EEG Signals Decoding of Freely Moving Rats Based on Support Vector Machine Algorithm

Brain-Computer Interface (BCI) technology can record spike signals and local field potential signals (LFP). The research on spike signals have achieved results, but the effect of spike recording will still decrease over time. The LFP can record the signal stably for a long time and make up for the deficiency of recording the spike signal alone. This paper took rats as its research object and researched the spike signals and LFP of the primary motor cortex of the rats after gait training. Meanwhile, the gait signals of the rats are collected. Based on the gait data, the rats’ “running” and “standing” motion states are divided. The corresponding time periods are determined to obtain LFP in different motion states. The time domain, frequency domain, and nonlinear neural dynamics characteristics of the two signals are analyzed respectively. The zero-crossing ratio, power spectral density and sample entropy characteristics of the signals under different behaviors are extracted. The support vector machine algorithm is used to train and predict the extracted features. The feature classification based on power spectrum density has achieved good results, and the classification accuracy can reach 79.14%. In addition, based on the spike signal to predict the movement speed of the rat, the predicted value and the actual value follow better, which can achieve the purpose of control of external equipment. This article provides a new complete method for identifying the movement intention of the hind limbs of freely moving rats from a variety of EEG signals and provides a research idea for the control of brain-controlled smart equipment.