Disentangling the Impacts of Language and Channel Variability on Speech Separation Networks
This work addresses domain mismatch for speech separation systems, focusing on channel variability, which is incremental as it builds on prior research on speaker and environment mismatches.
The study tackled the problem of domain mismatch in speech separation networks, specifically disentangling the impacts of language and channel variability, and found that channel differences have a much larger effect than language differences, with training on Android phone data yielding the best generalization and a new projection-based method improving performance on in-the-wild test data.
Because the performance of speech separation is excellent for speech in which two speakers completely overlap, research attention has been shifted to dealing with more realistic scenarios. However, domain mismatch between training/test situations due to factors, such as speaker, content, channel, and environment, remains a severe problem for speech separation. Speaker and environment mismatches have been studied in the existing literature. Nevertheless, there are few studies on speech content and channel mismatches. Moreover, the impacts of language and channel in these studies are mostly tangled. In this study, we create several datasets for various experiments. The results show that the impacts of different languages are small enough to be ignored compared to the impacts of different channels. In our experiments, training on data recorded by Android phones leads to the best generalizability. Moreover, we provide a new solution for channel mismatch by evaluating projection, where the channel similarity can be measured and used to effectively select additional training data to improve the performance of in-the-wild test data.