• 沒有找到結果。

RF-based Internet-of-Thing (IoT) Techniques

N/A
N/A
Protected

Academic year: 2022

Share "RF-based Internet-of-Thing (IoT) Techniques"

Copied!
30
0
0

加載中.... (立即查看全文)

全文

(1)

RF-based Internet-of-Thing 
 (IoT) Techniques

Kate C.-J. Lin Academia Sinica

!

2015.05.29

(2)

• Localization!

• Object Tracking!

• Gesture Detection!

• Human Computer Interaction!

• Health Care

(3)

See through walls

[sigcomm’13]

(4)

Applications

Police'avoids'ambush' Firefighters'check'for'humans'

Personal'security' Gesture'interface'from'

behind'a'wall'

(5)

Key Idea

Track people from WiFi signals reflected by human body

(6)

Challenges

Wall refection is 10,000x stronger than reflections 


coming from behind the wall!

(7)

How to Eliminate the Wall’s Reflection?

Idea: transmit two waves that cancel each other when they reflect off static objects, but not moving object

Wall is static

People tent to move

disappears

detectable

(8)

Transmit) Antennas) Receive)Antenna:)

αx # x #

Leveraging MIMO techniques

(9)

Eliminating Walls’ Reflection

h 2#

h 1#

y#=#h 1 #x#+#h 2 αx # 0 "

α"="%h 1 "/"h 2 " "

Receive#Antenna:#

αx "

x "

(10)

Eliminating All Static Reflection

(11)

Motion tracking via Radio Reflection

[NSDI’14]

(12)

WiTrack

• Centimeter-scale motion tracking using only radio reflections off the human body!

• Works behind walls and does not require

person to hold any device

(13)

How WiTrack Works?

Distance = Reflection Time x Speed of Light

Tx

Rx

(14)

Naive Solution:


Measuring Reflection time

Time

%

Tx#pulse# Rx#pulse#

Reflec)on%Time%

Transmit short pulse and listen for echo

(15)

Time

%

Tx#pulse# Rx#pulse#

Reflec)on%Time%

Naive Solution:


Measuring Reflection time

Transmit short pulse and listen for echo

Capturing the pulse needs sub-nanosecond sampling Multi-GHz samplers are expensive and

have high noise → Impractical

(16)

Frequency Modulated Carrier Wave (FMCW)

Time

%

Fr eq ue nc y

%

Transmi/ed

%

t t+ΔT%

Received

%

Reflec%on(Time(

ΔF(

slope( ΔF(

=(

ΔF → Reflection Time → Distance !

(17)

How to Measure ΔF

• Subtracting frequency is easy!

• Done using a mixer → signals whose frequency is ΔF

Mixer &

Transmi,ed

&

Received

&

FFT &

Power

&

ΔF#

(18)

Removing Static Multi-Path

0 5 10 15 20

Time (seconds) 0

5 10 15 20 25 30

Distance (meters)

(a) Spectrogram

0 5 10 15 20

Time (seconds) 0

5 10 15 20 25 30

Distance (meters)

(b) After Background Subtraction

0 5 10 15 20 25 30

0 5 10 15 20

Distance (in meters)

Time (in seconds)

Contour Denoised Contour

(c) Contour Tracking

Figure 3—Obtaining the Time-of-Flight (TOF) Estimates. WiTrack takes an FFT of the received signal in baseband over every sweep period to generate the spectrogram in (a). Then, by subtracting out a given frame from the frame that precedes it, WiTrack eliminates static multipath as in (b).

The blue plot in (c) shows how WiTrack can address dynamic multipath by tracking the bottom contour of (b), and then denoise the signal (red plot) to obtain a clean TOF estimate.

and sweep its carrier frequency across a wide bandwidth of multiple GHz.

In our design we chose the following parameter for our FMCW. We have built an FMCW system that sweeps a total bandwidth of 1.69 GHz from 5.56 GHz to 7.25 GHz, and transmits at 0.75 milliwatt. The choice of this band- width has been dictated by the FCC regulations for civil- ian use of spectrum [9]. Specifically, it is the largest con- tiguous bandwidth below 10 GHz which is available for civilian use at low power.

Based on Eq. 3, our sweep bandwidth allows us to obtain a distance resolution of 8.8 cm. Hence the aver- age error in mapping TOF to distance in 1D is about 4.4 cm. Note that the above derivation neglects the im- pact of noise, and hence provides a lower bound on the achievable resolution. In practice, the system’s resolution is affected by the noise level. It also depends on the geo- metric model that maps TOFs to 3D locations.

3.2 Addressing Static Multi-path

The next step in WiTrack’s operation is to distinguish a human’s reflections from reflections off other objects in the environment, like furniture and walls. Recall from the previous section that every reflector in the environ- ment contributes a component to the overall received sig- nal, and that component has a frequency shift that is lin- early related to the time-of-flight of the reflection based on Eq. 1. Typically, reflections from walls and furniture are much stronger than reflections from a human, especially if the human is behind a wall. Unless these reflections are removed, they would mask the signal coming from the human and prevent sensing her motion. This behavior is called the “Flash Effect”.

To remove reflections from all of these static objects (walls, furniture), we leverage the fact that since these reflectors are static, their distance to the WiTrack device does not change over time, and therefore their induced fre- quency shift stays constant over time. Fig. 3(a) plots the spectrogram of the received signal as a function of time, for one of the receive antennas of WiTrack. In particular,

we take the FFT of the received signal every sweep win- dow, and compute the power in each frequency as a func- tion of time. Note that there is a linear relation between frequency shifts and the traveled distances as follows:

distance = C ⇥ TOF = C ⇥ Df

slope . (4)

Thus, instead of plotting the power in each frequency as a function of time, we can use the above equation to plot the power reflected from each distance as a function of time, as shown in Fig. 3(a). The color code of the plot corre- sponds to a heat-map of the power in the reflected signal.

Strong reflectors are indicated by red and orange colors, weaker reflectors are indicated by yellow and green, and the absence of a reflector is indicated by blue at the corre- sponding frequency. The figure indicates the presence of very strong static reflectors in the environment. Specifi- cally, it has many horizontal stripes; each of these stripes signifies the presence of a reflector at the corresponding round-trip distance. Because these stripes are horizontal, their corresponding reflectors are stationary over time.

Hence, we eliminate the power from these static reflec- tors by simply subtracting the output of the FFT in a given sweep from the FFT of the signal in the previous sweep.

This process is called background subtraction because it eliminates all the static reflectors in the background.

Fig. 3(b) is the result of applying background subtrac- tion to Fig. 3(a). The figure shows that all static reflec- tors corresponding to the horizontal lines have been elim- inated. This makes it easier to see the much weaker reflec- tions from a moving human. Specifically, we see that the distance of the dominant reflector (the red color signal) is varying with time, indicating that the reflector is moving.

3.3 Addressing Dynamic Multi-path

By eliminating all reflections from static objects, WiTrack is left only with reflections from a moving hu- man (see Fig. 3(b)). These reflections include both signals that bounce off the human body to the receive antennas, and those that bounce off the human then bounce off other

4

Time of Flight (ToF) reflected from static objects stays constant!

→ Subtract the background signals and extract moving objects

(19)

From Distance to Localization

Can be anywhere on an ellipse whose foci are (Tx, Rx)!

Rx# Tx#

d

#

One ellipse is not enough for localization

(20)

Using Multiple Receive Antennas

Rx# Tx# Rx’#

d

#

in#beam#

d’

#

Extend to 3D by using 3 Rx antennas and taking

the intersection of ellipsoids

(21)

Monitoring Breathing and Heart Rate

[CHI’15]

Smart Homes that Monitor Breathing and Heart Rate

Fadel Adib Hongzi Mao Zachary Kabelac Dina Katabi Robert C. Miller Massachusetts Institute of Technology

32 Vassar Street, Cambridge, MA 02139

{ fadel,hongzi,zek,dk,rcm } @mit.edu

ABSTRACT

The evolution of ubiquitous sensing technologies has led to intelligent environments that can monitor and react to our daily activities, such as adapting our heating and cooling sys- tems, responding to our gestures, and monitoring our elderly.

In this paper, we ask whether it is possible for smart en- vironments to monitor our vital signs remotely, without in- strumenting our bodies. We introduce Vital-Radio, a wire- less sensing technology that monitors breathing and heart rate without body contact. Vital-Radio exploits the fact that wire- less signals are affected by motion in the environment, in- cluding chest movements due to inhaling and exhaling and skin vibrations due to heartbeats. We describe the operation of Vital-Radio and demonstrate through a user study that it can track users’ breathing and heart rates with a median ac- curacy of 99%, even when users are 8 meters away from the device, or in a different room. Furthermore, it can monitor the vital signs of multiple people simultaneously. We envision that Vital-Radio can enable smart homes that monitor peo- ple’s vital signs without body instrumentation, and actively contribute to their inhabitants’ well-being.

Author Keywords Wireless; Vital Signs; Breathing; Smart Homes; Seeing Through Walls; Well-being

Categories and Subject Descriptors H.5.2. Information Interfaces and Presentation: User Interfaces - Input devices and strategies. C.2.2. Network Architecture and Design:

Wireless Communication.

INTRODUCTION

The past few years have witnessed a surge of interest in ubiq- uitous health monitoring [22, 25]. Today, we see smart homes that continuously monitor temperature and air quality and use this information to improve the comfort of their inhab- itants [46, 32]. As health-monitoring technologies advance further, we envision that future smart homes would not only monitor our environment, but also monitor our vital signals, like breathing and heartbeats. They may use this information to enhance our health-awareness, answering questions like

“Do my breathing and heart rates reflect a healthy lifestyle?”

They may also help address some of our concerns by an- swering questions like “Does my child breathe normally dur- ing sleep?” or “Does my elderly parent experience irregular

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full cita- tion on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or re- publish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from Permissions@acm.org.

CHI 2015, April 18 - 23 2015, Seoul, Republic of Korea

Copyright is held by the owner/author(s). Publication rights licensed to ACM.

ACM 978-1-4503-3145-6/15/04...$15.00 http://dx.doi.org/10.1145/2702123.2702200

(a) Inhale Motion (b) Exhale Motion

Figure 1—Chest Motion Changes the Signal Reflection Time. (a) shows that when the person inhales, his chest expands and becomes closer to the antenna, hence decreasing the time it takes the signal to reflect back to the device. (b) shows that when the person ex- hales, his chest contracts and moves away from the antenna, hence the distance between the chest and the antenna increases, causing an increase in the reflection time.

heartbeats?” Furthermore, if non-intrusive in-home continu- ous monitoring of breathing and heartbeats existed, it would enable healthcare professionals to study how these signals correlate with our stress level and evolve with time and age, which could have a major impact on our healthcare system.

Unfortunately, typical technologies for tracking vital signals require body contact, and most of them are intrusive. Specif- ically, today’s breath monitoring sensors are inconvenient:

they require the person to attach a nasal probe [19], wear a chest band [43], or lie on a special mattress [3]. Some heart- rate monitoring technologies are equally cumbersome since they require their users to wear a chest strap [18], or place a pulse oximeter on their finger [21]. The more comfortable technologies such as wristbands do not capture breathing and have lower accuracy for heart rate monitoring [12]. Addition- ally, there is a section of the population for whom wearable sensors are undesirable. For example, the elderly typically feel encumbered or ashamed by wearable devices [20, 37], and those with dementia may forget to wear them. Children may remove them and lose them, and infants may develop skin irritation from wearable sensors [40].

In this paper, we ask whether it’s possible for smart homes to monitor our vital signs remotely – i.e., without requiring any physical contact with our bodies. While past research has investigated the feasibility of sensing breathing and heart rate without direct contact with the body [17, 16, 15, 34, 27, 48, 14], the proposed methods are more appropriate for con- trolled settings but unsuitable for smart homes: They fail in the presence of multiple users or extraneous motion. They typically require the user to lie still on a bed facing the device.

Furthermore, they are accurate only when they are within

close proximity to the user’s chest.

(22)

How? Measuring Variation

reflection times. Since wireless signals travel at the speed of light, signals reflected off objects at different distances would fall into different buckets.

However, in contrast to past work on localization, which uses FMCW to sense the amount of power arriving from different distances to localize the users, Vital-Radio uses the FMCW technique as a filter –i.e., it uses it to isolate the reflected sig- nals arriving from different distances in the environment into different buckets, before it proceeds to analyze the signals in each of these buckets to extract the vital signs (step 2 below).

Our implementation of FMCW follows the system in [6], where the resolution of FMCW buckets is about 8 cm. This has two implications:

• Reflections from two objects that are separated by at least 8 cm would fall into different buckets. Hence, two users that are few feet apart would naturally fall into different buckets. For example, in Fig. 2, the wall, Bob, the table, and Alice are at different distances from our device, and hence FMCW isolates the signals reflected from each of these entities into different buckets, allowing us to focus on each of them separately.

• Using FMCW as a filter also allows us to isolate some of the limb motion from chest movements due to breathing and heartbeats. For example, the signal reflected off the user’s feet will be in a different bucket from that reflected off the user’s chest. Thus, having the user move his feet (in place) does not interfere with Vital-Radio’s ability to extract the breathing and heart rate.

After bucketing the reflections based on the reflector’s dis- tance, Vital-Radio eliminates reflections off static objects like walls and furniture. Specifically, since static objects don’t move, their reflections don’t change over time, and hence can be eliminated by subtracting consecutive time measurements.

At the end of this step, Vital-Radio would have eliminated all signal reflections from static objects (e.g., walls and furni- ture), and is left with reflections off moving objects separated into buckets.

2

Step 2: Identifying Reflections Involving Breathing and Heart Rate

After Vital-Radio isolates reflections from different moving users into separate buckets, it proceeds by analyzing each of these buckets to identify breathing and heart rate. For exam- ple, in Fig. 2, we would like to identify whether the user in bucket 2 is quasi-static and his motion is dominated by his vi- tal signs, or whether he is walking around or moving a limb.

To do that, Vital-Radio zooms in on the signal reflection which it isolated in the corresponding bucket. This wireless reflection is a wave; the phase of the wave is related to the distance traveled by the signal as follows [39]:

( t) = 2⇡ d(t)

, (1)

2

While unlikely, it is possible that multiple users are at the same distance from the device and hence fall into the same bucket. To deal with such cases, one may deploy multiple devices so that if two users are at the same distance with respect to one device, they are at different distances with respect to another device.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4

Phase (in radians)

Time (in minutes)

-1.5 -1 -0.5 0 0.5 1 1.5

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Phase (in radians)

Time (in minutes)

Exhale'

Inhale'

Heartbeats'

Figure 3—Phase variation due to vital signs. The figure shows the variations in phase due to breathing and heartbeats, where peaks and valleys in the phase correspond to exhale and inhale motions respectively; also, zooming in on the signal allows us to observe the heartbeats modulated on top of the breathing motion.

where is the wavelength of the transmitted signal, and d(t) is the traveled distance from the device to the reflector and back to the device. The above equation shows that one can identify variations in d(t) due to inhaling, exhaling, and heart- beats, by measuring the resulting variations in the phase of the reflected signal.

To illustrate how the phase varies with vital signs, let us con- sider the example in Fig. 1, where a user sits facing the device.

When the person inhales, his chest expands and gets closer to the device; and when he exhales, his chest contracts and gets further away from the device. Because the phase and the dis- tance to a reflector are linearly related, Vital-Radio can track a person’s breathing. Fig. 3 shows the phase of the captured re- flection as a function of time. Specifically, a peak in the phase corresponds to an exhale (highest distance from the device), and a valley in the phase corresponds to an inhale (smallest distance from the device). We note that our implementation uses a wavelength around 4.5 cm. According to the above equation, sub-centimeter variations in the chest distance due to breathing cause sub-radian variations in the phase, which is what we observe in the figure.

Similarly, a person’s heartbeats cause minute movements of different parts of his body. Specifically, the physiological phenomenon that allows Vital-Radio to extract heart rate from signal reflections is ballistocardiography (BCG). BCG refers to movements of the body synchronous with the heartbeat due to ventricular pump activity [36]. Past work has documented BCG jitters from the head, torso, buttock, etc. [5, 8]. Periodic jitters cause periodic variations in the wireless signal allow- ing us to capture the heart rate. These movements translate to smaller fluctuations on top of the breathing motion in the wireless reflection as we can see from local peaks in Fig. 3.

Note that the periodicity of breathing and heartbeats is inde- pendent of the user’s orientation. For example, if the user has his back to the device, the valleys become peaks and vice versa, but the same periodicity persists.

Still, an important question to answer is: what happens when a person moves around or moves a limb, and how can Vital- Radio distinguish such motions from breathing and heart-

• First localize the user and identify its reflection!

• Zoom in on the reflected signal and analyze variations

(23)

Gesture Recognition Using Wireless Signals

[MobiCom’13]

rithm in more detail and show how to make it applicable to existing 802.11 frames.

(b) How can we deal with other humans in the environment?

A typical home may have multiple people who can affect the wireless signals at the same time. WiSee uses the MIMO capability that is inherent to 802.11n, to focus on gestures from a particular user. MIMO provides throughput gains by enabling multiple transmitters to concurrently send packets to a MIMO receiver. If we consider the wireless reflections from each human as signals from a wireless transmitter, then they can be separated using a MIMO receiver.

Traditional MIMO decoding, however, relies on estimating the channel between the transmitter and receiver antennas.

These channels are typically estimated by sending a distinct known preamble from each transmitter. Such a known signal structure is not available in our system since the human body reflects the same 802.11 transmitter’s signals.

Our solution to this problem is inspired by the trigger approach taken by many multi-user games that use Xbox Kinect, in which a user gains control of the interface by performing a specific gesture pattern. In WiSee the target human performs a repetitive gesture, which we use as that person’s preamble. A WiSee receiver leverages this preamble to estimate the MIMO channel that maximizes the energy of the reflections from the user. Once the receiver locks on to this channel, the user performs normal (non-repetitive) gestures that the receiver classifies using the Doppler shifts.

In §3.3, we explore this idea further and show how to ex- tract the preamble without requiring the human to perform gestures at a pre-determined speed.

The WiSee proof-of-concept is implemented in GNURadio using the USRP-N210 hardware. We classify the gestures from the Doppler shifts using a simple pattern-matching algorithm described in §3.2. We evaluated WiSee with a total of five users in both an office environment and a two-bedroom apartment whose layout is shown in Fig. 7.

We performed gestures in a number of scenarios including line-of-sight, non-line-of-sight, and through-the-wall scenar- ios where the person is in a different room from the wireless transmitter and the receiver. The users perform a total of 900 gestures across the locations.

Our findings are as follows:

WiSee can classify the nine whole-body gestures shown in Fig. 1, with an average accuracy of 94%. This is promising, given that the accuracy for random guesses is 11.1%.

Using a 4-antenna receiver and a single-antenna trans- mitter placed in the living room, WiSee can achieve the above classification accuracy in 60% of the home loca- tions. Adding an additional single-antenna transmitter to the living room achieves the above accuracy in locations across all the rooms. Thus, with a WiSee-enabled Wi-Fi AP acting as a receiver and a couple of mobile devices act- ing as transmitters, WiSee can enable whole-home gesture recognition.

Over a 24-hour period, WiSee’s average false positive rate—events that detect a gesture in the absence of the target human—is 2.63 events per hour when using a preamble with two gesture repetitions. This goes down to 0.07 events per hour, when the number of repetitions is increased to four.

Using a 5-antenna receiver and a single-antenna transmit- ter, WiSee can successfully perform gesture classification, in the presence of three other users performing random

Figure 1—Gesture sketches: WiSee can detect and clas- sify these nine gestures in line-of-sight, non-line-of-sight, and through-the-wall scenarios with an average accuracy of 94%.

gestures. However, the classification accuracy reduces as we further increase the number of interfering users. This is a limitation of WiSee: Given a fixed number of trans- mitters and receiver antennas, the accuracy reduces with the number of users. However, since typical home scenar- ios do not have a large number of users in a single room, WiSee can enable a large set of interaction applications for always-available computing in home environments.

Contributions: We make the following contributions:

We introduce the first wireless system that enables gesture recognition in line-of-sight, non-line-of-sight, and through- the-wall scenarios.

We present algorithms to extract gesture information from communication-based wireless signals. Specifically, we show how to extract minute Doppler shifts from wide- band OFDM transmissions that are typical to most mod- ern communication systems including Wi-Fi.

Finally, using a proof-of-concept prototype, we demon- strate that our system can detect a set of nine whole-body gestures in typical environments.

This paper takes the first step towards leveraging existing wireless networks to enable novel human-computer interac- tion mechanisms such as whole-home gesture recognition.

We hope that this line of work would open up a number of research opportunities at the intersection of wireless net- working and HCI, and bridge the two communities.

2. Related Work

Our work is related to prior art in both wireless systems and in-air gesture recognition systems.

(a) Wireless Systems: One can classify the related work in this domain into three main categories: wireless localiza- tion, wireless tomography, and through-the-wall radar sys-

(24)

Idea: Doppler shift

• Frequency change of a wave occurs as its source moves relative to the observer!

• Different moving patterns result in various Doppler shifts!

• Waves propagating at the speed of light (C m/s)!

• Assume a person moves at a speed of v!

• Doppler shift Δf=2fv/C (e.g., 0.5m/s corresponding

to 17Hz Doppler shift on a 5GHz band)

(25)

Challenges

• Small Doppler shift (e.g., 10-20Hz shift) is not detectable in wide-band OFDM channels (e.g., 20MHz/64 ~313KHz per subcarrier)

many more point object moving at the same time. Thus, it creates Doppler signals with much larger energy than when the human uses only parts of her body.

Challenge: Human motion results in a very small Doppler shift that can be hard to detect from a typical wireless transmission (e.g., Wi-Fi, WiMax, LTE, etc.). For instance, consider a user moving her hand towards the receiver at 0.5 m/sec. From Eq. 1, this results in a Doppler shift of about 17 Hertz for a Wi-Fi signal transmitted at 5 GHz (θ = 0). Since the bandwidth of Wi-Fi’s transmissions is at least 20 MHz, the resulting Doppler shift is orders of magni- tude smaller than Wi-Fi’s bandwidth. Identifying such small Doppler shifts from these transmissions can be challenging.

Our Solution: WiSee presents a receiver design that can identify Doppler shifts at the resolution of a few Hertz from Wi-Fi signals. The basic idea underlying WiSee is to trans- form the received Wi-Fi signal into a narrowband pulse with the bandwidth of a few Hertz. The receiver then tracks the frequency of this narrowband pulse to detect the small Doppler shifts.

WiSee is designed for OFDM-based systems – OFDM is the modulation of choice for most modern wireless sys- tems including 802.11 a/g/n, WiMAX, and LTE. OFDM divides the used RF bandwidth into multiple sub-channels and modulates data in each sub-channel. For instance, Wi-Fi typically divides the 20 MHz channel into 64 sub-channels each with a bandwidth of 312.5 KHz. The time-domain OFDM symbol is generated at the transmitter by taking an FFT over a sequence of modulated bits transmitted in each OFDM sub-channel. Specifically, the transmitter takes blocks of N modulated bits (N = 64 in 802.11), and applies an N -point Inverse Fast Fourier Transform (IFFT),

x k =

N

X

n =1

X n e i 2πkn/N

where X n is the modulated bit sent in the n th OFDM sub-channel. Each block of x 1 , · · · x N forms a time-domain OFDM symbol that the receiver decodes by performing the FFT operation, i.e.,

X n =

N

X

k =1

x k e −i 2πkn /N (2)

To demonstrate how WiSee’s receiver works on these OFDM signals, we first consider the scenario where the transmitter repeatedly sends the same OFDM symbol. We then generalize our approach to arbitrary OFDM symbols, making the scheme applicable to existing 802.11 frames.

Case 1: Transmitter sends the same OFDM symbol.

In this case, instead of performing an FFT over each OFDM symbol, WiSee’s receiver performs a large FFT over M con- secutive OFDM symbols. As a consequence of this operation, the bandwidth of each OFDM sub-channel is reduced by a factor of M . To see this, say the receiver performs a 2N - point FFT over two consecutive, identical OFDM symbols.

The output of the FFT can be written as, 1 X n =

N

X

k =1

x k e −i 2πkn /2N +

2N

X

k =N +1

x k e −i 2πkn /2N

1 For simplicity, we ignore the noise term in the above equa- tion. However, as with standard OFDM decoding, these lin- ear equations hold even in the presence of noise.

0 0.1 0.2 0.3 0.4

-30 -20 -10 0 10 20 30

Amplitude

OFDM Sub-channel

0 0.2 0.4 0.6 0.8

-60 -40 -20 0 20 40 60

Amplitude

OFDM Sub-channel

Figure 2—Creating a narrowband signal using WiSee: The first subplot shows the output of an FFT taken over an OFDM symbol. The second subplot shows the out- put of an FFT taken over two identical OFDM symbols.

They show that taking a larger FFT over identical OFDM symbols, reduces each subchannel’s bandwidth.

Since the first N transmitted samples are identical to the last N samples, i.e., x k = x k +N , for k = 1 to N , we can re-write the above equation as,

X n =

N

X

k =1

x k e −i 2πkn /2N +

N

X

k =1

x k e −i 2π(k +N )n /2N

After simplification, we get:

X n =

N

X

k =1

x k e −i 2πkn /2N (1 + e −i πn )

Now, when n is an even number, (1+e −i πn ) = 2, but when n is an odd number, (1+e −i πn ) = 0. Thus, the above equation can be re-written as,

X 2l = 2

N

X

k =1

x k e −i 2πkl /N , X 2l+1 = 0

Thus, as shown in Fig. 2, the odd sub-channels are zero and the even sub-channels capture the output (Eq. 2) of an N - point FFT on a single OFDM symbol. Intuitively, this hap- pens because in each sub-channel, the same modulated in- formation is transmitted in both the OFDM symbols. Thus, the bandwidth used by each sub-channel effectively halves.

More generally, when the receiver performs an MN -point FFT over an OFDM symbol that is repeated M times, the bandwidth of each sub-channel is reduced by a factor of M . Thus, WiSee can create multiple narrowband signals cen- tered at each sub-channel by repeating an OFDM symbol and performing a large FFT operation.

Now, by performing a large FFT over an one-second du- ration, the WiSee receiver can create a one-Hertz wide nar- rowband signal. The WiSee receiver tracks this narrowband signal to capture the Doppler shift (see §3.2). Note that one can average the Doppler shifts observed across all the OFDM sub-channels to significantly reduce the noise in the Doppler measurements.

Case 2: Transmitter sends arbitrary OFDM sym- bols. Our description so far assumes that the transmitter repeatedly sends the same OFDM symbol. Typical 802.11 transmitters however send different data across symbols. We

312.5KHz per subcarrier

(26)

How to Identify Small Shift

even in Wideband Channels?

Idea : Transform the WiFi signals to narrowband pulses via large FFT!

many more point object moving at the same time. Thus, it creates Doppler signals with much larger energy than when the human uses only parts of her body.

Challenge: Human motion results in a very small Doppler shift that can be hard to detect from a typical wireless transmission (e.g., Wi-Fi, WiMax, LTE, etc.). For instance, consider a user moving her hand towards the receiver at 0.5 m/sec. From Eq. 1, this results in a Doppler shift of about 17 Hertz for a Wi-Fi signal transmitted at 5 GHz (θ = 0). Since the bandwidth of Wi-Fi’s transmissions is at least 20 MHz, the resulting Doppler shift is orders of magni- tude smaller than Wi-Fi’s bandwidth. Identifying such small Doppler shifts from these transmissions can be challenging.

Our Solution: WiSee presents a receiver design that can identify Doppler shifts at the resolution of a few Hertz from Wi-Fi signals. The basic idea underlying WiSee is to trans- form the received Wi-Fi signal into a narrowband pulse with the bandwidth of a few Hertz. The receiver then tracks the frequency of this narrowband pulse to detect the small Doppler shifts.

WiSee is designed for OFDM-based systems – OFDM is the modulation of choice for most modern wireless sys- tems including 802.11 a/g/n, WiMAX, and LTE. OFDM divides the used RF bandwidth into multiple sub-channels and modulates data in each sub-channel. For instance, Wi-Fi typically divides the 20 MHz channel into 64 sub-channels each with a bandwidth of 312.5 KHz. The time-domain OFDM symbol is generated at the transmitter by taking an FFT over a sequence of modulated bits transmitted in each OFDM sub-channel. Specifically, the transmitter takes blocks of N modulated bits (N = 64 in 802.11), and applies an N -point Inverse Fast Fourier Transform (IFFT),

xk =

N

X

n=1

Xnei2πkn/N

where Xn is the modulated bit sent in the nth OFDM sub-channel. Each block of x1, · · · xN forms a time-domain OFDM symbol that the receiver decodes by performing the FFT operation, i.e.,

Xn =

N

X

k=1

xke−i 2πkn /N (2)

To demonstrate how WiSee’s receiver works on these OFDM signals, we first consider the scenario where the transmitter repeatedly sends the same OFDM symbol. We then generalize our approach to arbitrary OFDM symbols, making the scheme applicable to existing 802.11 frames.

Case 1: Transmitter sends the same OFDM symbol.

In this case, instead of performing an FFT over each OFDM symbol, WiSee’s receiver performs a large FFT over M con- secutive OFDM symbols. As a consequence of this operation, the bandwidth of each OFDM sub-channel is reduced by a factor of M . To see this, say the receiver performs a 2N - point FFT over two consecutive, identical OFDM symbols.

The output of the FFT can be written as,1 Xn =

N

X

k=1

xke−i 2πkn /2N +

2N

X

k=N +1

xke−i 2πkn /2N

1For simplicity, we ignore the noise term in the above equa- tion. However, as with standard OFDM decoding, these lin- ear equations hold even in the presence of noise.

0 0.1 0.2 0.3 0.4

-30 -20 -10 0 10 20 30

Amplitude

OFDM Sub-channel

0 0.2 0.4 0.6 0.8

-60 -40 -20 0 20 40 60

Amplitude

OFDM Sub-channel

Figure 2—Creating a narrowband signal using WiSee: The first subplot shows the output of an FFT taken over an OFDM symbol. The second subplot shows the out- put of an FFT taken over two identical OFDM symbols.

They show that taking a larger FFT over identical OFDM symbols, reduces each subchannel’s bandwidth.

Since the first N transmitted samples are identical to the last N samples, i.e., xk = xk+N, for k = 1 to N , we can re-write the above equation as,

Xn =

N

X

k=1

xke−i 2πkn /2N +

N

X

k=1

xke−i 2π(k +N )n /2N

After simplification, we get:

Xn =

N

X

k=1

xke−i 2πkn /2N(1 + e−i πn)

Now, when n is an even number, (1+e−i πn) = 2, but when n is an odd number, (1+e−i πn) = 0. Thus, the above equation can be re-written as,

X2l = 2

N

X

k=1

xke−i 2πkl /N, X2l+1 = 0

Thus, as shown in Fig. 2, the odd sub-channels are zero and the even sub-channels capture the output (Eq. 2) of an N - point FFT on a single OFDM symbol. Intuitively, this hap- pens because in each sub-channel, the same modulated in- formation is transmitted in both the OFDM symbols. Thus, the bandwidth used by each sub-channel effectively halves.

More generally, when the receiver performs an MN -point FFT over an OFDM symbol that is repeated M times, the bandwidth of each sub-channel is reduced by a factor of M . Thus, WiSee can create multiple narrowband signals cen- tered at each sub-channel by repeating an OFDM symbol and performing a large FFT operation.

Now, by performing a large FFT over an one-second du- ration, the WiSee receiver can create a one-Hertz wide nar- rowband signal. The WiSee receiver tracks this narrowband signal to capture the Doppler shift (see §3.2). Note that one can average the Doppler shifts observed across all the OFDM sub-channels to significantly reduce the noise in the Doppler measurements.

Case 2: Transmitter sends arbitrary OFDM sym- bols. Our description so far assumes that the transmitter repeatedly sends the same OFDM symbol. Typical 802.11 transmitters however send different data across symbols. We many more point object moving at the same time. Thus, it creates Doppler signals with much larger energy than when the human uses only parts of her body.

Challenge: Human motion results in a very small Doppler shift that can be hard to detect from a typical wireless transmission (e.g., Wi-Fi, WiMax, LTE, etc.). For instance, consider a user moving her hand towards the receiver at 0.5 m/sec. From Eq. 1, this results in a Doppler shift of about 17 Hertz for a Wi-Fi signal transmitted at 5 GHz (θ = 0). Since the bandwidth of Wi-Fi’s transmissions is at least 20 MHz, the resulting Doppler shift is orders of magni- tude smaller than Wi-Fi’s bandwidth. Identifying such small Doppler shifts from these transmissions can be challenging.

Our Solution: WiSee presents a receiver design that can identify Doppler shifts at the resolution of a few Hertz from Wi-Fi signals. The basic idea underlying WiSee is to trans- form the received Wi-Fi signal into a narrowband pulse with the bandwidth of a few Hertz. The receiver then tracks the frequency of this narrowband pulse to detect the small Doppler shifts.

WiSee is designed for OFDM-based systems – OFDM is the modulation of choice for most modern wireless sys- tems including 802.11 a/g/n, WiMAX, and LTE. OFDM divides the used RF bandwidth into multiple sub-channels and modulates data in each sub-channel. For instance, Wi-Fi typically divides the 20 MHz channel into 64 sub-channels each with a bandwidth of 312.5 KHz. The time-domain OFDM symbol is generated at the transmitter by taking an FFT over a sequence of modulated bits transmitted in each OFDM sub-channel. Specifically, the transmitter takes blocks of N modulated bits (N = 64 in 802.11), and applies an N -point Inverse Fast Fourier Transform (IFFT),

xk =

N

X

n=1

Xnei2πkn/N

where Xn is the modulated bit sent in the nth OFDM sub-channel. Each block of x1, · · · xN forms a time-domain OFDM symbol that the receiver decodes by performing the FFT operation, i.e.,

Xn =

N

X

k=1

xke−i 2πkn /N (2)

To demonstrate how WiSee’s receiver works on these OFDM signals, we first consider the scenario where the transmitter repeatedly sends the same OFDM symbol. We then generalize our approach to arbitrary OFDM symbols, making the scheme applicable to existing 802.11 frames.

Case 1: Transmitter sends the same OFDM symbol.

In this case, instead of performing an FFT over each OFDM symbol, WiSee’s receiver performs a large FFT over M con- secutive OFDM symbols. As a consequence of this operation, the bandwidth of each OFDM sub-channel is reduced by a factor of M . To see this, say the receiver performs a 2N - point FFT over two consecutive, identical OFDM symbols.

The output of the FFT can be written as,1 Xn =

N

X

k=1

xke−i 2πkn /2N +

2N

X

k=N +1

xke−i 2πkn /2N

1For simplicity, we ignore the noise term in the above equa- tion. However, as with standard OFDM decoding, these lin- ear equations hold even in the presence of noise.

0 0.1 0.2 0.3 0.4

-30 -20 -10 0 10 20 30

Amplitude

OFDM Sub-channel

0 0.2 0.4 0.6 0.8

-60 -40 -20 0 20 40 60

Amplitude

OFDM Sub-channel

Figure 2—Creating a narrowband signal using WiSee: The first subplot shows the output of an FFT taken over an OFDM symbol. The second subplot shows the out- put of an FFT taken over two identical OFDM symbols.

They show that taking a larger FFT over identical OFDM symbols, reduces each subchannel’s bandwidth.

Since the first N transmitted samples are identical to the last N samples, i.e., xk = xk+N, for k = 1 to N , we can re-write the above equation as,

Xn =

N

X

k=1

xke−i 2πkn /2N +

N

X

k=1

xke−i 2π(k +N )n /2N

After simplification, we get:

Xn =

N

X

k=1

xke−i 2πkn /2N(1 + e−i πn)

Now, when n is an even number, (1+e−i πn) = 2, but when n is an odd number, (1+e−i πn) = 0. Thus, the above equation can be re-written as,

X2l = 2

N

X

k=1

xke−i 2πkl /N, X2l+1 = 0

Thus, as shown in Fig. 2, the odd sub-channels are zero and the even sub-channels capture the output (Eq. 2) of an N - point FFT on a single OFDM symbol. Intuitively, this hap- pens because in each sub-channel, the same modulated in- formation is transmitted in both the OFDM symbols. Thus, the bandwidth used by each sub-channel effectively halves.

More generally, when the receiver performs an MN -point FFT over an OFDM symbol that is repeated M times, the bandwidth of each sub-channel is reduced by a factor of M . Thus, WiSee can create multiple narrowband signals cen- tered at each sub-channel by repeating an OFDM symbol and performing a large FFT operation.

Now, by performing a large FFT over an one-second du- ration, the WiSee receiver can create a one-Hertz wide nar- rowband signal. The WiSee receiver tracks this narrowband signal to capture the Doppler shift (see §3.2). Note that one can average the Doppler shifts observed across all the OFDM sub-channels to significantly reduce the noise in the Doppler measurements.

Case 2: Transmitter sends arbitrary OFDM sym- bols. Our description so far assumes that the transmitter repeatedly sends the same OFDM symbol. Typical 802.11 transmitters however send different data across symbols. We

FFT over one symbol FFT over two identical symbol

(27)

Capturing movement via Large FFT

FFT over 10,000 symbols → 31.25Hz per subcarrier

−3 −2 −1 0 1 2 3

x 104 0

0.2 0.4 0.6 0.8

OFDM Sub−channels

Amplitude

−3 −2 −1 0 1 2 3

x 104 0

0.2 0.4 0.6 0.8

OFDM Sub−channels

Amplitude

without movement

with movement

(28)

Capturing over Time

time (second)

frequency (Hz)

1.25 2.5 3.75 5 6.25 7.5 8.75

30 20 10 0

−10

−20

−30 8

16 24 32 40dB

Figure 5—Frequency-time Doppler profile of an ex- ample gesture. The user moves her hand towards the re- ceiver.

to be a specific kind of discontinuity between the OFDM symbols. Thus, we can perform interpolation between the OFDM symbols as described earlier. We note, however, that since all the CPs have a fixed length, such an interpolation is equivalent to resampling the OFDM symbols at a constant rate given by

Symbol length+CP length

Symbol length

, where Symbol length and CP length denote the length of the OFDM symbol and CP respectively. Since such resampling of the symbols does not change the doppler pattern, in practice we simply skip the CPs to reduce the computation.

3.2 Mapping Doppler Shifts to Gestures

So far we described how to transform the wideband 802.11 transmissions into a narrowband signal at the receiver. In this section, we show how to extract the Doppler informa- tion and map it to the gestures. Specifically, we describe the following three steps: (1) Doppler extraction which computes the Doppler shifts from the narrowband signals, (2) Segmen- tation which identifies a set of segments that correspond to a gesture, and (3) Classification which determines the most likely gesture amongst a set of gestures. We describe how WiSee performs each of these steps. We focus on the sin- gle user case; in §3.3, we extend our design to work in the presence of other users.

(1) Doppler Extraction: WiSee extracts the Doppler in- formation by computing the frequency-time Doppler profile of the narrowband signal. To do this, the receiver computes a sequence of FFTs taken over time. Specifically, it computes an FFT over samples in the first half-a-second interval. Such an FFT give a Doppler resolution of 2 Hertz. The receiver then moves forward by a 5 ms interval and computes an- other FFT over the next overlapping half-a-second interval.

It repeats this process to get a frequency-time profile.

Fig. 5 plots the frequency-time Doppler profile (in dB) of a user moving her hand towards the receiver. The plot shows that, at the beginning of the gesture most of the energy is concentrated in the DC (zero) frequency. This corresponds to the signal energy between the transmitter and the receiver, on paths that do not include the human. However, as the user starts moving her hand towards the receiver, we first see increasing positive Doppler frequencies (corresponding to hand acceleration) and then decreasing positive Doppler frequencies (corresponding to hand deceleration).

We note that the WiSee receiver is only interested in the Doppler shifts produced by human gestures. Since the speeds at which a human can typically perform gestures are between 0.25 m/sec and 4 m/sec [12], the Doppler shift of interest at 5 GHz is between 8 Hz and 134 Hz. Thus, the WiSee receiver reduces its computational complexity by analyzing the FFT output corresponding to only these frequencies.

(2) Segmentation: To do this, WiSee leverages the struc- ture of the Doppler profiles, shown in Fig. 6. These corre-

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

Figure 6—Frequency-time Doppler profiles of the gestures in Fig. 1. WiSee segments the profiles into sequences of positive and negative Doppler shifts, which uniquely identify each gesture.

spond to the gestures in Fig 1. The plots show that the profiles are a combination of positive and negative Doppler shifts. Further, each gesture comprises of a set of segments that have positive and negative Doppler shifts. For example, the profile in Fig. 6(a) has just one segment with positive Doppler shift. However, Fig. 6(b) has two segments each of which has a positive and a negative Doppler shift. Further, within each segment, the Doppler energy first increases and then decreases (which correspond to acceleration and decel- eration of human body parts).

A WiSee receiver leverages these properties to first find segments and then cluster segments into a gesture. Our pro- cess of finding segments is intuitively similar to packet detec- tion in wireless communication systems. In communication, to detect the beginning of a packet, the receiver computes the average received energy over a small duration. If the ratio between this energy and noise level is greater than a thresh- old, then the receiver detects the beginning of a packet. Sim- ilarly, if this ratio falls below a threshold, the receiver detects the end of the packet. Likewise, in our system, the energy in each segment first increases and then decreases. So the WiSee receiver computes the average energy in the positive and negative Doppler frequencies (other than the DC and the four frequency bins around it). If the ratio between this average energy and the noise level is greater than 3 dB, the receiver detects the beginning of a segment. When this ratio falls below 3 dB, the receiver detects the end of the segment.

3

To cluster segments into a single gesture, WiSee’s receiver uses a simple algorithm: if two segments are separated by less than one second, we cluster them into a single gesture.

(3) Gestures Classification: As described earlier, the Doppler profiles in Fig. 6 can be considered as a sequence of positive and negative Doppler shifts. Further, from the plots, we see that the patterns are unique and different across the nine gestures. Thus, the receiver can classify gestures by matching the pattern of positive and negative Doppler shifts. Specifically, there are three types of segments: seg- ments with only positive Doppler shifts, segments with only

3

The noise level is calibrated at the receiver by computing the energy in the non-DC frequencies, in the absence of ges- tures.

Frequency-time Doppler profile of an example gesture (push)

(29)

Detection by Classification

time (second)

frequency (Hz)

1.25 2.5 3.75 5 6.25 7.5 8.75

30 20 10 0

−10

−20

−30

8 16 24 32 40dB

Figure 5—Frequency-time Doppler profile of an ex- ample gesture. The user moves her hand towards the re- ceiver.

to be a specific kind of discontinuity between the OFDM symbols. Thus, we can perform interpolation between the OFDM symbols as described earlier. We note, however, that since all the CPs have a fixed length, such an interpolation is equivalent to resampling the OFDM symbols at a constant rate given by

Symbol length+CP length

Symbol length

, where Symbol length and CP length denote the length of the OFDM symbol and CP respectively. Since such resampling of the symbols does not change the doppler pattern, in practice we simply skip the CPs to reduce the computation.

3.2 Mapping Doppler Shifts to Gestures

So far we described how to transform the wideband 802.11 transmissions into a narrowband signal at the receiver. In this section, we show how to extract the Doppler informa- tion and map it to the gestures. Specifically, we describe the following three steps: (1) Doppler extraction which computes the Doppler shifts from the narrowband signals, (2) Segmen- tation which identifies a set of segments that correspond to a gesture, and (3) Classification which determines the most likely gesture amongst a set of gestures. We describe how WiSee performs each of these steps. We focus on the sin- gle user case; in §3.3, we extend our design to work in the presence of other users.

(1) Doppler Extraction: WiSee extracts the Doppler in- formation by computing the frequency-time Doppler profile of the narrowband signal. To do this, the receiver computes a sequence of FFTs taken over time. Specifically, it computes an FFT over samples in the first half-a-second interval. Such an FFT give a Doppler resolution of 2 Hertz. The receiver then moves forward by a 5 ms interval and computes an- other FFT over the next overlapping half-a-second interval.

It repeats this process to get a frequency-time profile.

Fig. 5 plots the frequency-time Doppler profile (in dB) of a user moving her hand towards the receiver. The plot shows that, at the beginning of the gesture most of the energy is concentrated in the DC (zero) frequency. This corresponds to the signal energy between the transmitter and the receiver, on paths that do not include the human. However, as the user starts moving her hand towards the receiver, we first see increasing positive Doppler frequencies (corresponding to hand acceleration) and then decreasing positive Doppler frequencies (corresponding to hand deceleration).

We note that the WiSee receiver is only interested in the Doppler shifts produced by human gestures. Since the speeds at which a human can typically perform gestures are between 0.25 m/sec and 4 m/sec [12], the Doppler shift of interest at 5 GHz is between 8 Hz and 134 Hz. Thus, the WiSee receiver reduces its computational complexity by analyzing the FFT output corresponding to only these frequencies.

(2) Segmentation: To do this, WiSee leverages the struc- ture of the Doppler profiles, shown in Fig. 6. These corre-

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

Figure 6—Frequency-time Doppler profiles of the gestures in Fig. 1. WiSee segments the profiles into sequences of positive and negative Doppler shifts, which uniquely identify each gesture.

spond to the gestures in Fig 1. The plots show that the profiles are a combination of positive and negative Doppler shifts. Further, each gesture comprises of a set of segments that have positive and negative Doppler shifts. For example, the profile in Fig. 6(a) has just one segment with positive Doppler shift. However, Fig. 6(b) has two segments each of which has a positive and a negative Doppler shift. Further, within each segment, the Doppler energy first increases and then decreases (which correspond to acceleration and decel- eration of human body parts).

A WiSee receiver leverages these properties to first find segments and then cluster segments into a gesture. Our pro- cess of finding segments is intuitively similar to packet detec- tion in wireless communication systems. In communication, to detect the beginning of a packet, the receiver computes the average received energy over a small duration. If the ratio between this energy and noise level is greater than a thresh- old, then the receiver detects the beginning of a packet. Sim- ilarly, if this ratio falls below a threshold, the receiver detects the end of the packet. Likewise, in our system, the energy in each segment first increases and then decreases. So the WiSee receiver computes the average energy in the positive and negative Doppler frequencies (other than the DC and the four frequency bins around it). If the ratio between this average energy and the noise level is greater than 3 dB, the receiver detects the beginning of a segment. When this ratio falls below 3 dB, the receiver detects the end of the segment.

3

To cluster segments into a single gesture, WiSee’s receiver uses a simple algorithm: if two segments are separated by less than one second, we cluster them into a single gesture.

(3) Gestures Classification: As described earlier, the Doppler profiles in Fig. 6 can be considered as a sequence of positive and negative Doppler shifts. Further, from the plots, we see that the patterns are unique and different across the nine gestures. Thus, the receiver can classify gestures by matching the pattern of positive and negative Doppler shifts. Specifically, there are three types of segments: seg- ments with only positive Doppler shifts, segments with only

3

The noise level is calibrated at the receiver by computing the energy in the non-DC frequencies, in the absence of ges- tures.

Different gestures correspond to 


various frequency-time Doppler profiles

(30)

Other Applications

• Sleep Apnea Detection!

• Hand Writing / Drawing!

• WiFi Imaging!

• Keystroke!

• …

參考文獻

相關文件

了⼀一個方案,用以尋找滿足 Calabi 方程的空 間,這些空間現在通稱為 Calabi-Yau 空間。.

Wang, Solving pseudomonotone variational inequalities and pseudocon- vex optimization problems using the projection neural network, IEEE Transactions on Neural Networks 17

Hope theory: A member of the positive psychology family. Lopez (Eds.), Handbook of positive

volume suppressed mass: (TeV) 2 /M P ∼ 10 −4 eV → mm range can be experimentally tested for any number of extra dimensions - Light U(1) gauge bosons: no derivative couplings. =>

Define instead the imaginary.. potential, magnetic field, lattice…) Dirac-BdG Hamiltonian:. with small, and matrix

• Formation of massive primordial stars as origin of objects in the early universe. • Supernova explosions might be visible to the most

正向成就 (positive accomplishment) 正向目標 (意義) (positive purpose) 正向健康 (positive health).. Flourish: A visionary new understanding of happiness

2-1 註冊為會員後您便有了個別的”my iF”帳戶。完成註冊後請點選左方 Register entry (直接登入 my iF 則直接進入下方畫面),即可選擇目前開放可供參賽的獎項,找到iF STUDENT