Efficient acoustic feature transformation in mismatched environments using a Guided-GAN

Walter Heymans, Marelie H. Davel, Charl Johannes van Heerden

Abstract: Highlights•Generative adversarial network-based method for acoustic feature enhancement.•A trained acoustic model guides the generative adversarial network during training.•The network is a computationally efficient alternative to multi-style training.•No parallel corpora are required to train the generative adversarial network.•The performance of a strong baseline is improved in new mismatched environments.