Denoising of discrete-time chaotic signals using echo state networks

André L. O. Duarte,Marcio Eisencraft

DOI: https://doi.org/10.1016/j.sigpro.2023.109252

2023-09-20

Abstract:Noise reduction is a relevant topic when considering the application of chaotic signals in practical problems, such as communication systems or modeling biomedical signals. In this paper an echo state network (ESN) is employed to denoise a discrete-time chaotic signal corrupted by additive white Gaussian noise. The choice for applying ESNs in this context is motivated by their successful exploitation for separation and prediction of chaotic signals. The results show that the processing gain of ESN is higher than that of the Wiener filter, especially when the power spectral density of the chaotic signals is white.

Signal Processing,Chaotic Dynamics

What problem does this paper attempt to address?

The paper primarily explores how to use Echo State Networks (ESN) to denoise discrete-time chaotic signals. Specifically, the paper addresses the following key issues: 1. **Background and Motivation**: Chaotic signals are widely used in fields such as communication systems, biomedical signal modeling, and time series prediction, but these signals are often affected by noise. Traditional linear filtering techniques like the Wiener Filter (WF) are effective but have limited performance in nonlinear scenarios. Given this, researchers consider using machine learning methods, particularly Echo State Networks (a special type of recurrent neural network), to achieve better denoising performance. 2. **Research Object**: The paper selects the skew-tent map as the chaotic signal generator because the power spectral density (PSD) of the chaotic signals generated by this map can be directly determined by its single parameter. This facilitates the analysis of denoising tasks under different spectral characteristics. 3. **Methodology**: By comparing the denoising performance of Echo State Networks with the optimal linear filtering technique—the Wiener Filter—under different parameter settings, the effectiveness and superiority of Echo State Networks are validated. Specifically, the researchers adjusted key parameters of the Echo State Network (such as the leakage rate and the spectral radius of the internal matrix) through a series of experiments to optimize its denoising capability and compared it with the Wiener Filter. 4. **Results and Analysis**: Experimental results show that for all tested parameter settings, Echo State Networks provide higher processing gain (i.e., improvement in signal-to-noise ratio after denoising) than the Wiener Filter. Especially when the power spectral density of the chaotic signal tends towards white noise (i.e., when the parameter α approaches 0), the performance of the Wiener Filter significantly degrades, while the Echo State Network still maintains good denoising performance. In summary, the main contribution of this paper is to demonstrate that Echo State Networks, as a nonlinear machine learning technique, have significant advantages in the task of denoising discrete-time chaotic signals, particularly when dealing with chaotic signals with narrowband spectral characteristics.

Denoising of discrete-time chaotic signals using echo state networks

Control of discrete chaotic systems based on echo state network modeling with an adaptive noise canceler

Optimizing Echo State Networks for Enhancing Large Prediction Horizons of Chaotic Time Series

Chaotic Time Series Prediction Using a Novel Echo State Network Model with Input Reconstruction, Bayesian Ridge Regression and Independent Component Analysis

Control-aware echo state networks (Ca-ESN) for the suppression of extreme events

Chaotic signal denoising based on energy selection TQWT and adaptive SVD

Delayed self-feedback echo state network for long-term dynamics of hyperchaotic systems

Echo state networks for modeling turbulent convection

Improved wavelet packet denoising algorithm using fuzzy threshold and correlation analysis for chaotic signals

Edge of stability echo state networks

Echo State Property upon Noisy Driving Input

Physics-Informed Echo State Networks for Chaotic Systems Forecasting

Sequential Learning from Noisy Data: Data-Assimilation Meets Echo-State Network

Physics-informed hierarchical echo state network for predicting the dynamics of chaotic systems

Predicting Chaotic Systems with Quantum Echo-state Networks

Adversarial Signal Denoising with Encoder-Decoder Networks

A time-frequency denoising method for single-channel event-related EEG

Sparse compressed deep echo state network with improved arithmetic optimization algorithm for chaotic time series prediction

A novel ECG signal denoising filter selection algorithm based on conventional neural networks

A Dynamic Threshold Value Control Method for Chaotic Neural Networks

Hybrid Regularized Echo State Network for Multivariate Chaotic Time Series Prediction