[30] |
J. Feng, et al. (2011) |
TGS826, TGS813, TGS825, TGS800, TGS816, TGS2620, TGS822, TGS2602, TGS2600, QS01, WSP2111, MQ138, MQ135, SP3S-AQ2 and AQ sensor |
Pump suction |
Wound detection (P. aeruginosa, E. coli and S. aureus) |
The smell of mice themselves |
Wavelet transform, RBF |
Recognition rate: RBF with ‘Leave-one-out’ method: 95%; RBF with ‘40 Training + 40 Test’ method: 97.5%. |
[31] |
F. Tian, et al. (2012) |
TGS826, TGS813, TGS825, TGS800, TGS816, TGS2620, TGS822, TGS2602, TGS2600, QS01, WSP2111, MQ138, MQ135, SP3S-AQ2 and AQ sensor |
Pump suction |
Wound detection (P. aeruginosa, E. coli and S. aureus) |
The smell of mice themselves |
ICA, RBF |
Recognition rate: 96.25%. |
[35] |
J. Feng, et al. (2014) |
TGS826, TGS813, TGS825, TGS800, TGS816, TGS2620, TGS822, TGS2602, TGS2600, QS01, WSP2111, MQ138, MQ135, SP3S-AQ2 and AQ sensor. |
Pump suction |
Wound detection (P. aeruginosa, E. coli and S. aureus) |
The smell of mice themselves |
OSC, RBF, PSO |
Recognition rate: 97.5%. |
[36] |
F. Tian, et al. (2012) |
TGS2602, TGS2620, TGS2201 |
Diffusion sampling |
Formaldehyde and benzene |
Noise interference |
PCA, ICA, RBF |
Average relative prediction error: formaldehyde: 30.4%; benzene: 10.726%. |
[37] |
R. Gutierrez-Osuna, et al. (2004) |
TGS2602, TGS2610, TGS2611, TGS2620 |
Pump suction |
Acetone, isopropyl alcohol and ammonia |
Background chemicals |
Generalization Fisher’s linear discriminants |
Cancel the effect of both single and mixture backgrounds |
[38] |
R. Gutierrez-Osuna, et al. (2003) |
TGS2602, TGS2610, TGS2611, TGS2620 |
Pump suction |
Acetone, isopropyl alcohol and ammonia |
Background odors |
Linear discriminant function, KIII model |
Eliminate the memory effect of previously detected |
[39] |
A. Gutierrez-Galvez, et al. (2006) |
TGS2602, TGS2610, TGS2611, TGS2620 |
Pump suction |
Acetone, isopropyl alcohol and ammonia |
Background odors |
KIII model |
Anti-Hebbian term can reduce the overlap between patterns |