CITATION — REFERENCE ENTRY
SDR – Half-baked or Well Done? — ICASSP 2019: IEEE International Conference on Acoustics, Speech and Signal Processing
- Key
- leroux2019-si-sdr
- Authors
- Le Roux, Jonathan; Wisdom, Scott; Erdogan, Hakan; Hershey, John R.
- Issued
- 2019-5
- Type
- paper-conference
- Container
- ICASSP 2019: IEEE International Conference on Acoustics, Speech and Signal Processing
- Pages
- 626-630
Raw CSL JSON
{
"DOI": "10.1109/ICASSP.2019.8683855",
"URL": "https://arxiv.org/abs/1811.02508",
"note": "Claims verified from arXiv 1811.02508 v1; page locators refer to that version.",
"page": "626-630",
"type": "paper-conference",
"title": "SDR – Half-baked or Well Done?",
"author": [
{
"given": "Jonathan",
"family": "Le Roux"
},
{
"given": "Scott",
"family": "Wisdom"
},
{
"given": "Hakan",
"family": "Erdogan"
},
{
"given": "John R.",
"family": "Hershey"
}
],
"issued": {
"date-parts": [
[
2019,
5
]
]
},
"container-title": "ICASSP 2019: IEEE International Conference on Acoustics, Speech and Signal Processing"
}
Claims
-
bss_eval_sources completely forgives channel errors that a time-invariant 512-tap filter can account for.
"bss_eval_sources completely forgives channel errors that can be accounted for by a time-invariant 512-tap filter, modifying the reference to best fit each estimate."
-
Le Roux and colleagues argue that the SDR implemented in the BSS Eval toolkit has been improperly used, especially for single-channel separation, giving misleading results.
"We argue here that the signal-to-distortion ratio (SDR) implemented in the BSS eval toolkit has generally been improperly used and abused, especially in the case of single-channel separation, resulting in misleading results."
-
The authors propose a simpler, more robust measure, scale-invariant SDR (SI-SDR).
"We propose to use a slightly modified definition, resulting in a simpler, more robust measure, called scale-invariant SDR (SI-SDR)."
Available in