Wireless Channel classification with Entropy Enhanced Supervised Time Series Forest (EESTSF)

The classification of wireless channel propagation scenarios is an important function in wireless communication design and machine learning (ML) has been increasingly used to implement such function. In recent times, Deep Learning (DL) models like Convolutional Neural Networks (CNN) or Long Short Term Memory (LSTM) have achieved a higher classification performance in comparison to feature extraction methods at a significant computational complexity and time. An alternative approach would be to rely on Time Series Classification (TSC) algorithms but they may also incur in high computational costs. A new computing efficient TSC algorithm called Supervised Time Series Forest (STSF) has been recently introduced in the literature for a different set of applications and this paper provides an evaluation in this specific context. In addition, this paper introduces an improvement on the initial design of the STSF with the addition of entropy measures to generate the Entropy Enhanced STSF (EESTSF) algorithm. EESTSF is applied to an open source data set with 9 different channel propagation scenarios, where it is shown to outperform significantly not only RF, KNN and CNN but also STSF, especially in low Signal to Noise (SNR) conditions.

BALDINI Gianmarco;  KANDEEPAN Sithamparanathan; 

2025-07-30

IEEE

JRC140923

https://ieeexplore.ieee.org/document/11059599,    https://publications.jrc.ec.europa.eu/repository/handle/JRC140923,   

10.1109/IWCMC65282.2025.11059599 (online),