FilterNet: A deep convolutional neural network for measuring plastic deformation from raw Barkhausen noise waveform

•A convolutional neural network named FilterNet which contains three filter-type convolutional kernels was proposed to learn feature representation directly from raw magnetic Barkhausen noise (MBN).•The phenomenon of performance of a trained model declined due to the change in the angle between the electromagnetic sensor and specimens was studied.•The performance of three regression models, which are multiple linear regression, BP neural network, and the FilterNet, were compared using multi-angles MBN signals.