mTrack: High-Precision Passive Tracking Using Millimeter Wave Radios

Teng Wei and Xinyu Zhang

University of Wisconsin – Madison

Near-field Wireless Tracking

Tracking objectives at mm-level accuracy

Turn any surface into interactive virtual touchscreen

Enable a new form of pervasive user-computer interface

Virtual Trackpad Interactive Display Tracking Whiteboard

State-of-the Art

C. Xu, etalSCPL: Indoor Device-free Multi-

subject Counting and Localization Using Radio

Signal Strength IEEE IPSN, 2013.

TagoramMobiCom

Radio-based tracking system

m-level dm-level cm-level mm-level

RF-IDrawSIGCOMM

PinLocMobiSys

H.Fang, 60GHz RSS Localization with

Omni-directional and Horn Antennas, Ph.D.

dissertation, 2010.

WiViSIGCOMM

WiTrackNSDI ?

Active

Passive

New Challenges

Weak signal intensity of passive reflection

Passive Fine-grained Tracking

Target does not modulate and emit signals

Irrelevant reflection from unintended objectives

Time-varying multipath reflection from background

Locating initial position with few number of devices

Especially from small objects, like pen

Costly to deploy substantial nodes

Overview the Basic Idea

60GHz laser-like directional beam ❶

Tx

RxPen

Rx2

Flexible beam-steering capability❷

5mm extremely short wavelength❸

Quasi-omni-directional illumination❺

Interactive diffusion from small objects❹

Understanding mmWave Passive Tracking

Feasibility Study

Tx

30cmPen 0.8cm

Rx

Diffusive Reflection

15~20dB

Tx

50cm60cm

Moving

Rx

Fine-grained Tracking

Tx

50cm

Rx

Initial Locating

Key Challenge: Background Reflection

Tx

Rx

Objects in the background

Background Reflection

Target Reflection

0

2𝜋

λ /2 λ 3 λ /2 2 λTarget movement

Phase ofReceived

Signal

Background Dominated

0

2𝜋

λ /2 λ 3 λ /2 2 λTarget movement

Phase ofReceived

Signal

Less than 2

Target Dominated

Rx

Target DominatedBackground Dominated

Naïve Solution

DC-filter the decoded symbols

Received signal Target

reflection

Backgroundreflection

I

Q

1, 0, 1, 0, …

Unmodulated

Modulated

Filter the received waveform (RFID)

Require target to modulate the reflect signal

I

Q

static background removed

Dual-differential Background Removal (DDBR)

Key Observation

Background reflection remains similar in consecutive samples

Differential cancels the background reflection

12[Δarg(�⃗�𝒕𝒓𝒈)𝑡 −1

𝑡 +Δarg (�⃗�𝒕𝒓𝒈)𝑡𝑡+1]=arg (�⃗�𝑟𝑒𝑐

𝑡+1− �⃗�𝑟𝑒𝑐𝑡 )−arg (�⃗�𝑟𝑒𝑐

𝑡 − �⃗�𝑟𝑒𝑐𝑡 −1 )

Lemma (DDBR): The average phase shift among three consecutive samples is

Average phase shift Diff. phase of sample differential

Sample differential

531

-1-3-4

Target movement

DDBR received signals

Phas

e

Advantage and Limitation

Handle time-varying background reflection

Simple computation of processing

Suitable for hardware implementation

Cons of DDBR

Vulnerable to the phase noise

60GHz COTS device has non-negligible phase noise

phase noise > phase shift

Pros of DDBR

Phase Counting and Regeneration (PCR)

Periodicity Pattern of Phase I (TD) II (BD) III (ITM)

0

2𝜋

λ /2 λ 3 λ /22 λTarget movement

Pha

se

0

2𝜋

λ /2 λ 3 λ /2 2 λTarget movement

Pha

se

0

2𝜋

λ /2 λ 3 λ /22 λ

Pha

se

Target movement

0 50 100 150 200 250 300 350

Case (I ) Case ( II )∧( III ) Case (I )

Sample index

30

-3

1030

-3

30

-5

PCR Algorithm

Reducing ITM to BDStep 1

Periodicity Counting Step 2

RegenerationStep 3

Input phase

Anchor Point Acquisition (APA)

Complementary to Tracking

Initial location for successive tracking

Prevent error accumulation

Calibrate tracking result

Discrete Beam Steering

Spline interpolation improves granularity of APA

reduce error

True direction

Background Reflection

Enhance 10dB

contrast

RSS subtraction improves contrast of APA

BG Pen

Touch Event Detection

e.g., start/pause of tracking

Detect touch gestures as control command

Gesture and Feature Space

Touch

LiftClick

Phase shift❶

Variance of phase shift❷

RSS❸

Event detection: Variance of phase shift

Event Classification: RSS

Decisiontreerule

Touch Lift Click

Implementation and Evaluation

Horn Antenna

MotorizedRotator

60 GHz RFFront-end (Rx)

High SpeedADC/DACWARP Board

PHY Extraction

Tracking

Locating

Touchdetection

AppsmTrack

60GHz SDR testbed Algorithm implementation

Testing objects

Metal-surfaced pen

Marker

Pencil

Passive Tracking

Rx 1Tx

Rx 2

Drywall

Cabinet

1m 1m⨉10cm10cm2m

1.5mExample trajectory of tracking

Error map over tracking region

Tracking Setup Result

Achieve high-precision tracking

1cm

3cm

Anchor Positioning and Event Detection

Event Touch Lift Click ND

Touch 94.0% 0 0 6.0%

Lift 0 93.5% 0 6.5%

Click 0 0 94.8% 5.2%

APA Performance

Randomly placed 30 positions

Beam-steering at step of

RSS: 12.3dB, 10.1dB and 4.7dBAverage error of 1.5 cm, 2 cm

and 6 cm

Event Detection

7 users

Each provides a 10-sampletraining set

20~50-sample testing set

Application: Trackpad

Experiment Setup

Example word Recognition Accuracy

Integrate mTrack into word-recognition application

Record hand-writing trace from mTrack

Export and control mouse of a PC

MyScript© Stylus for word detection

Conclusion

First RF-based system that achieves sub-centimeter scale passive object tracking

Implement on a configurable 60GHz radio testbed

Validate performance in a wireless trackpad setup

Resolve new practical challenges in passive tracking/locating

DDBR algorithm for addressing background reflection

PCR algorithm for mitigating phase noise issue

RSS interpolation and subtraction for improving granularity and contrast.

Questions?

Thank you