Multi-channel Speech Separation Using Spatially Selective Deep Non-Linear Filters
This website accompanies our ACM/IEEE Transactions of Audio, Speech and Language Processing submission:
Kristina Tesch, Timo Gerkmann, "Multi-channel Speech Separation Using Spatially Selective Deep Non-linear Filters">
In this work, we propose to use a DNN-based non-linear spatially selective filter (SSF) to perform speech separation by spatial filtering. We evaluate the proposed method based on two network architectures (McNet and JNF) and compare with a direct separation (DS) approach trained with utterance-wise permutation invariant training (PIT).