A one-dimensional signal data restoration method based on convolution neural network

The invention discloses a one-dimensional signal data restoration method based on a convolution neural network. Convolution neural network model based on an Encoder-decoder architecture is constructed, the loss function in the convolution neural network model is weighted, The invention makes the repaired data have more instantaneous characteristics. The method of the invention restores all the lost data excellently through fitting the damaged signal without any prior knowledge, and solves the data packet loss problem caused by factors such as unstable link, node failure and the like in wireless sensor network. L2 regular term is added to avoid over-fitting phenomenon of neural network, choosing iterative number and loss function change rate as training stop condition is helpful to stabilize the performance of neural network and improve the efficiency of data restoration. Adam optimization algorithm has the advantages of fast convergence and good anti-noise performance, and can modify the convolution neural ne