Wavelet transform - artificial neural network receiver with adaptive equalisation for a diffuse indoor optical wireless OOK link

Zabih Ghassemlooy, Robert Dickenson, Sujan Rajbhandari

Research output: Contribution to journalArticlepeer-review

Abstract

This paper presents an alternative approach for signal detection and equalization using the continuous wavelet transform (CWT) and the artificial neural network (ANN) in diffuse indoor optical wireless links (OWL). The wavelet analysis is used for signal preprocessing (feature extraction) and the ANN for signal detection. Traditional receiver architectures based on matched filter (MF) experience significant performance degradation in the presence of artificial light interference (ALI) and multipath induced intersymbol interference (ISI). The proposed receiver structure reduces the effect of ALI and ISI by selecting a particular scale of CWT that corresponds to the desired signal and classifying the signal into binary 1 and 0 based on an observation vector. By selecting particular scales corresponding to the signal, the effect of ALI is reduced. We show that there is little variation when using 30 and 5 neurons in the first layer, with one layer ANN model showing a consistently worse BER performance than other models, whilst the 15 neuron model show some behaviour anomalies from a BER of approximately 10-3. The simulation results show that the Wavelet-ANN architecture outperforms the traditional MF based receiver even with the filter is matched to the ISI affected pulse shape. The Wavelet-ANN receiver is also capable of providing a bit error rate (BER) performance comparable to the equalized forms of traditional receiver structure.
Original languageEnglish
Pages (from-to)1-9
JournalMediterranean Journal of Computers and Networks
Volume5
Issue number1
Publication statusPublished - Jan 2009

Fingerprint Dive into the research topics of 'Wavelet transform - artificial neural network receiver with adaptive equalisation for a diffuse indoor optical wireless OOK link'. Together they form a unique fingerprint.

Cite this