Skip to contentScholarGate
LibraryBookshelfDeskReview StudioAssistant
Sign in
On this page
IntuitionHow it worksWhen to use itStrengths & limitationsCommon pitfallsApplicationsFrequently asked🔒 Read the full methodSourcesRelated methods
Cite this pageSpotted an issue on this page? Report or suggest a fix →
Home›Music Information Retrieval›Vocal Separation
Machine learningSource separation and demixing

Vocal Separation

Vocal Separation and Source Separation Algorithm · Also known as: singing voice extraction, voice isolation, source demixing

Vocal separation is the task of isolating the singing voice from a mixed music recording, leaving the instrumental accompaniment. Introduced formally by Han et al. (2012), it is critical for music editing, remixing, karaoke generation, and music analysis. Modern deep learning approaches (Défossez et al., 2021) have achieved impressive quality, enabling practical applications in music production and streaming services. Vocal separation is a special case of source separation, where the goal is to isolate the most perceptually salient source.

ScholarGate
  1. Machine learning
  2. v1
  3. 3 Sources
  4. PUBLISHED
Cite this page →
Tools & resources
Download slides
Learn & explore

Read the full method

Members only

Sign in with a free account to read this section.

Sign in

Method map

The neighbourhood of related methods — select a node to explore.

Vocal Separation
Automatic Music Transcri…Beat TrackingMelody ExtractionMusic SegmentationPitch Detection AlgorithmTimbre Analysis

When to use it

Use vocal separation for music remixing, karaoke generation, music-to-lyrics synchronization, and vocal-focused analysis. It works well on commercially produced music with clear vocal-instrumental separation. Avoid it for heavily overlapped or dense arrangements, music with harmonies and backing vocals indistinguishable from leads, or genres where vocals blend seamlessly (some jazz, ambient).

Strengths & limitations

Strengths
  • Enables music remixing and creative production without manual multitrack recording.
  • Useful for karaoke and music-to-lyrics alignment applications.
  • Modern deep learning models achieve high quality (SDR > 6 dB) on standard benchmarks.
  • Can isolate other sources (bass, drums) with the same architecture.
Limitations
  • Separation quality degrades on music with tightly integrated vocals and accompaniment.
  • Backing vocals and harmonies often leak into either separated source.
  • Requires clean reference separated data for training; large annotated datasets are expensive.
  • Phase reconstruction artifacts can introduce distortion even with good magnitude estimates.

Frequently asked

Can vocal separation isolate individual backing vocals or harmonies?

Not reliably with standard methods. Standard approaches treat all vocals as one source. Isolating specific vocal layers requires specialized training data and architectures designed for multi-voice scenarios.

What is a spectral mask and how does it work?

A spectral mask is a time-frequency matrix (same shape as the spectrogram) with values 0–1 indicating the fraction of each time-frequency cell belonging to the target source (vocals). Multiplying the mix spectrogram by the mask yields the estimated source.

How is the separated audio reconstructed from a spectrogram?

Using inverse STFT (Griffin-Lim or learned phase reconstruction). Griffin-Lim is fast but introduces artifacts; modern methods learn phase from data, reducing distortion.

How good must vocal separation be for practical use?

For remixing and karaoke, SDR > 6 dB (good separation) is acceptable. For commercial-grade production, SDR > 8–10 dB is expected. Subjective listening tests often find lower objective SDR acceptable if artifacts are minimal.

Sources

  1. Han, Y., Qin, Z., & Kang, Z. (2012). Singing voice separation using spectral floor filtered spectrograms. In Proceedings of the International Society for Music Information Retrieval Conference. link ↗
  2. Huang, P. S., Kim, M., Hasegawa-Johnson, M., & Smaragdis, P. (2015). Joint optimization of masks and deep recurrent neural networks for monaural source separation. IEEE Transactions on Audio, Speech, and Language Processing, 23(12), 2136-2147. DOI: 10.1109/taslp.2015.2468583 ↗
  3. Défossez, A., Usunier, N., Bottou, L., & Bach, F. (2021). Music source separation in the waveform domain. In International Conference on Learning Representations. link ↗

How to cite this page

ScholarGate. (2026, June 3). Vocal Separation and Source Separation Algorithm. ScholarGate. https://scholargate.app/en/music-information-retrieval/vocal-separation

Related methods

Automatic Music TranscriptionBeat TrackingMelody ExtractionMusic SegmentationPitch Detection Algorithm

Which method?

Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.

  • Automatic Music TranscriptionMusic Information Retrieval↔ compare
  • Beat TrackingMusic Information Retrieval↔ compare
  • Melody ExtractionMusic Information Retrieval↔ compare
  • Music SegmentationMusic Information Retrieval↔ compare
  • Pitch Detection AlgorithmMusic Information Retrieval↔ compare
Compare side by side →

Referenced by

Melody ExtractionPitch Detection AlgorithmTimbre Analysis

Similar methods

Melody ExtractionAutomatic Music TranscriptionPitch Detection AlgorithmInstrument RecognitionMusic SegmentationChord RecognitionCepstral AnalysisMusic Genre Classification

Related reference concepts

Automatic Speech RecognitionSpeech SynthesisThe Source-Filter Model of SpeechAcoustic Cues and FormantsSupport Vector Machines and Kernel MethodsSpeech Perception and Intelligibility

Spotted an issue on this page? Report or suggest a fix →

ScholarGate — Vocal Separation (Vocal Separation and Source Separation Algorithm). Retrieved 2026-07-21 from https://scholargate.app/en/music-information-retrieval/vocal-separation · Dataset: https://doi.org/10.5281/zenodo.20539026
Quick facts
Originator
Yonggang Han
Subfamily
Source separation and demixing
Year
2012
Type
Audio source separation
Related methods
Automatic Music TranscriptionBeat TrackingMelody ExtractionMusic SegmentationPitch Detection Algorithm
ScholarGate

A content-first reference library for research methods — what each one is, how it works, and where it comes from.

Open data (CC-BY)

Explore

  • Library
  • Search the library…
  • Browse by field
  • Fields
  • Journey
  • Compare
  • Which method?

Reference

  • Subjects
  • Atlas
  • Glossary
  • Methodology
  • Philosophy

Your tools

  • Bookshelf
  • Desk
  • Chat

Company

  • About
  • Pricing
  • Contact
  • Suggest a method

Entries are compiled from published sources for reference. Verifying the accuracy and suitability of any information for your own use remains your responsibility.

© 2026 ScholarGate · A research-method reference library
  • Privacy
  • Cookies
  • Terms
  • Delete account