Neural Decoding Mechanism Exploration Of Primary Motor Cortex

Organisms pass information mainly through generating neural impulse signals,and the neural impulse signals of the motor cortex contain larges of corresponding motion information,which especially catches the attention of the researchers.With the continuous development of electronic science and technology and neuroscience,researchers can use the method of implanting microelectrode array to extract neural signals from the motor cortex,and use some mathematical methods to analyze the neural signals to gain the movement information from them.Then they convert the movement information to instructions through the neural interface,so they can achieve the target of helping the disabled people recovering their motion abilities.The process of analyzing of neural signals to get the motion information is called neural decoding.By analyzing and processing the neural signals extracted from the monkey's primary motor cortex,we predict the movement variables and track the trajectory in the grasping movement of the monkey respectively to study the neural decoding mechanism.Firstly,several different pattern recognition algorithms,including SVM and KNN and ELM have been utilized to predict the movement directions and angles of the hand of the monkey.The prediction results of the Angles and directions have similar laws that the predicted average accuracies increased with the increase of number of neurons used in the prediction.The highest average accuracy can be achieved when using SVM,but the time cost is a little high.When KNN is used,the decoding process can be very fast,but the average accuracy is a little lower.In addition,based on the analysis of the signal of single neuron,kalman filter and sparse coding have been utilized to the tracking of the monkey's hand trajectories.Choosing the number of neural pulse in a short period of time as the feature,we use Least squares(LS)method estimate the parameters in the model of Kalman filter and build the filter model;in contrast,we use sparse coding to transform the feature and build another model and use two models to track the hand trajectories of the monkey.The results suggest that the use of sparse coding in feature extraction can enhance the accuracy of trajectory prediction.