Conference proceeding
The Use of Multiclass Support Vector Machines and Probabilistic Neural Networks for Signal Classification and Noise Detection in PLC/OFDM Channels
PROCEEDINGS OF THE 2020 30TH INTERNATIONAL CONFERENCE RADIOELEKTRONIKA (RADIOELEKTRONIKA), pp 41-46
01 Jan 2020
Abstract
For the past many years, Artificial Neural Networks (ANNs) have shown powerful performance in many applications. In this paper, the usage of ANNs in pattern recognition (discriminant analysis) have been studied and examined. For the purpose of detecting noise that presents in OFDM signals after being transmitted over a PLC channel, two classification learners were proposed. These classifiers are multiclass support vector machines (SVMs) with the error-correcting output codes (ECOC) and probabilistic neural networks (PNNs). A training dataset of 5,000 randomly generated signals transmitted over PLC channels, where each received signal is associated with its category based on its amplitude, was used to train the proposed classifiers. The purpose of this study was to decide on the optimum classification scheme among the proposed methods in terms of computational cost and classification accuracy. In general, our research demonstrated that our proposed algorithms trained on the PLC signals features achieved high classification accuracy, for instance the PNN obtained classification accuracy of 94.3% whilst the classification accuracy produced by the SVM using fine Gaussian kernel function was 96.4%. Therefore, they can be viewed as robust supervised classification learners.
Metrics
Details
- Title
- The Use of Multiclass Support Vector Machines and Probabilistic Neural Networks for Signal Classification and Noise Detection in PLC/OFDM Channels
- Creators
- Dalal H. Baroud - University of JohannesburgAli N. Hasan - University of JohannesburgT. Shongwe - University of Johannesburg
- Publication Details
- PROCEEDINGS OF THE 2020 30TH INTERNATIONAL CONFERENCE RADIOELEKTRONIKA (RADIOELEKTRONIKA), pp 41-46
- Publisher
- IEEE
- Number of pages
- 6
- Resource Type
- Conference proceeding
- Language
- English
- Academic Unit
- Engineering Leadership and Society/Engineering Technology
- Web of Science ID
- WOS:000610803200008
- Scopus ID
- 2-s2.0-85085730137
- Other Identifier
- 991022004635704721
InCites Highlights
Data related to this publication, from InCites Benchmarking & Analytics tool:
- Web of Science research areas
- Engineering, Electrical & Electronic
- Telecommunications