ZCR-aided neurocomputing: A study with applications

Abstract

This paper covers a particular area of interest in pattern recognition and knowledge-based systems (PRKbS), being intended for both young researchers and academic professionals who are looking for a polished and refined material. Its aim, playing the role of a tutorial that introduces three feature extraction (FE) approaches based on zero-crossing rates (ZCRs), is to offer cutting-edge algorithms in which clarity and creativity are predominant. The theory, smoothly shown and accompanied by numerical examples, innovatively characterises ZCRs as being neurocomputing agents. Source-codes in C/C++ programming language and interesting applications on speech segmentation, image border extraction and biomedical signal analysis complement the text.

A tutorial review on entropy-based handcrafted feature extraction for information fusion

“…This is the third in a set of tutorials I have recently published with the same objective: innovative usage of humble and wellknown concepts for the benefit of both digital signal processing (DSP) and pattern recognition (PR) communities. The preceding texts, [23] and [24] , were respectively dedicated to the exploration of relevant aspects of energy by means of proposed methods A 1 , A 2 and A 3 , and zero-crossing rates (ZCRs), according to the techniques introduced as B 1 , B 2 and B 3 . Successfully, I employed those formulations for neurophysiological signal analysis, texture characterisation, text-dependent speaker verification, speech classification and segmentation, image border extraction and biomedical signal processing.…”

“…Particularly, I demonstrate that H , by itself, obtained based on two proposed approaches for FE from both unidimensional (1D) and bidimensional (2D) data, has flagrant potential, as also evidenced in relevant scientific articles published last year [11,17,20,42,50,51,53,81,84,86] and a few years ago [6,15,21,37,56,57,62,63,73,75,83] . Similarly to the characterization of ZCRs as neurocomputing agents [24] , H is shown to be the outcome of a specifically tuned deep neural network (DNN) that fuses important information, bringing an innovative point-of-view for both DSP and PR communities. Furthermore, experiments and applications on restricted-vocabulary speech recognition and image synthesis reassure the efficacy of the proposed techniques.…”

“…Compromised with a balance among creativity, simplicity and accuracy , exactly as in [23] and [24] , this paper is organised as follows. A review on H accompanied by some of its recent applications is the theme of the next subsection.…”

“…Illustrative experiments and applications involving 1D and 2D signals can be found in Section 4 and, lastly, I conclude the paper. Following my previous rationale, as indicated in [24] -pp.1 with respect to article [23] , I emphasise the importance of those articles, suggesting their readings beforehand for a better understanding of the ideas discussed herein.…”

“…Fundamentally and in comparison to the concepts of ene.g. [23] and ZCR [24] , entropy can reveal characteristics from the signal under analysis which are unlikely found based on those features. The humble example shown in Fig.…”