collaborative activity

8/2/2019 Collaborative Activity

1/17

-1- Georgia State Universityensorweb Research Laboratory

Multi-Target Tracking in Smart Environments

using Collaborative Sensing

Debraj De

Presentation 11/22/2011


2/17


Presentation Outline

Background of research problem

Related works

Proposed solution

System evaluation plan


3/17



Multi-user tracking in smart environments only with

binary motion sensor network

Not sufficient solution

Error in path disambiguation if a number of usersoverlap/crossover at the same time


4/17



Need collaborative sensing framework

Binary motion sensing + user worn accelerometer

Research problem: how to disambiguate path of eachuser

Challenge: only a part of the users use accelerometer


5/17


Related works

Toward Cooperative Localization of Wearable Sensors

using Accelerometer and Camera, Infocom 2010

PEM-ID: Identifying People by Gait-Matching usingCameras and Wearable Accelerometers, ICDSC 2009

Adaptive Calibration for Fusion-based Wireless SensorNetworks, Infocom 2010

Energy-efficient Trajectory Tracking for Mobile Devices,MobiSys 2011


6/17


Related works

Problem with existing works: most of them use camera

and accelerometer collaboration

No work found (to my knowledge till now) using tworelatively low-end sensors

Also assumes each user wears at least oneaccelerometer


7/17-7- Georgia State Universityensorweb Research Laboratory

Proposed solution: system model

Path disambiguated multi-user tracking in smart

environments using collaborative sensing betweenaccelerometer and static binary motion sensor network

Only a subset of users are wearing accelerometers

(more practical scenario participatory sensing)

We allow some users carrying phone or some devicesthat can communicate with motion sensor nodes (thus

we allow heterogeneity)



Proposed solution: problem formulation

System model: N motion sensor nodes

K users

K users with accelerometers

Example:



Proposed solution: problem formulation

Motion signature:

Accelerometer signature:



Proposed solution: problem formulation Distance function between the signatures will denote

how correlated they are:

Problem formulation:

[P1] Distance Function calculation

[P2] Displacement computation from accelerometer data

[P3] Assignment problem: combinatorial optimization



Proposed solution: [P1] distance function

Pearsons correlation coefficient is popular metric for

signal similarity

We use this as our distance function D( )



Proposed solution: [P2] displacement calculation

The displacement calculation using accelerometer should

be independent of: Orientation angle

User gait size

User walking speed



Proposed solution: [P3] Assignment Problem

Classical assignment problem: Consider the situation of assigning n jobs to n machines. When a job i (=1,2,....,n) is assigned to machine j (=1,2, .....n) that incurs

a cost Cij.

The objective is to assign the jobs to machines at the least possible totalcost.

We will propose an iterative approach of matching tracki with track i iteratively, such that activity transitiongraph is satisfied.



Proposed solution: including communication

Problem: Shift, jitter and out of range motion of users

with accelerometers

Some users accelerometers will be able to communicatewith motion sensor nodes

This requires change in the accelerometer signaturedefinition



Evaluation plan

Experiment setup:

6 users walking in 14th floor in random paths

Collect ground truth

3 users wearing accelerometers

Collect motion data and acc data in valve

Parameters:

Tracking error vs time

Tracking error CDF Tracking error vs #of users with accelerometer



Evaluation plan

Comparison study

User tracking evaluation in regular working day on 14thfloor

Show live GUI


17/1717

Questions

Email: [email protected]

collaborative activity

Documents