MMG/EMG Mapping with Reservoir Computing (original) (raw)

Ding, Yuqi, Liang, Xiangpeng, Middelmann, Thomas, Marquetand, Justus and Heidari, Hadi ORCID: https://orcid.org/0000-0001-8412-8164(2022) MMG/EMG Mapping with Reservoir Computing. In: 29th IEEE International Conference on Electronics, Circuits and Systems (ICECS 2022), Glasgow, UK, 24-26 October 2022, ISBN 9781665488235(doi: 10.1109/ICECS202256217.2022.9971109)

Full text not currently available from Enlighten.

Abstract

Magnetomyography(MMG) is the method that measures the magnetic field around the human muscle as an informative bio-signal that has received considerable attention in recent years. However, the noise compositions of MMG signals are complex and difficult to be removed, thus hindering the application of MMG. To extract muscle movement information for MMG and attenuate the effect of noise, this paper proposes a method to convert noisy MMG to clean electromyography (EMG) that also stems from muscle activities. The conversion is done by using a recently proposed electronic Rotating Neuron Reservoir (eRNR) model with high efficiency and strong system approximation ability. This model is trained with our self-collected MMG data as input and the corresponding EMG as target output. After training, the model can successfully map the MMG signal to EMG with acceptable normalised root mean square error (0.3894), offering a new pathway for extracting desirable information from the noisy bio-signal.

University Staff: Request a correction | Enlighten Editors: Update this record

Deposit and Record Details