| Abstract: |
| In this talk, we deal with the Discrete Shapelet Transform (DST), that estimates the degree of similarity between the signal under analysis and a pre-specified shape. This transform, based on a fractal-based criterion, allows us to to obtain a time-frequency-shape joint analysis. We prove that the replacement of the fractal-based criterion with a correlation based formulation improves the DST. Moreover, the recent results concerning the fractal nature of prime numbers suggest us to reformulate the fractal-based criterion, in order to improve this technique. In addition, we discuss the application of the DST in image in imagine fusion, PDEs and more generally in any fractal-like set. |
|