d).
  • the stronger signal that was reconstructed and cancelled is not identical to the true signal due to estimation errors. Energy leakage to the residual signal has a large effect on the estimation of the weaker signal, which result in a large estimation error, as shown in Figure 5(b).
  • the signal x (t) is now refined, containing less interference compared to the raw signal X[(t) in which weaker signals are also considered as noise.
  • the path parameters of xi ' (t), u j ' are re-estimated using the multi-dimensional estimator tool 40 described above.
  • the background noise is re-calculated and updated.
  • the noise of the (l s + l) th path can be updated as per Equation (14):
  • V ' is fed back into Equation 13 to start another round of estimation.
  • the above steps of multipath signal re-estimation and reconstruction are iterated and stop when the difference of each path parameter between two consecutive iterations is smaller than a predefined threshold.
  • the threshold can be flexibly tuned to meet different levels of accuracy requirements.
  • the threshold is set as the step size of the parameter search above, yielding accurate results for passive localisation and motion tracking as will be explained in more detail below.
  • Figure 6 illustrates the workflow of the iterative process 70 based on an example of superimposed signals from two paths.
  • the received signal Y(t) which comprises all of the multipath signals plus noise is considered to be the first signal x ⁇ t) in step 72 and is fed into the multi-dimensional estimator tool 40 as described above in relation to Figure 3 in step 74 so as to calculate an initial estimate of the parameters for the first signal path.
  • the first signal path is then reconstructed in step 76 using the parameters and the reconstructed first signal s ⁇ t; ! ⁇ ) is cancelled from the received signal Y(t) in a step 78 to leave a residual signal x 2 (t) in step 80.
  • the residual signal x 2 (t) is then fed into the multi-dimensional estimator tool 40 as described above in relation to Figure 3 in step 82 so as to calculate an initial estimate of the parameters ⁇ 2 for the second signal path.
  • the second signal path is then reconstructed in step 84 using the parameters ⁇ 2 and then both the reconstructed signals s ⁇ t Ui) and s 2 (t; u 2 ) are cancelled from the received signal Y(t) in a step 86 to update the background noise W(t).
  • the updated noise W(t) is then added to the reconstructed first signal s ⁇ t; in step 88 and fed into the multi-dimensional estimator tool 40 as described above in relation to Figure 3 in step 90 so as to calculate a refined estimate of the parameters v t for the first signal path.
  • the first signal path is then reconstructed in step 92 using the refined parameters v t and the reconstructed refined first signal s ⁇ t; 3 ⁇ 4 ) along with the reconstructed second signal s 2 (t; u 2 ) is cancelled from the received signal Y(t) in a step 94 to yield an updated background noise W(t).
  • the updated noise is then added to the reconstructed second signal s 2 (t; u 2 ) in step 96 and fed into the multidimensional estimator tool 40 as described above in relation to Figure 3 in step 98 so as to calculate a refined estimate of the parameters ⁇ 2 ' for the second signal path.
  • the second signal path is then reconstructed in step 100 using the parameters ⁇ 2 ' and then both the reconstructed refined signals si ⁇ Ui ' ) and s 2 ' (t; u 2 ) are cancelled from the received signal Y(t) in a step 102 to further update the background noise W(t) and the process continues iteratively until a convergence condition is met.
  • SIC gives the coarse estimates v t and ⁇ 2 for the two paths. From there, the iterative process 70 reconstructs the path signals x 1 (t) and x 2 (t) , re- estimates the path parameters and ⁇ 2 , and updates the residual signal (i.e. noise) W(t) in each iteration.
  • Figures 5(c) and 5(d) present the z-function of the multi-dimensional estimator tool 40 after three rounds of iteration. It is clear that the peaks in Figures 5(c) and 5(d) are much closer to the true path parameters compared with Figures 5(a) and 5(b). Thus, demonstrating that the iterative process helps to better refine the estimated parameters.
  • the multi-dimensional estimator tool 40 contributes a significant portion of the computation of mDTrack.
  • Each multidimensional estimator tool 40 solves a maximum likelihood problem by an exhaustive search.
  • n. P0 ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇
  • the combination ⁇ increases exponentially with the number of dimensions and causes significant overhead. From Figure 6, it is observed that the estimator tool 40 is applied N it erL times in a system with L paths and N it er rounds of iteration.
  • each multi-dimensional estimator tool 40 is actually the signal from one path plus the noise, as shown in Figure 6.
  • the power level of the signal is tens of dB higher than noise so that a dominant peak in the output z-function of the estimator is observed as per Figures 5(c) and 5(d).
  • a coordinate descent method may be used to approach the maximum.
  • the four-dimensional search of Equation 7 is replaced with four one-dimensional searches. In practice this means that three of the four parameters (e.g. ⁇ , ⁇ , and ⁇ ) are fixed while a search for the value of the fourth parameter (e.g., 0) is conducted and a maximal output can be generated as per Equation (16):
  • Embodiments of the present invention have been implemented using an HP desktop computer as a backend server with an Intel i7 central processing unit (i.e. processor) and 8GB random access memory (i.e. data storage device). Examples using both WARP v3 boards [reference 26] and commodity Wi-Fi routers (TP-Link, TL- WDR4300 and Compex WPJ558) were employed for collecting the channel measurements and relaying them to the server for processing in real-time.
  • WARP v3 boards [reference 26]
  • commodity Wi-Fi routers TP-Link, TL- WDR4300 and Compex WPJ558
  • the WARP v3 board collects time domain information samples and sends them to a computer connected with the WARP via an Ethernet cable.
  • the computer detects the beginning of a packet, extracts the LTS (long training symbol) in the preamble and sends time domain LTS to the server through the internet.
  • All of the commercial access points are equipped with Atheros Wi- Fi NIC (network interface controller) or SoC (System on Chip) (AR9340, AR9580, and QCA9558), and run a customized OpenWRT system.
  • the Wi-Fi driver is modified in kernel space of OpenWRT to enable CSI (channel state information) collection.
  • a user-space application was also built to retrieve the frequency domain LTS by multiplying the transmitted LTS (defined in 802.1 1 standard) with received CSI (empty subcarriers of CSI and LTS are padded with zeros).
  • the frequency domain LTS is then transformed into time domain information via IFFT (inverse Fast Fourier Transform), which is sent to the server through the internet.
  • IFFT inverse Fast Fourier Transform
  • Figure 7(a) plots a histogram of the estimated absolute ToF values of the signal from the shorter cable, which exhibits a large variation across different transmissions.
  • Figure 7(b) presents a cumulative distribution function (CDF) of a ToF estimation error. This shows that even when the absolute ToF measurements are inaccurate (i.e. with a median error of 13ns), the relative ToF estimates can be very accurate (i.e. with a median error of 0.48ns).
  • CDF cumulative distribution function
  • the ToF measurement of the direct path can be used as a reference basis to calibrate the ToF estimations of other reflection paths.
  • the ground-truth ToF, AoA and AoD of the direct path need to be calculated based on the locations of the transmitter 14 and receiver 12. From all signal path outputs from mD-Track, the path with 1) the shortest ToF, 2) the highest amplitude, or 3) the smallest AoA and AoD difference, can be identified as the direct path measurement and the ⁇ between the measured and ground-truth ToF can therefore be determined. The derived ⁇ can then be used to correct the ToF measurements for all reflection paths. In this way, time synchronisation is not required and it is still possible to obtain accurate ToF information.
  • the WiFi transmitter 14 and receiver 12 may be located at high positions, e.g., on a ceiling, to ensure a direct path between them exists.
  • the targets being monitored may be at a different height and the path parameters (ToF, AoA, AoD) need to be projected to the 2D horizontal plane at the ceiling. This can be done with knowledge of the height of the ceiling.
  • ToF, AoA, AoD path parameters
  • the system can be aware of that because in such conditions no detected path will match the ground-truth direct path, and, in this case, it is possible to omit ToF as a parameter and employ a lower dimensional estimator tool 40.
  • LO Local oscillator
  • the antennas of each Wi-Fi frontend are phase-locked to different oscillator phases.
  • the LO offset between antennas must be eliminated to provide accurate angle estimation (AoA and AoD).
  • the constant LO offset can be measured by connecting the TX-chain of the transmitter antennas with the RX-chain of the receiver antennas (e.g. of the WARP or COTS Wi-Fi devices) via coaxial cable.
  • the three TX-chains and three RX-chains are connected together via a single power splitter/combiner.
  • spatial multiplexing fails and no packets can be correctly received with such a setup as the channel matrix becomes singular.
  • FIGs 8(b) and 8(c) the setups in Figures 8(b) and 8(c) are used where one TX-chain is connected with two RX-chains and the remaining two TX- chains are connected with the remaining RX-chain.
  • the LO offset is measured between: 1) TX-chain 2 and TX-chain 3, as denoted as a 2 ; and 2) RX-chain 1 and RX-chain 2, as denoted as Similarly, LO offset is measured between: 1) TX- chain 1 and TX-chain 2, as denoted as t ; and 2) RX-chain 2 and RX-chain 3, as denoted as ⁇ 2 in Figure 8(c).
  • CFO between the transmitter 14 and the receiver 12 may confuse Doppler shift estimation since CFO and Doppler frequency shift have the same effect on a carrier frequency.
  • Doppler shift is introduced by a moving target, so the Doppler effect only occurs on the signal reflected off the target.
  • CFO is introduced by the transceivers, so it affects all signals. Therefore, the frequency change introduced by CFO has the same value for all signals. Based on this observation, the frequency shifts caused by CFO and Doppler shift can be distinguished. The frequency shift caused by CFO is then subtracted to obtain an accurate Doppler shift estimate.
  • mD-Track has a very high resolution for Doppler estimation, i.e., 1 Hz, so subtle frequency shifts caused by moving targets (e.g. in the order of tens of Hz) are still accurately estimated in the presence of a large CFO value.
  • Figure 9(a) shows an experimental setup in an indoor meeting room. This comprises a transmitter access point (TX AP), a transmitter array (TX array), a receiver array (RX array) and a receiver access point (RX AP).
  • TX AP transmitter access point
  • TX array transmitter array
  • RX array receiver array
  • RX AP receiver access point
  • Figure 9(b) shows a commercial AP having a three-antenna array constituting the RX AP and the RX array, respectively, in some experiments.
  • Figure 9(c) shows a WARP AP having an eight- antenna array constituting the RX AP and the RX array, respectively, in some experiments. In both cases, channel measurements were collected and sent to a server via the internet, where the path parameters were estimated for target tracking and localisation in real time.
  • mD-Track Three versions of mD-Track were tested, as shown in Table 1.
  • the input signal to each version of mD-Track is different.
  • An LTS symbol consists of 64/128 information samples if a 20/40 MHz channel is used.
  • the LTS symbols received by M receiving antennas are required to estimate AoA plus ToF.
  • the input is a matrix of information samples with size M x 64 (20 MHz) to mD-Track 2D.
  • the matrix size increases to M x 64 x T and M x N x 64 x T for mD-Track 3D and 4D respectively, where N is the number of transmitting antennas and T is the number of packets collected for a Doppler shift estimation.
  • mD-Track requires estimation over multiple packets when and only when Doppler shift is jointly estimated.
  • SpotFi estimates ToF plus AoA and takes CSI as input which is calculated using transmitted and received LTS at WARP.
  • the size of the input CSI matrix is M x 56 and M x 114 when using 20 and 40 MHz, respectively.
  • 1000 packets were transmitted and the two signals were resolved with mD-Track 1000 times using the received packets.
  • Figure 10 plots the output parameters for each variant tested.
  • the resolvability of the mD-Track and SpotFi (using JADE) versions listed in Table 1 are compared also with a 1 D MUSIC algorithm, separately for ToF and AoA. More specifically, in Figure 10 white indicates that the two signals are "non-resolvable" and black indicates that the two signals are "fully resolvable”.
  • Figure 10(a) shows the resolvability of a MUSIC 1 D estimator for ToF
  • Figure 10(b) shows the resolvability of a MUSIC 1 D estimator for AoA
  • Figure 10(c) shows the resolvability of a SpotFi 2D estimator for jointly estimating AoA+ToF
  • Figure 10(d) shows the resolvability of mD-Track 2D to jointly estimate AoA+ToF
  • Figure 10(e) shows the resolvability of mD-Track 3D to jointly estimate AoA+ToF+Doppler, with Doppler unchanged
  • Figure 10(f) shows the resolvability of mD-Track 4D to jointly estimate AoA+ToF+Doppler+AoD, with AoD and Doppler unchanged
  • Figure 10(g) shows the resolvability of mD-Track 4D to jointly estimate AoA+ToF+Doppler+AoD, with ToF and Doppler unchanged
  • mD-Track 2D still outperforms SpotFi 2D, as illustrated in a comparison of Figures 10(c) and (d).
  • the improvement stems from the iterative interference cancellation and path refinement process in mDTrack as described above.
  • Both MUSIC and SpotFi estimate parameters of all resolvable paths simultaneously and ignore the mutual interference between paths.
  • mD-Track only estimates one path (the strongest one) at a time and subtracts the estimated signal from the input before estimating the next path. As long as parameters of the stronger path are accurate, the nearby weaker signal is preserved in the residual signal and can be estimated accurately, as shown in Figure 5(c) and (d). Furthermore, increasing the parameter dimensions helps to increase the resolvability significantly.
  • mD-Track estimates signal path parameters with high accuracies.
  • Channel measurements were collected using a Compex WPJ558 router, which has three antennas.
  • the ground truth of the location related parameters, including AoA, AoD, and ToF, were calculated based on the actual positions of the transceivers.
  • a comparison of AoA and ToF estimation is made with MUSIC and SpotFi.
  • the comparison favors MUSIC and SpotFi by selecting the path with the AoA estimate closest to the ground truth as their estimate.
  • the ground truth of Doppler is determined by motion and is hard to measure. An evaluation of mD-Track's Doppler estimation is discussed later.
  • Figure 1 1 shows a cumulative distribution function (CDF) of AoA estimation error of the direct path signal for mD-Track 4D, mD-Track 2D, SpotFi and MUSIC.
  • Figure 11 shows a cumulative distribution function (CDF) of AoA estimation error of reflection path signals for mD-Track 4D, mD-Track 2D, SpotFi and MUSIC.
  • Figure 1 1 shows a cumulative distribution function (CDF) of ToF estimation error for mD-Track 4D, mD-Track 2D, SpotFi and MUSIC.
  • Figure 1 1 (d) shows a cumulative distribution function (CDF) of AoD estimation error, for mD-Track 4D LOS (i.e. an active set-up) and mD-Track 4D passive (i.e. using a reflected signal).
  • CDF cumulative distribution function
  • FIGS 11 (a) and (b) AoA estimation results are presented in Figures 11 (a) and (b). From these it is observed that mD-Track provides much more accurate AoA estimates: with only three antennas, mD-Track's median AoA error for the direct path (Figure 11 (a)) is as small as 4.4° and 6.2° for 4D and 2D versions respectively, compared to 13.4° and 17.1 ° for SpotFi and MUSIC, respectively. The AoA estimates of the reflection paths ( Figure 1 1 (b)) are less accurate because reflection signals are weaker in most cases. Despite that, mD-Track still achieves a median AoA error of 5.6° and 7.3° for 4D and 2D versions respectively, compared with 16.9° and 24.2° for SpotFi and MUSIC respectively.
  • mD-Track calculates the ToF differences between the reflection paths and direct path. From Figure 11 (c), it is clear that mD-Track 4D and 2D are able to estimate the ToF difference between the reflection path and direct path with a median error of 1.23 ns and 1.85 ns, while SpotFi and MUSIC achieve a median error of 3.8 ns and 6.7 ns.
  • Figure 1 1 (d) shows the AoD estimation results for a direct (LoS) path and a (passive) reflection path using mD-Track 4D.
  • the median errors are 3.3° and 7.6° respectively. It is believed that mD-Track is the first implemented system based on COTS Wi-Fi hardware that can estimate AoD at such a high accuracy.
  • the localisation error results obtained using commodity Wi-Fi APs are presented in Figure 13.
  • the commodity AP is equipped with a limited number of antennas, i.e., three compared with eight at WARP, so all three antennas are used in these experiments.
  • the size of the bandwidths is varied along with the number of dimensions of mD-Track to evaluate the performance.
  • mD-Track 3D and 4D using the commodity AP are still able to achieve a median error of 0.67 m and 0.48 m with 40 MHz bandwdith and only three antennas.
  • Figures 12 and 13 depict the localisation error with variant signal bandwidth (20 and 40 MHz), antenna count (3, 4 and 8) and dimensionality (2D, 3D and 4D). From the results, a trend is observed that a higher number of dimensions leads to higher accuracy. With the same bandwidth and antenna count (e.g. 8 antennas and 40 MHz), mD-Track 4D achieves a much smaller median error (0.28 m) compared with mDTrack 3D (0.36 m) and 2D (1.16 m). It is also noted that increasing the number of antennas or bandwidth can improve the performance by improving the basic resolution in that dimension.
  • mD-Track 3D By fixing the bandwidth (40 MHz) and doubling the antenna count from 4 to 8 using WARP, mD-Track 3D reduces the median error from 0.57 m to 0.36 m. By fixing the antenna count (3), and doubling the bandwidth from 20 MHz to 40 MHz using COS Wi-Fi AP, mD-Track 3D reduces the error from 0.89 m to 0.67 m. It is noted that increasing the number of antennas (or radio chains) and bandwidth significantly increases the hardware overhead, especially for the COTS device. Increasing the dimensionality, however, does not incur additional hardware overhead. Effect of location and Doppler shift
  • Figure 14(a) shows an indoor floorplan map of a meeting room of 60 m 2 as used in this set-up. The floorplan was divided into 1 m x 1 m grids and a human target was stood in the center of each grid waving his hands.
  • Figure 14(b) shows the localisation error of mD-Track 2D in locating the person in each grid and
  • Figure 14(c) shows the localisation error of mD-Track 3D in locating the person in each grid, where the lightest grids depict the highest errors.
  • mD-Track is able to resolve signals from two close-by targets and successfully localise both of them, achieving a median error of 0.51 m.
  • mD-Track is able to locate four targets simultaneously with a median error of 0.47 m when they are 3 m apart from each other. The error is 0.94 m even when they are 0.5 m apart.
  • Embodiments of the invention like mD-Track are not only capable of tracking the location but also the motion of a target.
  • the capability of detecting four types of motions using mD-Track was investigated.
  • scenario one the transmitter was moved and the Doppler shift was estimated as shown in Figure 16(a).
  • scenario two the locations of the transceivers were fixed and a human target moved towards and away from the receiver, with the Doppler shift results being shown in Figure 16(b).
  • the human target stood still and only waved his hand back and forth and the results are shown in Figure 16(c).
  • the human target stood still and only moved his fingers back and forth and the results are shown in Figure 16(d).
  • mD-Track estimated the Doppler shift introduced by the different motions and the results are depicted in Figure 16.
  • mD-Track according to embodiments of the invention is very sensitive to motion and can accurately identify signals reflected from a target's body, his moving hands or fingers.
  • mD-Track can use Doppler shifts to detect motion, as well as ToF and AoA to locate multiple movements at the same time.
  • Doppler shifts to detect motion, as well as ToF and AoA to locate multiple movements at the same time.
  • ToF and AoA ToF and AoA
  • Figure 18(c) shows end-to-end execution time of SpotFi and mD-Track with varied granularities (i.e. step sizes).
  • the dimensionality and step size substantially affect the number of combinations for searching.
  • the step sizes of frequency were fixed to 0.1 Hz while the step size of angle and time was varied in order to evaluate the computational complexity of mD-Track 2D, 3D, and 4D, in comparison with SpotFi.
  • mD-Track 2D estimates AoA and ToF jointly, as SpotFi does, but runs much faster than SpotFi, as depicted in Figure 18(c).
  • the end-to-end execution time of mD-Track does not increase much when decreasing the step size or increasing the dimensionality.
  • a step size of [0.02 rad, 0.5 ns, 0.1 Hz] was used and as a result mD-Track 2D, 3D, and 4D were found to run 30x, 20x and 14x faster than SpotFi.
  • Received signal strength indication (RSSI) based indoor localisation techniques [references 5, 45] provide coarse localisation accuracies. More advanced techniques [references 10, 42] make use of an antenna array at the latest 802.11 n/802.1 1ac APs [references 31 , 47] to estimate the AoA of a signal emitted from a mobile device. These systems are able to achieve localisation accuracies of a few decimeters, which, however, are still vulnerable to rich multipaths because of the poor resolvability with limited number of antennas at the commodity Wi-Fi APs.
  • RSSI Received signal strength indication
  • Wi-Vi [reference 3] employs a signal nulling technique to cancel signals from static objects and then estimate an AoA of a signal reflected off a human. Wi-Vi is able to track the direction of a human's movement, but no location information can be obtained.
  • WiTrack [references 1 , 2] and other systems [references 6, 14, 21 , 33, 48, 49] use dedicated hardware with more than 1 GHz bandwidth to achieve a high resolution ToF for human tracking which is not possible with Wi-Fi. While some recent systems [references 16, 17] consider the use of both AoA and ToF for object tracking, their systems are designed to estimate two- dimensional parameters, and cannot be generalised easily to estimate parameters of more dimensions. Also the computational load in these systems is very high [references 16, 17].
  • MUSIC [reference 27] is widely employed to obtain AoA or ToF information.
  • SpotFi [reference 17] employs a modified version of MUSIC to jointly search two dimensions, which is similar to JADE [reference 34].
  • Wi-Deo [reference 16] estimates AoA and ToF using compressive sensing.
  • embodiments of the present invention like mDTrack provide a general platform that can be adapted to work with any number of dimensions.
  • mD-Track significantly reduces the computational load for a multidimensional search, e.g., mDTrack is 30 times faster than SpotFi.
  • embodiments of the present invention propose a maximum-likelihood based iterative path parameter estimation algorithm.
  • the described embodiment first adopts a coarse cancellation algorithm that resembles successive interference cancellation (SIC) for data communications, but is applied for multipath cancellation.
  • the interference cancellation based method is applied to resolve signals from one path, cancelling the interference from other paths and is repeated to resolve every resolvable path.
  • an iterative cancellation algorithm that recalls the structure of a Turbo Decoder, operating in a loop and iteratively refining the signal parameters is employed.
  • Embodiments of the present invention also perform a multi-dimensional search over each resolved signal to estimate its parameters. The above steps together form the proposed multidimensional parameter estimator according to embodiments of the invention.
  • embodiments of the invention such as the described mD-Track provide a general platform capable of incorporating information from as many dimensions as possible, to achieve the finest resolution in resolving multipath signals.
  • the experiments described demonstrate greatly improved performance for active and passive multi-target localisation and motion tracking.
  • the platform can be easily extended to support a myriad of applications including gesture recognition and Wi-Fi imaging.
  • JADE Job angle and delay estimation
  • JADE Job angle and delay estimation

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

L'invention concerne un appareil et un procédé de localisation et/ou de suivi. L'appareil comprend : un récepteur permettant de recevoir un signal reçu, le signal reçu comprenant une superposition de multiples signaux dérivés d'un signal sans fil transmis par un émetteur le long d'une pluralité de trajets différents ; un processeur informatique ; et un dispositif de stockage de données. Le dispositif de stockage de données comprend des instructions pouvant être exploitées par le processeur informatique afin de séparer le signal reçu en multiples signaux selon les étapes suivantes : a) estimer des paramètres de trajet initiaux : par l'estimation d'un nombre (L) de trajets de signal du signal reçu ; par l'estimation des paramètres de chaque trajet de signal à l'aide d'un outil d'estimation ; et par la dérivation d'un bruit de fond initial ; et b) raffiner les paramètres : i) par la combinaison du bruit de fond initial avec un trajet de signal reconstruit à l'aide de paramètres estimés courants et par la ré-estimation d'un ou de plusieurs paramètres de chaque trajet de signal à l'aide de l'outil d'estimation ; ii) par la ré-estimation du bruit de fond en reconstruisant chaque trajet de signal à l'aide des paramètres ré-estimés à partir de (i) et en soustrayant les L trajets de signal construits du signal reçu ; et iii) en répétant l'étape (i) et (ii) à l'aide du bruit de fond ré-estimé comme bruit de fond initial jusqu'à ce que les paramètres de chaque trajet de signal satisfont un critère de convergence ; et en utilisant les paramètres de chaque trajet de signal pour déterminer la localisation et/ou le mouvement d'une cible.
PCT/SG2018/050101 2017-03-03 2018-03-05 Appareil et procédé de localisation et/ou de suivi WO2018160141A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
SG10201701703V 2017-03-03
SG10201701703V 2017-03-03

Publications (1)

Publication Number Publication Date
WO2018160141A1 true WO2018160141A1 (fr) 2018-09-07

Family

ID=63369541

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/SG2018/050101 WO2018160141A1 (fr) 2017-03-03 2018-03-05 Appareil et procédé de localisation et/ou de suivi

Country Status (1)

Country Link
WO (1) WO2018160141A1 (fr)

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020099646A1 (fr) * 2018-11-16 2020-05-22 Sony Corporation Dispositif et procédé de communication
CN111289967A (zh) * 2020-03-31 2020-06-16 四川长虹电器股份有限公司 基于毫米波雷达的人员检测跟踪与计数算法
US20200309938A1 (en) * 2017-09-12 2020-10-01 Senstation Technologies Ltd. A method and system for detection of a target by a passive radar system exploiting multichannel-per-carrier illuminator sources
CN112315452A (zh) * 2020-10-29 2021-02-05 中国科学技术大学 基于多路径相位相消的人体呼吸追踪方法、装置及系统
TWI721665B (zh) * 2019-11-22 2021-03-11 光禾感知科技股份有限公司 基於磁場強度的定位系統和定位方法
CN112924928A (zh) * 2021-01-26 2021-06-08 西安交通大学 一种基于路径分离的室内Wi-Fi多人检测方法
CN113923589A (zh) * 2021-09-24 2022-01-11 郑州大学 一种非接触式多目标定位方法
US20220078050A1 (en) * 2018-12-17 2022-03-10 U-Blox Ag Estimating one or more characteristics of a communications channel
CN114175534A (zh) * 2019-07-24 2022-03-11 赛普拉斯半导体公司 利用频谱分集进行基于机器学习的射频信号参数估计
CN114185002A (zh) * 2021-12-09 2022-03-15 重庆邮电大学 一种基于波束空间矩阵束的三维参数估计方法
CN114884781A (zh) * 2022-03-30 2022-08-09 西安电子科技大学 一种基于线性规划和多信号分类算法的载波频偏估计方法
WO2022235347A1 (fr) * 2021-05-07 2022-11-10 Qualcomm Incorporated Génération de cartes intérieures sur la base d'une détection de fréquence radio
EP3852398B1 (fr) * 2020-01-15 2023-09-20 Nokia Technologies Oy Détection de localisation sans fil
CN117452332A (zh) * 2023-12-26 2024-01-26 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) 一种位置确定方法及装置

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102098705A (zh) * 2011-03-15 2011-06-15 北京邮电大学 多维信道参数提取方法

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102098705A (zh) * 2011-03-15 2011-06-15 北京邮电大学 多维信道参数提取方法

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
CHONG C. C. ET AL.: "Joint detection-estimation of directional channel parameters using the 2-D frequency domain SAGE algorithm with serial interference cancellation", IEEE INTERNATIONAL CONFERENCE ON COMMUNICATIONS, 2002, vol. 2, 2 May 2002 (2002-05-02), pages 906 - 910, XP010589625, [retrieved on 20180427] *
FLEURY B. H. ET AL.: "Channel Parameter Estimation in Mobile Radio Environments Using the SAGE Algorithm", IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, vol. 17, no. 3, 31 March 1999 (1999-03-31), pages 434 - 445, XP055539098, [retrieved on 20180427] *
LI J. ET AL.: "Joint Estimation of Channel Parameters for MIMO Communication Systems", 2ND INTERNATIONAL SYMPOSIUM ON WIRELESS COMMUNICATION SYSTEMS , 2005, 7 September 2005 (2005-09-07), pages 22 - 26, XP055539106, [retrieved on 20180427] *
WAX M. ET AL.: "Joint Estimation of Time Delays and Directions of Arrival of Multiple Reflections of a Known Signal", IEEE TRANSACTIONS ON SIGNAL PROCESSING, vol. 45, no. 10, 31 October 1997 (1997-10-31), pages 2477 - 2484, XP055539104, [retrieved on 20180427] *

Cited By (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200309938A1 (en) * 2017-09-12 2020-10-01 Senstation Technologies Ltd. A method and system for detection of a target by a passive radar system exploiting multichannel-per-carrier illuminator sources
US11460541B2 (en) * 2017-09-12 2022-10-04 Senstation Technologies Ltd. Method and system for detection of a target by a passive radar system exploiting multichannel-per-carrier illuminator sources
WO2020099646A1 (fr) * 2018-11-16 2020-05-22 Sony Corporation Dispositif et procédé de communication
US11841451B2 (en) 2018-11-16 2023-12-12 Sony Corporation Communication device and method
US11601307B2 (en) * 2018-12-17 2023-03-07 U-Blox Ag Estimating one or more characteristics of a communications channel
US20220078050A1 (en) * 2018-12-17 2022-03-10 U-Blox Ag Estimating one or more characteristics of a communications channel
CN114175534A (zh) * 2019-07-24 2022-03-11 赛普拉斯半导体公司 利用频谱分集进行基于机器学习的射频信号参数估计
CN114175534B (zh) * 2019-07-24 2024-06-07 赛普拉斯半导体公司 用于射频信号参数的估计的方法、系统和设备
TWI721665B (zh) * 2019-11-22 2021-03-11 光禾感知科技股份有限公司 基於磁場強度的定位系統和定位方法
EP3852398B1 (fr) * 2020-01-15 2023-09-20 Nokia Technologies Oy Détection de localisation sans fil
CN111289967A (zh) * 2020-03-31 2020-06-16 四川长虹电器股份有限公司 基于毫米波雷达的人员检测跟踪与计数算法
CN112315452A (zh) * 2020-10-29 2021-02-05 中国科学技术大学 基于多路径相位相消的人体呼吸追踪方法、装置及系统
CN112924928A (zh) * 2021-01-26 2021-06-08 西安交通大学 一种基于路径分离的室内Wi-Fi多人检测方法
CN112924928B (zh) * 2021-01-26 2023-05-02 西安交通大学 一种基于路径分离的室内Wi-Fi多人检测方法
US11885899B2 (en) 2021-05-07 2024-01-30 Qualcomm Incorporated Generating indoor maps based on radio frequency sensing
WO2022235347A1 (fr) * 2021-05-07 2022-11-10 Qualcomm Incorporated Génération de cartes intérieures sur la base d'une détection de fréquence radio
CN113923589B (zh) * 2021-09-24 2023-05-23 郑州大学 一种非接触式多目标定位方法
CN113923589A (zh) * 2021-09-24 2022-01-11 郑州大学 一种非接触式多目标定位方法
CN114185002A (zh) * 2021-12-09 2022-03-15 重庆邮电大学 一种基于波束空间矩阵束的三维参数估计方法
CN114884781A (zh) * 2022-03-30 2022-08-09 西安电子科技大学 一种基于线性规划和多信号分类算法的载波频偏估计方法
CN117452332A (zh) * 2023-12-26 2024-01-26 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) 一种位置确定方法及装置
CN117452332B (zh) * 2023-12-26 2024-03-12 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) 一种位置确定方法及装置

Similar Documents

Publication Publication Date Title
Xie et al. mD-Track: Leveraging multi-dimensionality for passive indoor Wi-Fi tracking
WO2018160141A1 (fr) Appareil et procédé de localisation et/ou de suivi
Soltanaghaei et al. Multipath triangulation: Decimeter-level wifi localization and orientation with a single unaided receiver
Chen et al. Achieving centimeter-accuracy indoor localization on WiFi platforms: A multi-antenna approach
Kotaru et al. Spotfi: Decimeter level localization using wifi
Chen et al. $ M^ 3$ M 3: Multipath assisted Wi-Fi localization with a single access point
Karanam et al. Tracking from one side: Multi-person passive tracking with WiFi magnitude measurements
Gjengset et al. Phaser: Enabling phased array signal processing on commodity WiFi access points
Xie et al. xD-track: Leveraging multi-dimensional information for passive wi-fi tracking
WO2015114355A1 (fr) Appareil et procédé permettant d'étalonner un point d'accès sans fil comprenant un réseau de plusieurs antennes
EP3100069A1 (fr) Appareil et procédé de détermination de l'emplacement d'un dispositif mobile à l'aide de plusieurs points d'accès sans fil
KR100979608B1 (ko) 두개의 안테나들 또는 이의 균등물을 사용하여 네비게이션비컨 신호를 검출하는 방법
Pizarro et al. Accurate ubiquitous localization with off-the-shelf IEEE 802.11 ac devices
Chen et al. AWL: Turning spatial aliasing from foe to friend for accurate WiFi localization
Yu et al. Compressive sensing for MIMO radar
Bnilam et al. ANGLE: ANGular location estimation algorithms
Wang et al. CSI component reconstruction-based AoA estimation for subtle human-induced reflection under the TTW scenario
Venkatnarayan et al. Leveraging polarization of WiFi signals to simultaneously track multiple people
Zhu et al. WiseFi: Activity localization and recognition on commodity off-the-shelf WiFi devices
Rahman et al. Lochunt: Angle of arrival based location estimation in harsh multipath environments
Wang et al. TWPalo: Through-the-wall passive localization of moving human with Wi-Fi
Sun et al. An indoor environment sensing and localization system via mmWave phased array
Yang et al. Multiple wifi access points co-localization through joint aoa estimation
JP2018112539A (ja) センサおよび位置推定方法
Xiong Pushing the limits of indoor localization in today’s Wi-Fi networks

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18761268

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18761268

Country of ref document: EP

Kind code of ref document: A1