Detection, classification, and recognition based on the detection of energy features of Ultra Wide Band (UWB) vs. signals emitted in the Industrial Scientific and Medical (ISM) radio bands, such as Bluetooth, is a challenging issue. This work addressed this issue by analyzing the behavior of UWB versus Bluetooth signals in various noisy environments. The focus was on identifying robust feature extraction algorithms, that would enable encoding UWB and Bluetooth signals with features such as, for example, short time energy, Fast Fourier transform energy, and derivatives of short time energy. Results of experimental analysis showed that with respect to other signals, short-time energy of UWB over small overlapping time windows had acceptable discriminative performance. The different feature selection algorithms were tested with the following classifiers; Support Vector Machine with related kernel methods, Probabilistic Neural Networks, K Nearest Neighborhood, and Naive Bayes were tested in order to select the best option towards detection performance in different noisy conditions. © 2014 IEEE.
Ultra Wideband and Bluetooth detection based on energy features / Soleimani, Hossein; Caso, Giuseppe; DE NARDIS, Luca; DI BENEDETTO, Maria Gabriella. - ELETTRONICO. - (2014), pp. 96-101. (Intervento presentato al convegno IEEE International Conference on Ultra-Wideband tenutosi a Paris; France nel 1-3 Settembre) [10.1109/ICUWB.2014.6958958].
Ultra Wideband and Bluetooth detection based on energy features
SOLEIMANI, HOSSEIN;CASO, GIUSEPPE;DE NARDIS, LUCA;DI BENEDETTO, Maria Gabriella
2014
Abstract
Detection, classification, and recognition based on the detection of energy features of Ultra Wide Band (UWB) vs. signals emitted in the Industrial Scientific and Medical (ISM) radio bands, such as Bluetooth, is a challenging issue. This work addressed this issue by analyzing the behavior of UWB versus Bluetooth signals in various noisy environments. The focus was on identifying robust feature extraction algorithms, that would enable encoding UWB and Bluetooth signals with features such as, for example, short time energy, Fast Fourier transform energy, and derivatives of short time energy. Results of experimental analysis showed that with respect to other signals, short-time energy of UWB over small overlapping time windows had acceptable discriminative performance. The different feature selection algorithms were tested with the following classifiers; Support Vector Machine with related kernel methods, Probabilistic Neural Networks, K Nearest Neighborhood, and Naive Bayes were tested in order to select the best option towards detection performance in different noisy conditions. © 2014 IEEE.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
