Unveiling Designers' Cognitive States in Early Engineering Design Stages: An Autoencoder-DNN Approach based on EEG Data

Early engineering product design stages, such as innovative conceptual design and product form design, require intensive thinking by designers. During these tasks, designers experience constantly shifting cognitive states of either Trance, Concentration or Confusion. Accurately recognizing designer's cognitive states is a prerequisite task to provide assistance, e.g. design knowledge recommendation, to designers timely. Electroencephalogram (EEG) data is the external expression of designers' cognitive states. However, current research on applying EEG technology in engineering design scenarios lacks automatic selection of significant features from EEG information, and the connection with the design process is also limited. Faced with such issues, this study proposes a cognitive state recognition procedure for engineering design scenarios, including a recognition model and an experiment protocol. The Autoencoder-Deep Neural Network (DNN)-based recognition model can flexibly select significant features to achieve accurate cognitive state recognition, while the two-stage experiment protocol can collect standard cognitive state data and perform validation in the simulation of real design scenarios. The experiment results demonstrate the validity of the proposed recognition procedure, and confirmed that states of Concentration and Confusion can be utilized to determine whether designers need assistance.