CITATION — REFERENCE ENTRY

Conv-TasNet: Surpassing ideal time-frequency magnitude masking for speech separation — IEEE/ACM Transactions on Audio, Speech, and Language Processing

Revision 58b90a39-c1d5-4a54-ad5d-1ddb2f976704 · 11/21/2026, 6:45:55 AM UTC
Key
luo2019-conv-tasnet
Authors
Luo, Yi; Mesgarani, Nima
Issued
2019-8
Type
article-journal
Container
IEEE/ACM Transactions on Audio, Speech, and Language Processing
Volume
27
Issue
8
Pages
1256-1266
Raw CSL JSON
{
  "DOI": "10.1109/TASLP.2019.2915167",
  "URL": "https://arxiv.org/abs/1809.07454",
  "note": "Claims verified from arXiv 1809.07454 v3; page locators refer to that version.",
  "page": "1256-1266",
  "type": "article-journal",
  "issue": "8",
  "title": "Conv-TasNet: Surpassing ideal time-frequency magnitude masking for speech separation",
  "author": [
    {
      "given": "Yi",
      "family": "Luo"
    },
    {
      "given": "Nima",
      "family": "Mesgarani"
    }
  ],
  "issued": {
    "date-parts": [
      [
        2019,
        8
      ]
    ]
  },
  "volume": "27",
  "container-title": "IEEE/ACM Transactions on Audio, Speech, and Language Processing"
}

Claims

  1. In two-speaker separation, Conv-TasNet surpasses several ideal time-frequency magnitude masks by objective measures and in listening tests.
    "Conv-TasNet surpasses several ideal time-frequency magnitude masks in two-speaker speech separation as evaluated by both objective distortion measures and subjective quality assessment by human listeners."
    Locator: page: 1 (Abstract) · Quote language: en
  2. Conv-TasNet is trained to maximize the scale-invariant source-to-noise ratio (SI-SNR).
    "The objective of training the end-to-end system is maximizing the scale-invariant source-to-noise ratio (SI-SNR), which has commonly been used as the evaluation metric for source separation replacing the standard source-to-distortion ratio (SDR)"
    Locator: page: 5 (sec. III.C) · Quote language: en
  3. Time-frequency formulations of separation have drawbacks: decoupling of phase and magnitude, a suboptimal representation for separation, and long latency in computing spectrograms.
    "The majority of the previous methods have formulated the separation problem through the time-frequency representation of the mixed signal, which has several drawbacks, including the decoupling of the phase and magnitude of the signal, the suboptimality of time-frequency representation for speech separation, and the long latency in calculating the spectrograms."
    Locator: page: 1 (Abstract) · Quote language: en
  4. Conv-TasNet uses a learned linear encoder of the waveform and separates speakers by applying masks to the encoder output.
    "Conv-TasNet uses a linear encoder to generate a representation of the speech waveform optimized for separating individual speakers. Speaker separation is achieved by applying a set of weighting functions (masks) to the encoder output."
    Locator: page: 1 (Abstract) · Quote language: en
Available in