Object identification method based on eeg data

Abstract

The invention discloses an object identification method based on EEG data. The method comprises the following steps: firstly, expanding data by using a data enhancement and extraction technology; randomly dividing data into five equal parts, performing five-fold training, then using ResBlock as a basic structure, constructing a brand-new two-dimensional convolutional neural network, replacing common convolution with cavity convolution for the first three layers of the network, using PReLU as an activation function of the network, and using Focalloss as a loss function of a model; and carrying out model training by using an EEG data object identification data set released in 2017 in a Perceive laboratory. By means of data enhancement, the data of the small data set can be fully utilized. In addition, a deeper network model and fewer network parameters are adopted, although the number of EEG data in the data set is smaller, the features capable of being recognized in the EEG data can be learned as much as possible, and therefore the object recognition task can be efficiently and accurately achieved.

Description

technical field [0001] The invention belongs to the technical field of EEG signal recognition, and in particular relates to a scene recognition system based on EEG signals. Background technique [0002] Currently, humans are much better at classifying scenes than computers. Although the deep learning algorithm in recent years has been able to achieve a very good accuracy rate of scene recognition, the deep learning algorithm still cannot surpass the recognition accuracy rate of humans. Compared with the method of computer vision to directly extract the distinguishable features of the image for recognition, the visual recognition process of the human brain also includes the perception process and recognition theory, for example, the color and shape of the object stimulate the human brain, the human cerebral cortex response to these stimuli. Some studies in the field of neuroscience have pointed out that human brain activity has specific patterns of brain activity for specif...

AI Technical Summary

Problems solved by technology

Although these methods have achieved very good results at present, there is still a lot of room for improvement in the accuracy of recognition. The main reason is that the network still needs to be improved, and the second is that there is less data in the image. Distinguishing features pose difficulties

Images