Content Based Singing Voice Extraction From a Musical Mixture
This addresses source separation for audio processing, but it is incremental as it builds on existing deep learning approaches with a focus on content-based extraction.
The authors tackled the problem of extracting raw singing voices from musical mixtures using linguistic content, achieving results that were evaluated subjectively against state-of-the-art methods.
We present a deep learning based methodology for extracting the singing voice signal from a musical mixture based on the underlying linguistic content. Our model follows an encoder decoder architecture and takes as input the magnitude component of the spectrogram of a musical mixture with vocals. The encoder part of the model is trained via knowledge distillation using a teacher network to learn a content embedding, which is decoded to generate the corresponding vocoder features. Using this methodology, we are able to extract the unprocessed raw vocal signal from the mixture even for a processed mixture dataset with singers not seen during training. While the nature of our system makes it incongruous with traditional objective evaluation metrics, we use subjective evaluation via listening tests to compare the methodology to state-of-the-art deep learning based source separation algorithms. We also provide sound examples and source code for reproducibility.