Articolo in rivista, 2020, ENG, 10.1109/ACCESS.2020.3000693

PROLIFIC: A Fast and Robust Profile-Likelihood-Based Muscle Onset Detection in Electromyogram Using Discrete Fibonacci Search

Suviseshamuthu, Easter S.; Allexandre, Didier; Amato, Umberto; Della Vecchia, Biancamaria; Yue, Guang H.

Kessler Fdn; Rutgers New Jersey Med Sch; CNR; Univ Roma La Sapienza

A stochastic scheme, namely, PLM-Lap, has recently been propounded, which relies on the profile likelihood (PL) constructed with a Laplace distribution for estimating muscle activation onsets (MAOs) in surface electromyographic (sEMG) data. The MAO detection accuracy and robustness of the PLM-Lap have been empirically shown to be better than those of several state-of-the-art approaches. The algorithm designates the data point index associated with the maximum of the PL function as an onset occurrence by regarding every sEMG data point as a candidate onset and hence exhaustively evaluating the objective function. This article concerns an expedient and faster approach premised on the discrete Fibonacci search (DFS) to locate the maximum of the discrete PL function. The experimental results support that both the exhaustive and DFS procedures are equivalent in a statistical sense, whereas the latter offers impressive computational savings by a factor of approximately 90. Owing to the speed-up, the accuracy of MAO estimation may further be enhanced by modeling the sEMG data with a set of PL functions, each one built using a suitable probability distribution, and picking the estimate from the best model. Three statistical criteria, i.e., Kolmogorov-Smirnov, Lilliefors, and Anderson-Darling test, for choosing the probability distribution are recommended. A freely downloadable MATLAB package, namely PROLIFIC, meant for sEMG onset detection is available on MATLAB File Exchange from the following link: https://www.mathworks.com/matlabcentral/fileexchange/76495-prolific-profile-likelihoodbased-on-fibonacci-search.

IEEE access 8 , pp. 105362–105375

Keywords

Anderson-Darling test, discrete Fibonacci search, Kolmogorov-Smirnov test, Lilliefors test, muscle activation onset, profile likelihood, surface electromyography

CNR authors

Amato Umberto

CNR institutes

ID: 451153

Year: 2020

Type: Articolo in rivista

Creation: 2021-04-02 00:13:20.000

Last update: 2021-04-02 00:13:20.000

CNR authors

External links

OAI-PMH: Dublin Core

OAI-PMH: Mods

OAI-PMH: RDF

DOI: 10.1109/ACCESS.2020.3000693

External IDs

CNR OAI-PMH: oai:it.cnr:prodotti:451153

DOI: 10.1109/ACCESS.2020.3000693

ISI Web of Science (WOS): 000541044200066