Toward Sensor to Reproduce Human Sense - The Electronic Nose Artificial Olfaction Technology - page 135

and their anatomical changes carried out by this signals. Piezo electric
transducer worked as probes in this type of sensory system. e.g., Sonorgraphy
machine

Characteristics of biosignals:

• Deterministic signal (man-made).

- Classified by mathematical function or rules.
- Periodic signal, transient signal.
• Random signal (stochastic signal).

- Contain uncertainty in the parameters that describe it.
• Real biological signals almost always have some unpredictable noise or change
in parameters.
• e.g., EMG, ECG, EEG.

5.10 Signal Analysis Techniques

Basically multivariate data analysis involves data reduction, i.e., it reduces high
dimensionality in a multivariate problem where variables are somewhat correlated
(e.g., sensors with overlapping sensitivities), allowing the information to be dis-
played in a smaller dimension (typically two or three) [ 19 ]. There are many
multivariate techniques available to choose from, which is shown in Fig. 5.12 .A
subset, called PARC techniques, is normally used in ''E-nose'' data analysis.
Multidimensional scaling (MDS), principal components analysis (PCA), self
organizing maps (SOM), independent component analysis (ICA), Cluster analysis
(CA), linear discriminate analysis (LDA), partial least squares (PLS), feature
subset selection (FSS), principal component regression (PCR), multiple linear

Fig. 5.12 Multivariate pattern analysis techniques methods