A hybrid approach with multi-channel i-vectors and convolutional neural networks for acoustic scene classification.