The aim of this activity is to propose new solutions for the functionalization of THz devices in order to be able to respond to applicative as well as academic demands. Initially, we have more precisely targeted two promising application areas: product identification and non-destructive testing.
These two application fields allow us to efficiently use the advantages of the THz field (material transparency and short wavelengths) compared to those of our main RF or optical "competitors". Three complementary tasks are carried out in parallel.
• Device functionalization, whose aim is to design a device whose signature carries sufficiently rich information. Priority is given to all-dielectric structures, since structures incorporating metal are less "discrete" and serve as fall-back solutions either for non-destructive testing or for identification.
• Several characterization techniques are also envisaged to obtain the THz signature: THz-TDS to obtain wide band information and imaging, possibly multi-spectral, to obtain information relating to the structuring or constitution of the structure. If necessary, systems with particular and specific characteristics could be developed (coll. startup(TIHIVE).
• Finally, in partnership with laboratories specialized in the field, signal processing and statistical analysis bricks adapted to THz signals are being investigated. In a first step, the performances of several methods will be evaluated: a) analysis of parsimonious signal components (GIPSA-LAB coll.): phase diagrams, dictionary projection (search for invariants, i.e. characteristics of the signal allowing for example its identification), 2D wavelet decomposition (applied to imaging), learning by tests (data mining), analysis of transient signals and b) classification analysis (Coll. INRIA): multivariate statistics (Principal Component Analysis-ACP, Linear Discriminant Analysis-LDA, ...) and inferential statistics (ROC curve), applied to different metrics (correlation coefficient, Euclidean distance, ...).