学术文献库

Computer-vision technologies have emerged to assist security surveillance. However, automation alert/alarm systems often apply a low-beta threshold to avoid misses and generates excessive false alarms. This study proposed an adaptive hazard diagnosis and alarm system with adjustable alert threshold levels based on environmental scenarios and operator's hazard recognition performance. We recorded electroencephalogram (EEG) data during hazard recognition tasks. The linear ballistic accumulator model was used to decompose the response time into several psychological subcomponents, which were further estimated by a Markov chain Monte Carlo algorithm and compared among different types of hazardous scenarios. Participants were most cautious about falling hazards, followed by electricity hazards, and had the least conservative attitude toward structural hazards. Participants were classified into three performance-level subgroups using a latent profile analysis based on task accuracy. We applied the transfer learning paradigm to classify subgroups based on their time-frequency representations of EEG data. Additionally, two continual learning strategies were investigated to ensure a robust adaptation of the model to predict participants' performance levels in different hazardous scenarios. These findings can be leveraged in real-world brain-computer interface applications, which will provide human trust in automation and promote the successful implementation of alarm technologies.