Design of Multiple-Input Single-Output System for EEG Signals.

Existing electroencephalogram (EEG) collection devices primarily include EEG collection systems for medical scientific research and BrainLink or earphones for commercial advertising. The EEG collection systems employ multiple-input multiple-output connections, making the circuit and wiring complex. However, the BrainLink or earphones are usually designed for entertainment mainly collecting one or two EEG locations. To address these problems, in this paper, a multiple-input single-output collection device (MISOCD) is developed that can obtain a single-channel mixed signal, and a multiple-input single-output blind separation (MISOBS) algorithm is proposed to separate multiple-channel signals from a single-channel mixed signal by introducing an improved parallel dual generative adversarial network that can build one-to-multiple mapping. An experiment is conducted on synthetic signals to illustrate the separation performance of the proposed MISOBS algorithm, which significantly outperforms the state-of-the-art algorithms. The real signals collected by the MISOCD are used for gesture recognition. [ABSTRACT FROM AUTHOR]