44 results about "Nonlinear signal processing" patented technology

An electroencephalogram signal decoding method based on a deep convolutional generative adversarial neural network comprises: firstly, converting an electroencephalogram signal into a time-frequency domain image signal from a time sequence signal through short-time Fourier transform; and integrating the convolutional neural classification network and the adversarial generation network, performingdata enhancement on the small sample electroencephalogram data by using the adversarial generation network, putting the enhanced data into the convolutional neural classification network, and performing classification to realize decoding. The structural advantages of the convolutional neural classification network and the adversarial generative network are combined, the defects existing in small sample data processing through the neural network are overcome, the problems existing in the non-stationary and non-linear signal processing process are effectively solved, and the spontaneous electroencephalogram signal decoding precision is improved; and meanwhile, a new solution is provided for electroencephalogram signal data enhancement, and a new thought is provided for reducing the calibration time and improving the generalization ability of the classification model in the actual operation process in the future.