The Additive Margin SincNet (AM-SincNet) is a new approach for speaker recognition problems which is based in the neural network architecture SincNet and the additive margin softmax (AM-Softmax) loss function. It uses the architecture of the SincNet, but with an improved AM-Softmax layer.