Breath Sounds by Kostas N. Priftis Leontios J. Hadjileontiadis & Mark L. Everard

Author:Kostas N. Priftis, Leontios J. Hadjileontiadis & Mark L. Everard
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

9.9 Emerging Approaches

9.9.1 Swarm Decomposition (SwD)–Swarm Transform (SwT)

In Sects. 9.5 and 9.7, the MRD–MRR and the EMD/EEMD multiresolution analysis schemes were referenced, respectively, as bases to build upon the analysis of LSs and derive new diagnostic features. The issue behind the use of MRD–MRR is the demand for an a priori selection of the analysis basis (wavelet) from a given library (e.g., Daubechies, Morlet), whereas in the EMD/EEMD case, this requirement is waived, as the analysis basis is produced from the signal itself, via the sifting process. Nevertheless, as the name of EMD/EEMD denotes, it is an empirical approach.

In the light of the aforementioned, a new multiresolution analysis scheme has been recently proposed [161], based on the swarm perception, namely, the swarm decomposition (SwD). The latter circumvents the empirical character of EMD by proposing a more deterministic approach to achieve effective multiresolution decomposition.

In the SwD, a swarming model is used as the analysis basis, where the processing is intuitively considered as a virtual swarm–prey hunting, where the prey is the signal itself. According to the state of the swarm, i.e., relations of its members, different oscillatory characteristics of the signal are revealed (e.g., its low/high frequencies when the swarm has low/high coherence between its members). To systematically use this type of filtering, the relations between the swarm parameters and particular responses are needed; these are derived using a genetic algorithm [161]. Eventually, the SwD is realized by iteratively applying the swarm filtering, where, at each iteration, it is properly parametrized, so as to result in an oscillatory mode (OM) sequence (roughly corresponding to the IMFs of the EMD/EEMD). The main advantage of the proposed swarm-based perspective is that SwD allows for the efficient decomposition of a signal into components that preserve physical meaning, likewise EMD/EEMD, being yet based on a rigid mathematical model [161]. By applying Hilbert transform to the resulted OM components, a time-frequency representation is formed, namely, swarm transform (SwT), which allows for better discrimination of the localization in time and frequency of the inherent oscillations of the input signal (e.g., pitched LSs, such as wheezes).

Figure 9.6a displays the OMs derived from the analysis of an asthmatic wheeze (first panel of Fig. 9.6a) using the SwD, whereas Fig. 9.6b displays the corresponding output of the EEMD analysis (first seven IMFs) of the same input LS signal (first panel of Fig. 9.6b). In addition, Fig. 9.7a displays the estimated SwT corresponding to the OMs from the SwD of Fig. 9.6a, whereas Fig. 9.7b shows the Hilbert–Huang spectrum, as the Hilbert transform of all derived IMFs via the EEMD displayed (in part) in Fig. 9.6b.

Fig. 9.6Application of (a) SwD and (b) EEMD to an asthmatic wheeze recording. In both cases, the recorded LS signal along with the seven OMs and IMFs are depicted, respectively

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.