Edge Sensing of Mental States Using a Scale-Balancing Brain-Computer Interface.

Negative mental states often cause inappropriate behaviors or incorrect operations. Human factors, in particular, are responsible for more than half of all mining safety accidents in the dim and noisy mining environments, endangering workers’ lives and production safety. Although the audiovisual brain-computer interface (BCI) can collect electroencephalogram (EEG) signals, which objectively reflect an individual’s mental state without brain invasion, it is difficult to adapt it for the mining industry because of the lacked balances in the scales for the key electrode determination, EEG filtering, and feature optimization. This work proposes an edge-sensing method for mental states by using a scale-balancing BCI. First, the balanced electrode positions of a BCI are determined by using a fused node integrity evaluation algorithm, which overcomes the coarse-grained sorting problem of the k-shell analysis. Second, an improved filtering algorithm is used for the EEG signal filtering, in which a center frequency aliasing factor is defined to avoid the overlapping problem of the variation mode component analysis. Third, the feature optimal learning algorithm is formed by a hybrid loss function with manifold learning. Fourth, mental state sensing is implemented by using the proposed edge sensing method and the EEG features of intensity and connectivity. Experiments show that a noise intensity above 80 decibels is likely to evoke a negative mental state, and a noise intensity above 105 decibels is more likely to evoke a fearful mental state in the mining environment.