Influence on pattern detection for the solution of the nonlinear system on a discrete Shapelet Transform II

 

محفوظ في:
التفاصيل البيبلوغرافية
المؤلفون: Valdés-Santiago, Damian, León-Mecías, Ángela M., Baguer Díaz-Romañach, Marta L., González-Hidalgo, Manuel, Jaume-I-Capó, Antoni
التنسيق: artículo original
الحالة:Versión publicada
تاريخ النشر:2024
الوصف:The use of adapted wavelets for pattern recognition is very attractive because of the multiscalarity of the wavelet transform. However, the good performance of these algorithms in pattern detection strongly depends on the construction of the filters adapted to the pattern of interest. The Discrete Shapelet Transform II [9] (DST-II) is an algorithm inspired by the wavelet transform, which allows the design of tailored filters for pattern detection in one-dimensional signals. The construction of these filters requires the solution of a system of nonlinear equations, which according to [9] can be performed by any iterative method. This research presents a novel and comprehensive numerical study that demonstrates the impact of the choice of the appropriate numerical method for the solution of the nonlinear system in DST-II. The efficiency of the estimated filters has an impact on the performance of this transform in pattern detection. The best results are obtained by combining Newton’s method with preiteration using the continuation algorithm. The convergence achieved for 55,37% of the patterns suggests that DST-II could be suitable for patterns with specific shapes, useful in biomedical signal applications.
البلد:Portal de Revistas UCR
المؤسسة:Universidad de Costa Rica
Repositorio:Portal de Revistas UCR
اللغة:Español
OAI Identifier:oai:archivo.portal.ucr.ac.cr:article/53834
الوصول للمادة أونلاين:https://archivo.revistas.ucr.ac.cr/index.php/matematica/article/view/53834
كلمة مفتاحية:Diseño de filtros wavelet
Wavelet adaptada
Transformada shapelet discreta
Transformada wavelet discreta
Wavelet filter design
Matched wavelet
Discrete shapelet transform
Discrete wavelet transform