HUMANWARE - New Related Work - Activity and Location Recognition Using Wearable Sensors - Memories for Life - SharMe – Nokia Research Center La The Vinh.

HUMANWARE - New Related Work- Activity and Location Recognition Using Wearable Sensors- Memories for Life- SharMe – Nokia Research Center

La The Vinh

[email protected]

Activity and Location Recognition Using Wearable Sensors

Proposed by Seon-Woo Lee and Kenji Mase Using acceleration and angular velocity data Inexpensive, wearable sensors Determine user’s location base on relative

distance from a given place Integrate incremental motion information over

time

Activity and Location Recognition Using Wearable Sensors

The systemLinux-based PDA2xSensor module

MicrocontrollerBiaxial accelerometerDigital compass sensorAngular velocity sensor

RS232 communication

Activity and Location Recognition Using Wearable Sensors

Activity and Location Recognition Using Wearable Sensors

Unit motionsSittingStandingWalking - Steps

Walking on a level groundGoing upGoing down

Activity and Location Recognition Using Wearable Sensors

Training phaseRecord unit motions and heading

measurements while user walks from one location to another

Ax(t),Az(t), θ(t) – Leg ModuleNorth, East, South, West – Waist Module

Build the transition table

Activity and Location Recognition Using Wearable Sensors

Recognition phaseDetect walking behaviorConstruct the vectorMatch to the transition tableUpdate user position

Activity and Location Recognition Using Wearable Sensors

Motion Recognitionσx(t)

σy(t)

σθ(t)

Δθ1

Δθ2

Δθ3

Activity and Location Recognition Using Wearable Sensors

Motion RecognitionSitting

Standing

Activity and Location Recognition Using Wearable Sensors

Steps CountingUsing peaks of az(t) – the blue

Activity and Location Recognition Using Wearable Sensors

Moving position Cx(k+1) = Cx(k) + Slcos(2πAh)

Cy(k+1) = Cy(k) + Slsin(2πAh)

Cz(k+1) = Cz(k) + Ss

Sl – Normalized length of a step on a level ground

Ss – Normalized length of an up/down step

Activity and Location Recognition Using Wearable Sensors

Transition vectors from location i to jTi,j = {dx

ij,dyij,dz

ij}

Test condition |Cx – dij

x| < thresholdx

|Cy – dijy| < thresholdy

|Cz – dijz| < thresholdz

Activity and Location Recognition Using Wearable Sensors

Activity and Location Recognition Using Wearable Sensors

Activity and Location Recognition Using Wearable Sensors

ProblemsAccumulated errorLimited in the case of long pathProne to drift