Embedded Volterra for prediction of electromyographic signals during labour

Abstract

It has been demonstrated that the dynamics of abdominal electromyographic signals (AEMG) during labour contractions are multi-fractal chaotic. A new embedded multi-step Volterra structure, which exploits the non-linear signal dynamics embedded in the attractor and integrates them in the design of such structures to gauge the long-term behaviour of the dynamics, has been introduced. The long-term predictive capability of the structure is tested by using a closed-loop adaptation scheme without any external input signal applied to the structure. Evidence of long-term prediction of highly complex labour contraction signals using only a small fraction of this sample is provided. In this paper, the Non-linear Auto-Regressive with exogenous inputs (NARX) Recurrent Neural Network (RNN) Multi-Layer Perceptron (MLP) model and the embedded cubic Volterra structure for the reconstruction of the underlying dynamics of labour contraction signals are compared.