Fusing time, frequency and shape-related information: Introduction to the Discrete Shapelet Transform's second generation (DST-II)

Abstract

This article introduces the second generation of the Discrete Shapelet Transform (DST-II), a tool created for fusing three types of information: time, frequency and shape-related. Considered a particular Discrete Wavelet Transform (DWT), it allows a productive time-frequency-shape (TFS) joint analysis. In the proposed approach, both the procedure to attain the corresponding filters coefficients and the interpretation of the transformed signal are simplified in relation to the usage of its predecessor, i.e., the DST. Throughout the article, the DST-II formulation is described in detail, including a numerical example, a prototype for use in a diversity of fields and an application on spike and overlap sorting, reassuring the efficacy of the new transform.