Low-speed eye movement signal detection method, device and application

The invention relates to a low-speed eye movement signal detection method. The method comprises the following steps: acquiring a horizontal eye movement signal in real time, sliding on the horizontal eye movement signal by using a sliding window with preset duration and step length, acquiring a current window data sample of each sliding, and inputting the current window data sample into a trained multi-scale one-dimensional convolutional neural network model for prediction; the current window data sample passes through a plurality of down-sampling branches with preset down-sampling rates to generate a plurality of electro-oculogram signal branches with different scales; performing one-dimensional convolution operation and maximum pooling operation on each electro-oculogram signal branch to generate a corresponding one-dimensional feature map; and connecting the plurality of one-dimensional feature maps corresponding to the current window data sample to obtain a corresponding full-connection feature map, and j