State of the Art - Smartphone-Based Human Activity Recognition - page 44

Table 3.4 Summary of existing online HAR systems
HAR System

P ∗

Sensor Activities Machine learning Computing Acc. (%)

No. Type Location No. Type algorithm device

Karantonis et al.
( 2006 )

1 3D-acc Wa i s t 10 Locomotion, falls Threshold-based Microcontroller,

PC

90.8

Sama et al. ( 2012 ) 1 3D-acc Wa i s t 3 PD motor symptoms 20 SVM Microcontroller 94

Brezmes et al.
( 2009 )

1 3D-acc Pockets 6 Locomotion, falls N/A k-NN Smartphone 80

Bruno et al. ( 2013 ) 1 3D-acc Wrist 8 ADL, PTs,

ambulation activities
(AAs)

16 GMM, GMR PC 68

Maurer et al. ( 2006 ) 4 2D-acc,

microphone,
light sensor,
thermometer

Wrist 6 Locomotion 6 DT, k-NN, NB Wa t ch 92.8

Lara and Labrador
( 2012b )

7 3D-acc, vital

signs, GPS

Chest 3 Locomotion 8 DT Smartphone 92.64

Ermes et al. ( 2008 ) 3 3D-acc Chest, wrist,

ankle

5 Locomotion 3 DT PDA 94

Riboni and Bettini
( 2011 )

3 3D-acc, GPS Wrist, pocket 10 Locomotion, ADL 6 NB, DT, LR, SVM Smartphone,

server

93

Tapi a e t a l . ( 2007 ) 6 3D-acc, heart rate Wrist, arm,

ankle, thigh,
chest, hip

30 Gymnasium 21 NB, DT Laptop 58.6-97.6

L-HAR 2 3D-acc, 3D-gyro Wa i s t 6 Locomotion 30 MC-L1-L2-SVM Smartphone 96.4

PTA-HAR 2 3D-acc, 3D-gyro Wa i s t 7 Locomotion, PTs 30 MC-L1-SVM Smartphone 96.7

P ∗ Number of participants