Using Machine Learning and Non-Intervention Methods to Explore the Visual-Color Neural Sensing Underlying Interaction in the Temporal and Spatial Dimensions

Many previous studies have shown that visualcolor sensing enables the brain to interact with the outside world. However, little research has been done on the visual-color sensing underlying interaction mechanisms. Since color is usually defined as a psychological category in traditional understanding, exploring the human visual-color sensing phenomenon by biological methods usually requires other fields' theoretical methods intervened to achieve. In this work, we review related hybrid decoding methods that capture corresponding behavioral and imaginary responses to visual color stimuli and find that existing methods cannot independently explain the visual-color sensing underlying interaction mechanism and essential features. Recently, with the development of brain-computer interfaces, decoding visual sensing information by electroencephalogram (EEG) has become a more efficient method. Inspired by this, we propose a new method based on machine learning and sensing paradigm evolution to explore the visual-color neural underlying interaction mechanism and essential features. This method can not only completely distinguish the color information from other intervention information for independent analysis, but also obtain another clear result by the classification accuracy dimension. In addition, in the fast response communication based on event-related potential (ERP) theory, we successfully captured the visual-color sensing EEG information after P300. Compared with the EEG power information that is fast lost after P300, this continuously persistent information can help us better understand the visual-color interaction process in the brain. In fact, the simpler the sensing paradigm evolves, the more obvious underlying sensing features will be in the EEG. To express the difference between different colors of EEG information, we use extreme colors with strong contrast effects on the objects selected as stimulus targets. We choose three extremely pure basic colors to classify and decode them on fit support vector machines (fitSVM). The results show that the visual-color sensing EEG information can be well decoded by machine learning through paradigm evolution.