gary m. weiss fordham university [email protected]

Smart Phone-Based Sensor Mining

Gary M. WeissFordham University

[email protected]

Gary M. Weiss Einstein 2

Background and Motivation Smart phones are ubiquitous

As of 4th quarter 2010 outpaced PC sales We carry them everywhere at almost all times

Smart phones are powerful Increasing processing power and storage space Filled with sensors

Smart phones include the following sensors:▪ Tri-Axial Accelerometer▪ Location sensor (GPS, cell tower, WiFi)▪ Audio sensor (microphone), Image sensor (camera)▪ Proximity, light, temperature, magnetic compass

5/17/2012


Data Mining and Sensor Mining Data mining: application of computational

methods to extract knowledge from data Most data mining involves inferring predictive

models, often for classification Sensor mining: application of computational

methods to extract knowledge from sensor data

Supervised machine learning Obtain labeled time-series training data Create examples described by generated features Build model to predict example’s label

5/17/2012


The WISDM Project

Three years ago started what is now WISDM Began with focus on activity recognition

▪ Determine what a user is doing based on accelerometer Moved to an Android-based smartphone platform Expanded to other applications

▪ Biometric identification▪ Identifying user characteristics (soft biometrics)▪ Mining GPS data (project starting with Bronx Zoo)

Current focus on Actitracker▪ Track user activities and present info to user via the

web as a health app (NSF “Health and Well-Being Grant)

5/17/2012


The WISDM Platform

Based on Android Smartphones but could be extended to other mobile devices

Client/Server architecture Smartphones are the client (they run our app) We have a dedicated server Right now raw data is sent to the server and

processing occurs there Data can be streamed or sent on demand In future more responsibility moved to the phone

5/17/2012


WISDM Platform Continued Web Interface

Users can access their data via a web interface▪ Accessible from smartphone or full-screen

computer

Security Secure logins and data encrypted

Resource Issues: Power Power is an issue if collect GPS data and

maybe if we collect data 24x7, but not for periodic data collection

5/17/2012


Smart Phone Accelerometer

Measures acceleration along 3 spatial axes Detects/measures gravity (orientation matters) Measurement range typically -2g to +2g

Okay for most activities but falling yields higher values

Range & sensitivity may be adjustable Sampling rates ~20-50 Hz

Study found 20Hz required for activity recognition WISDM project found could not reliably sample

beyond 20Hz (50ms) and this may impact activity recognition

5/17/2012


Existing WISDM Applications

Activity Recognition Identify the activity a user is performing

(walking, jogging, sitting, etc.) Biometric Identification

Identify a user based on prior accelerometer data collected from that user

Trait Identification Identify characteristics about a user

based (height, weight, age)5/17/2012


Why is Activity Recognition Useful?

Context-sensitive applications Handle phone calls differently depending on

context Play music to suit your activity New & innovative apps to make phones smarter

Tracking & Health applications Track overall activity levels & generate fitness

profiles Care of elderly

▪ Detect dangerous situations like (falling)▪ Warn if some with Alzheimer’s wanders outside of area

5/17/2012


Accelerometer Data for Six Activites

Accelerometer data from Android phone Walking Jogging Climbing Stairs Lying Down Sitting Standing

5/17/2012


Accelerometer Data for “Walking”

5/17/2012


Accelerometer Data for “Jogging”

5/17/2012


Accelerometer Data for “Up Stairs”

5/17/2012


Accelerometer Data for “Lying Down”

5/17/2012


Accelerometer Data for “Sitting”

5/17/2012

Z axis


Accelerometer Data for “Standing”

5/17/2012


WISDM Activity Recognition

Six activities: walking, jogging, stairs, sitting, standing, lying down

Labeled data collected from over 50 users

Data transformed via 10-second windows Accelerometer data sampled (x,y,z)

every 50ms Features (per axis):

▪ average, SD, ave diff from mean, ave resultant accel, binned distribution, time between peaks

5/17/2012


WISDM Activity Recognition

The 43 features used to build a classifier WEKA data mining suite used, multiple

techniques Personal, universal, hybrid models built

Architecture (for now) uses “dumb” client

Basis of soon to be released actitracker service Provides web based view of activities

over time5/17/2012


WISDM Results

WISDM Results are shown for various things Personal, universal, and hybrid models Most results aggregated over all users

but a few per user to show how performance varies by user

Results for 6 activities (ones shown in the plots)

5/17/2012


WISDM Universal Model- IB3 Matrix

5/17/2012

72.4%Accuracy

Predicted Class

Walking

Jogging

Stairs

Sitting

Standing

LyingDown

Actual Class

Walking 2209 46 789 2 4 0

Jogging 45 1656 148 1 0 0

Stairs 412 54 869 3 1 0

Sitting 10 0 47 553 30 241

Standing 8 0 57 6 448 3

Lying Down 5 1 7 301 13 131


WISDM Personal Model- IB3 Matrix

5/17/2012

98.4%accuracy

Predicted Class

Walking Jogging Stairs

Sitting

Standing

LyingDown

Actual Class

Walking 3033 1 24 0 0 0

Jogging 4 1788 4 0 0 0

Stairs 42 4 1292 1 0 0

Sitting 0 0 4 870 2 6

Standing 5 0 11 1 509 0

Lying Down 4 0 8 7 0 442


WISDM Accuracy Results

5/17/2012

% of Records Correctly ClassifiedPersonal Universal Stra

w ManIB3 J48 NN IB3 J48 NN

Walking 99.2 97.5 99.1 72.4 77.3 60.6 37.7

Jogging 99.6 98.9 99.9 89.5 89.7 89.9 22.8

Stairs 96.5 91.7 98.0 64.9 56.7 67.6 16.5

Sitting 98.6 97.6 97.7 62.8 78.0 67.6 10.9

Standing 96.8 96.4 97.3 85.8 92.0 93.6 6.4

Lying Down 95.9 95.0 96.9 28.6 26.2 60.7 5.7

Overall 98.4 96.6 98.7 72.4 74.9 71.2 37.7


Biometric Identification

5/17/2012


Biometrics

Biometrics concerns unique identification based on physical or behavioral traits Hard biometrics involves traits that are

sufficient to uniquely identify a person▪ Fingerprints, DNA, iris, etc.

Soft biometric traits are not sufficiently distinctive, but may help▪ Physical traits: Sex, age, height, weight, etc.▪ Behavioral traits: gait, clothes, travel

patterns, etc.5/17/2012


Gait-Based Biometrics

Numerous accelerometer-based systems that use dedicated and/or multiple sensors See related work section of Cell Phone-Based

Biometric Identification for details Possible uses:

▪ Phone security (e.g., to automatically unlock phone)▪ Automatic device customization▪ To better track people for shared devices▪ Perhaps for secondary level of physical security

5/17/2012


WISDM Biometric System Same setup as WISDM activity recognition

Same data collection, feature extraction, WEKA, … Used for identification and authentication

Identification: predicting identity from pool of users Authentication is binary class prediction problem

Evaluate single and mixed activities Evaluate using 10 sec. and several min. of test data

▪ Longer sample classify with “Most Frequent Prediction” Results based on 36 users

But hold up on preliminary experiments with 200 users

5/17/2012


WISDM Biometric Prediction Results

Aggregate

Walk Jog Up Dow

n

Aggregate

(Oracle)

J48 72.2 84.0 83.0

65.8

61.0 76.1

Neural Net

69.5 90.9 92.2

63.3

54.5 78.6

Straw Man

4.3 4.2 5.0 6.5 4.7 4.3

5/17/2012

Aggregate

Walk

Jog Up Down

Aggregate

(Oracle)

J48 36/36 36/36

31/32

31/31 28/31 36/36

Neural Net

36/36 36/36

32/32

28.5/31

25/31 36/36

Based on 10 second test samples

Based on most frequent prediction for 5-10 minutes of data


WISDM Biometric Authentication Results

Authentication results: Positive authentication of a user

▪ 10 second sample: ~85%▪ Most frequent class over 5-10 min: 100%

Negative Authentication of a user (an imposter)▪ 10 second sample: ~96%▪ Most frequent class over 5-10 min: 100%

5/17/2012


Biometric Identification Summary

Can do remarkably well with short amounts of accelerometer data (10s – 2 min)

Since we can distinguish between ways different people walk may be able to distinguish between different gaits

5/17/2012


Trait Identification

5/17/2012


WISDM Trait Identification Data collected from ~70 people (now over

200) Accelerometer and survey data Survey data includes anything we could think

of that might somehow be predictable▪ Sex, height, weight, age, race, handedness, disability▪ Type of area grew up in {rural, suburban, urban}▪ Shoe size, footwear type, size of heels, type of

clothing▪ # hours academic work , # hours exercise

Too few subjects investigate all factors▪ Many were not predictable (maybe with more data)

5/17/2012


WISDM Trait Identification Results

Accuracy

71.2%

Male

Female

Male 31 7

Female 12 16

5/17/2012

Accuracy

83.3%

Short

Tall

Short 15 5

Tall 2 20

Accuracy

78.9%

Light Heavy

Light 13 7

Heavy 2 17Results for IB3 classifier. For height and weight middle categories removed.


Trait Identification Summary

A wide open area for data mining research A marketers dream

Clear privacy issues Room for creativity & insight for

finding traits Probably many interesting

commercial and research applications Imagine diagnosing back problems via

your mobile phone via gait analysis … 5/17/2012


Connections to Your Work

Can collect accelerometer data from patients On demand or in the background Data transmitted wirelessly or stored on

the phone for periodic download

Can extend study beyond gait Can monitor overall activity levels Can monitor daily routine

5/17/2012


Connections to Your Work cont. Facilitate quantitative analysis of gait

“Fourth, although experienced clinicians assessed gait, quantitative analysis of gait might be more reliable” (Verghese et al. 2002)

Accelerometer data can provide basis for gait classification

Can use data mining to learn a classifier for gait▪ Just need carefully selected training data▪ Yields consistent measure

5/17/2012


Connections to Your Work cont.

Can look at other neurological diseases besides non-Alzheimer’s dementia

Can try to track progression of Alzheimer’s

Note can monitor daily routine, travel, etc.

Smartphone can also administer surveys, record video, provide voice prompts, etc.

Besides diagnosis, can assist people suffering from these diseases

5/17/2012


My Contact Information

Gary Weiss Fordham University, Bronx NY 10458 [email protected] http://storm.cis.fordham.edu/~gweiss/

WISDM Information http://www.cis.fordham.edu/wisdm/

▪ WISDM papers available: click “About” then “Publications”

By end of summer Actitracker will allow you to track your activities via our Android app (actitracker.com)

5/17/2012


WISDM Members

WISDM research group Current Active Members

▪ Linna AI*, Shaun Gallagher*, Andrew Grosner*, Margo Flynn, Jeff Lockhart*, Paul McHugh*, Tony Pulickal*, Greg Rivas*, Isaac Ronan*, Priscilla Twum, Bethany Wolff

* Working full-time on the project at Fordham over the summer

5/17/2012


References

Available from: http://www.cis.fordham.edu/wisdm/publications

Kwapisz, J.R., Weiss, G.M., and Moore, S.A. 2010. Activity recognition using cell phone

accelerometers, Proceedings of the Fourth International Workshop on Knowledge Discovery from

Sensor Data, 10-18.

Kwapisz, J.R., Weiss, G.M., and Moore, S.A. 2010. Cell phone-based biometric identification,

Proceedings of the IEEE Fourth International Conference on Biometrics: Theory, Applications and

Systems.

Lockhart, J.W., Weiss, G.M., Xue, J.C., Gallagher, S.T., Grosner, A.B., and Pulickal, T.T. 2011. Design

considerations for the WISDM smart phone-based sensor mining architecture, In Proceedings of

the Fifth International Workshop on Knowledge Discovery from Sensor Data, San Diego, CA.

Weiss, G.M., and Lockhart, J.W. 2011. Identifying user traits by mining smart phone accelerometer

data, Proceedings of the 5th International Workshop on Knowledge Discovery from Sensor Data.

Weiss, G.M., and Jeffrey W. Lockhart (2012). The Impact of Personalization on Smartphone-Based

Activity Recognition, Proceedings of the AAAI-12 Workshop on Activity Context Representation:

Techniques and Languages, Toronto, CA. 5/17/2012

gary m. weiss fordham university [email protected]

Documents

data mining

raw data

data 24x7

weiss einstein2 slide

weiss einstein4 slide

weiss einstein8 slide

weiss einstein6 slide

weiss einstein5 slide