Designing simple nonlinear filters using hysteresis of single recurrent neurons for acoustic signal recognition in robots

Autor(en): Manoonpong, P.
Pasemann, F.
Kolodziejski, C.
Wörgötter, F.
Stichwörter: Acoustic signals; Autonomous robot; Discrete time dynamics; Dynamical properties; Filter designs; Hysteresis effect; Nonlinear filter; Signal filters; Single neuron; Transient dynamics, Acoustic waves; Bandpass filters; High pass filters; Hysteresis; Low pass filters; Robots; Signal filtering and prediction; Signal processing, Recurrent neural networks
Erscheinungsdatum: 2010
Journal: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volumen: 6352 LNCS
Ausgabe: PART 1
Startseite: 374
Seitenende: 383
Zusammenfassung: 
In this article we exploit the discrete-time dynamics of a single neuron with self-connection to systematically design simple signal filters. Due to hysteresis effects and transient dynamics, this single neuron behaves as an adjustable low-pass filter for specific parameter configurations. Extending this neuro-module by two more recurrent neurons leads to versatile high- and band-pass filters. The approach presented here helps to understand how the dynamical properties of recurrent neural networks can be used for filter design. Furthermore, it gives guidance to a new way of implementing sensory preprocessing for acoustic signal recognition in autonomous robots. © 2010 Springer-Verlag Berlin Heidelberg.
Beschreibung: 
Conference of 20th International Conference on Artificial Neural Networks, ICANN 2010 ; Conference Date: 15 September 2010 Through 18 September 2010; Conference Code:82138
ISBN: 9783642158186
ISSN: 03029743
DOI: 10.1007/978-3-642-15819-3_50
Externe URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-78049410282&doi=10.1007%2f978-3-642-15819-3_50&partnerID=40&md5=d3a80cdc0a03a9c6d97f8b9d01cafbc5

Show full item record

Page view(s)

1
Last Week
0
Last month
0
checked on Feb 22, 2024

Google ScholarTM

Check

Altmetric