Gary M. Weiss & Jeffrey W. Lockhart
Fordham University
{gweiss,lockhart}@cis.fordham.edu

 Biometrics concerns unique identification based on physical or behavioral traits
 Hard biometrics relies on uniquely identifying traits
▪ Fingerprints, DNA, iris, etc.
 Soft biometric traits are not distinctive enough for unique identification, but may help
▪ Physical traits: Sex, age, height, weight, etc.
▪ Behavioral traits: gait, clothes, travel patterns, etc.
8/21/2011 Gary M. Weiss SensorKDD 2011 2

 In earlier work 1 we showed that for 36 users we were able to identify the correct user using only accelerometer data:
 With a single 10 second walking sample: 84% - 91%
 With a 5-10 minute walking sample: 100%
 So if we can identify a user based on their movements, maybe we can identify user traits
1 Jennifer R. Kwapisz, Gary M. Weiss and Samuel A. Moore. Cell Phone-Based Biometric Identification, Proceedings of the
IEEE Fourth International Conference on Biometrics: Theory, Applications and Systems (BTAS-10) , Washington DC.
8/21/2011 Gary M. Weiss SensorKDD 2011 3

 To help identify a person (soft biometrics)
 But do we have better uses for these “soft” traits than for identification?
 As data miners, of course we do!
 We want to know everything we possibly can about a person. Somehow we will exploit this.
8/21/2011 Gary M. Weiss SensorKDD 2011 4

 Normally think about traits as being:
 Unchanging: race, skin color, eye color, etc.
 Slow changing: Height, weight, etc.
 But want to know everything about a person:
 What they wear, how they feel, if they are tired, etc.
 Our goal is to predict these too
 We have not seen this goal stated in context of mobile sensor data mining.
8/21/2011 Gary M. Weiss SensorKDD 2011 5



Very little explicit work on this topic
 Some work related to biometrics but incidental
▪ Work on gait recognition mentions factors that influence recognition, like weight of footwear & sex
Other communities work in related areas
 Ergonomics & kinesiology study factors that impact gait
▪ Texture of footwear, type of shoe, sex, age, heel height
8/21/2011 Gary M. Weiss SensorKDD 2011 6

 Data collected from ~70 people
 Accelerometer data while walking
 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)
8/21/2011 Gary M. Weiss SensorKDD 2011 7

Accuracy
71.2%
Male
Female
Male Female
31
12
7
16
Accuracy
83.3%
Short
Tall
Short Tall
15
2
5
20
Accuracy
78.9%
Light
Heavy
Light
13
2
Heavy
Results for IB3 classifier. For height and weight middle categories removed.
8/21/2011 Gary M. Weiss SensorKDD 2011
7
17
8

8/21/2011 Gary M. Weiss SensorKDD 2011 9





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 …
8/21/2011 Gary M. Weiss SensorKDD 2011 10




Will collect data from hundreds of users
 Getting a diverse sample a bit difficult (on campus)
Try to construct more useful features
Evaluate the ability to predict the dozens of user traits that we track
 Have begun to track shoe type and heel size
8/21/2011 Gary M. Weiss SensorKDD 2011 11