{"id":5359,"date":"2026-06-18T15:50:00","date_gmt":"2026-06-18T07:50:00","guid":{"rendered":"https:\/\/www.15zhi.net\/blog\/?p=5359"},"modified":"2026-06-18T15:50:00","modified_gmt":"2026-06-18T07:50:00","slug":"202606-%e8%ae%ba%e6%96%87%e7%a0%94%e8%af%bb-physically-informed-spatial-regularization-for-sound-event-localization-and-detection","status":"publish","type":"post","link":"https:\/\/www.15zhi.net\/blog\/202606-%e8%ae%ba%e6%96%87%e7%a0%94%e8%af%bb-physically-informed-spatial-regularization-for-sound-event-localization-and-detection\/","title":{"rendered":"202606 \u8bba\u6587\u7814\u8bfb-Physically Informed Spatial Regularization for Sound Event Localization and Detection"},"content":{"rendered":"\n<p>\u4f5c\u8005\uff1aHaocheng Liu, Diego Di Carlo, Aditya Arie Nugraha, Kazuyoshi Yoshii, Ga\u00ebl Richard, Mathieu Fontaine<\/p>\n\n\n\n<p>\u65f6\u95f4\uff1a2025<\/p>\n\n\n\n<p>\u6765\u6e90\uff1aWASPAA 2025<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u80cc\u666f<\/h2>\n\n\n\n<p>\u58f0\u97f3\u4e8b\u4ef6\u68c0\u6d4b\u4e0e\u5b9a\u4f4d\uff0cSELD\uff0c\u901a\u5e38\u540c\u65f6\u5b8c\u6210\u4e24\u4ef6\u4e8b\uff1a\u5224\u65ad\u5f53\u524d\u65f6\u95f4\u7247\u91cc\u53d1\u751f\u4e86\u4ec0\u4e48\u58f0\u97f3\u4e8b\u4ef6\uff0c\u4ee5\u53ca\u4f30\u8ba1\u5bf9\u5e94\u58f0\u6e90\u7684\u7a7a\u95f4\u65b9\u5411\u3002\u8fd1\u5e74\u6765\uff0c\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5df2\u7ecf\u53ef\u4ee5\u8f83\u597d\u5730\u5b8c\u6210 SELD \u4efb\u52a1\uff0c\u4f46\u5728\u590d\u6742\u623f\u95f4\u58f0\u5b66\u73af\u5883\u4e0b\u4ecd\u7136\u5bb9\u6613\u53d7\u5230\u53cd\u5c04\u3001\u6df7\u54cd\u548c\u8bad\u7ec3\u6570\u636e\u4e0d\u8db3\u7684\u5f71\u54cd\u3002\u4f20\u7edf\u505a\u6cd5\u5f80\u5f80\u4ece\u6570\u636e\u6216\u7f51\u7edc\u7ed3\u6784\u5165\u624b\uff0c\u4f8b\u5982\u589e\u52a0\u623f\u95f4\u54cd\u5e94\u6a21\u62df\u6570\u636e\u3001\u8bbe\u8ba1\u7a7a\u95f4\u97f3\u9891\u7279\u5f81\u3001\u6539\u8fdb CRNN\/Conformer\/Transformer \u7b49\u65f6\u5e8f\u9aa8\u5e72\u3002\u7136\u800c\uff0c\u8fd9\u4e9b\u65b9\u6cd5\u5927\u591a\u53ea\u662f\u8ba9\u6a21\u578b\u4ece\u6570\u636e\u4e2d\u81ea\u5df1\u5b66\u7a7a\u95f4\u89c4\u5f8b\uff0c\u5e76\u6ca1\u6709\u76f4\u63a5\u8981\u6c42\u6a21\u578b\u8f93\u51fa\u7b26\u5408\u771f\u5b9e\u58f0\u4f20\u64ad\u7684\u7269\u7406\u7ed3\u6784\u3002\u672c\u6587\u7684\u51fa\u53d1\u70b9\u662f\uff1a\u5982\u679c\u6a21\u578b\u9884\u6d4b\u4e86\u4e00\u4e2a\u58f0\u6e90\u65b9\u5411\uff0c\u90a3\u4e48\u8fd9\u4e2a\u65b9\u5411\u5bf9\u5e94\u7684\u591a\u901a\u9053\u7a7a\u95f4\u534f\u65b9\u5dee\u7ed3\u6784\u4e5f\u5e94\u8be5\u662f\u7269\u7406\u4e0a\u5408\u7406\u7684\u3002\u4e5f\u5c31\u662f\u8bf4\uff0c\u6a21\u578b\u4e0d\u5e94\u8be5\u53ea\u5728 DOA \u6570\u503c\u4e0a\u63a5\u8fd1\u6807\u7b7e\uff0c\u8fd8\u5e94\u8be5\u5728\u7a7a\u95f4\u7edf\u8ba1\u7ed3\u6784\u4e0a\u63a5\u8fd1\u771f\u5b9e\u58f0\u573a\u3002\u56e0\u6b64\uff0c\u4f5c\u8005\u63d0\u51fa\u5c06\u7269\u7406\u7a7a\u95f4\u5148\u9a8c\u4f5c\u4e3a\u8bad\u7ec3\u9636\u6bb5\u7684\u6b63\u5219\u7ea6\u675f\uff0c\u5f15\u5bfc SELD \u6a21\u578b\u5b66\u4e60\u66f4\u7b26\u5408\u58f0\u5b66\u4f20\u64ad\u89c4\u5f8b\u7684\u8868\u793a\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u6838\u5fc3\u8d21\u732e<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u63d0\u51fa\u4e00\u79cd physically informed spatial regularization \u65b9\u6cd5\uff1a\u5229\u7528\u7269\u7406\u6a21\u578b\u6784\u9020\u7a7a\u95f4\u534f\u65b9\u5dee\u77e9\u9635\uff0c\u7ea6\u675f SELD \u6a21\u578b\u7684\u9884\u6d4b\u7ed3\u679c\u3002<\/li>\n\n\n\n<li>\u5c06\u58f0\u6e90\u65b9\u5411\u4f30\u8ba1\u4e0e\u7a7a\u95f4\u534f\u65b9\u5dee\u77e9\u9635\uff0cSCM\uff0c\u8054\u7cfb\u8d77\u6765\uff0c\u4f7f\u6a21\u578b\u8f93\u51fa\u4e0d\u4ec5\u5728\u6807\u7b7e\u5c42\u9762\u6b63\u786e\uff0c\u4e5f\u5728\u7269\u7406\u7a7a\u95f4\u7ed3\u6784\u4e0a\u66f4\u5408\u7406\u3002<\/li>\n\n\n\n<li>\u8bbe\u8ba1\u4e24\u7c7b SCM \u5148\u9a8c\uff1a\u4e00\u79cd\u57fa\u4e8e\u89e3\u6790 steering vector\uff0c\u53e6\u4e00\u79cd\u57fa\u4e8e\u68c0\u7d22\u5f97\u5230\u7684\u623f\u95f4\u8109\u51b2\u54cd\u5e94\uff0cRIR\u3002<\/li>\n\n\n\n<li>\u5c06 inverse-Wishart log-likelihood \u5f15\u5165\u8bad\u7ec3\u76ee\u6807\uff0c\u7528\u4e8e\u8861\u91cf\u7ecf\u9a8c SCM \u4e0e\u7269\u7406 SCM \u4e4b\u95f4\u7684\u4e00\u81f4\u6027\u3002<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">\u65b9\u6cd5<\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"500\" height=\"265\" src=\"https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-39.png\" alt=\"\" class=\"wp-image-5362\" srcset=\"https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-39.png 500w, https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-39-300x159.png 300w\" sizes=\"auto, (max-width: 500px) 100vw, 500px\" \/><\/figure>\n\n\n\n<p>\u8f93\u5165\u7279\u5f81\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u8bba\u6587\u4f7f\u7528 SALSA-Lite \u7279\u5f81\u4f5c\u4e3a SELD \u6a21\u578b\u8f93\u5165\u3002<\/li>\n\n\n\n<li>SALSA-Lite \u672c\u8eab\u5305\u542b\u591a\u901a\u9053\u7a7a\u95f4\u4fe1\u606f\uff0c\u9002\u5408\u7528\u4e8e\u58f0\u97f3\u4e8b\u4ef6\u68c0\u6d4b\u4e0e\u65b9\u5411\u4f30\u8ba1\u3002<\/li>\n\n\n\n<li>\u57fa\u7840\u4efb\u52a1\uff1a\n<ul class=\"wp-block-list\">\n<li>\u6a21\u578b\u4ecd\u7136\u5b8c\u6210\u5e38\u89c4 SELD \u4efb\u52a1\u3002<\/li>\n\n\n\n<li>\u8f93\u51fa\u5305\u62ec\u58f0\u97f3\u4e8b\u4ef6\u6d3b\u52a8\uff0cSED\uff0c\u4ee5\u53ca\u65b9\u5411\u4f30\u8ba1\uff0cDOA\u3002<\/li>\n\n\n\n<li>\u57fa\u7840\u635f\u5931\u5305\u62ec SED \u7684\u5206\u7c7b\u635f\u5931\u548c DOA \u7684\u56de\u5f52\u635f\u5931\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u65b9\u6cd5\u4e00\uff1aAlgebraic steering-based SCM\n<ul class=\"wp-block-list\">\n<li>\u6839\u636e\u9ea6\u514b\u98ce\u9635\u5217\u51e0\u4f55\u548c\u9884\u6d4b DOA\uff0c\u76f4\u63a5\u7528 steering vector \u6784\u9020\u7406\u8bba SCM\u3002<\/li>\n\n\n\n<li>\u4f18\u70b9\u662f\u5f62\u5f0f\u7b80\u6d01\u3001\u7269\u7406\u542b\u4e49\u660e\u786e\u3001\u4e0d\u4f9d\u8d56\u989d\u5916 RIR \u6570\u636e\u5e93\u3002<\/li>\n\n\n\n<li>\u7f3a\u70b9\u662f\u4e3b\u8981\u63cf\u8ff0\u76f4\u8fbe\u58f0\uff0c\u5bf9\u590d\u6742\u53cd\u5c04\u548c\u6df7\u54cd\u5efa\u6a21\u6709\u9650\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>\u65b9\u6cd5\u4e8c\uff1aRIR-retrieved SCM\n<ul class=\"wp-block-list\">\n<li>\u4ece RIR \u6570\u636e\u5e93\u4e2d\u68c0\u7d22\u4e0e\u5f53\u524d\u9884\u6d4b\u65b9\u5411\u63a5\u8fd1\u7684\u623f\u95f4\u8109\u51b2\u54cd\u5e94\u3002<\/li>\n\n\n\n<li>\u518d\u5229\u7528\u8fd9\u4e9b RIR \u6784\u9020\u66f4\u63a5\u8fd1\u771f\u5b9e\u623f\u95f4\u4f20\u64ad\u7684 SCM\u3002<\/li>\n\n\n\n<li>\u4f18\u70b9\u662f\u53ef\u4ee5\u5305\u542b\u53cd\u5c04\u548c\u6df7\u54cd\u7b49\u590d\u6742\u58f0\u5b66\u56e0\u7d20\u3002<\/li>\n\n\n\n<li>\u7f3a\u70b9\u662f\u4f9d\u8d56 RIR \u6570\u636e\u5e93\u8d28\u91cf\uff0c\u82e5\u6570\u636e\u5e93\u4e0e\u6d4b\u8bd5\u73af\u5883\u4e0d\u5339\u914d\uff0c\u5148\u9a8c\u4e5f\u53ef\u80fd\u5931\u6548\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">\u5b9e\u9a8c<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u6570\u636e\u6765\u6e90\uff1a\n<ul class=\"wp-block-list\">\n<li>\u58f0\u97f3\u4e8b\u4ef6\u6765\u81ea NIGENS\u3002<\/li>\n\n\n\n<li>\u623f\u95f4\u8109\u51b2\u54cd\u5e94\u6765\u81ea TAU-SRIR\u3002<\/li>\n\n\n\n<li>\u901a\u8fc7\u5408\u6210\u65b9\u5f0f\u6784\u9020 SELD \u6570\u636e\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>\u5b9e\u9a8c\u8bbe\u7f6e\uff1a\n<ul class=\"wp-block-list\">\n<li>\u6bd4\u8f83\u666e\u901a SELD baseline \u4e0e\u52a0\u5165\u7269\u7406\u7a7a\u95f4\u6b63\u5219\u540e\u7684\u6a21\u578b\u3002<\/li>\n\n\n\n<li>\u5206\u522b\u6d4b\u8bd5 in-domain \u548c out-of-domain \u623f\u95f4\u6761\u4ef6\u3002<\/li>\n\n\n\n<li>\u989d\u5916\u6d4b\u8bd5 low-data regime\uff0c\u5373\u53ea\u4f7f\u7528\u5c11\u91cf\u8bad\u7ec3\u6570\u636e\u7684\u60c5\u51b5\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"295\" src=\"https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-40-1024x295.png\" alt=\"\" class=\"wp-image-5364\" srcset=\"https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-40-1024x295.png 1024w, https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-40-300x86.png 300w, https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-40-768x221.png 768w, https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-40.png 1300w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p>\u5b9e\u9a8c\u4e3b\u8981\u6bd4\u8f83\u4e09\u79cd\u65b9\u6cd5\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><code>Baseline<\/code>\uff1a\u4e0d\u52a0\u7269\u7406\u6b63\u5219\uff1b<\/li>\n\n\n\n<li><code>IW-Algebraic<\/code>\uff1a\u7528\u89e3\u6790 steering vector \u6784\u9020 SCM\uff1b<\/li>\n\n\n\n<li><code>IW-RIR<\/code>\uff1a\u7528 RIR \u6570\u636e\u5e93\u6784\u9020 SCM\u3002<\/li>\n<\/ul>\n\n\n\n<p>\u4ece Table 1 \u770b\uff0c<code>IW-RIR<\/code> \u5728 in-domain full-data \u548c low-data \u573a\u666f\u4e0b\u6574\u4f53\u6700\u597d\uff0c\u8bf4\u660e\u7269\u7406\u7a7a\u95f4\u6b63\u5219\u786e\u5b9e\u80fd\u63d0\u5347 SELD\uff0c\u5c24\u5176\u5728\u8bad\u7ec3\/\u6d4b\u8bd5\u73af\u5883\u4e00\u81f4\u65f6\u66f4\u6709\u6548\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"250\" src=\"https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-41-1024x250.png\" alt=\"\" class=\"wp-image-5365\" srcset=\"https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-41-1024x250.png 1024w, https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-41-300x73.png 300w, https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-41-768x187.png 768w, https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-41.png 1420w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p>\u5c55\u793a\u591a\u6b21\u5b9e\u9a8c\u7684\u5206\u5e03\u3002\u53ef\u4ee5\u770b\u5230\u6b63\u5219\u5316\u65b9\u6cd5\u6574\u4f53\u6709\u4e00\u5b9a\u6539\u5584\uff0c\u4f46\u63d0\u5347\u5e45\u5ea6\u4e0d\u5927\uff0c\u4e0d\u662f\u538b\u5012\u6027\u4f18\u52bf\uff1bDOA \u76f8\u5173\u6307\u6807\u7684\u6ce2\u52a8\u4e5f\u6bd4\u8f83\u660e\u663e\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"728\" height=\"308\" src=\"https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-42.png\" alt=\"\" class=\"wp-image-5366\" style=\"width:407px;height:auto\" srcset=\"https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-42.png 728w, https:\/\/www.15zhi.net\/blog\/wp-content\/uploads\/2026\/06\/image-42-300x127.png 300w\" sizes=\"auto, (max-width: 728px) 100vw, 728px\" \/><\/figure>\n\n\n\n<p>\u8bf4\u660e\u5728\u4e0d\u540c DOA threshold \u4e0b\uff0c\u7269\u7406\u6b63\u5219\u4f1a\u6539\u53d8 F1 \u548c ER \u7684\u6743\u8861\u3002low-data \u573a\u666f\u4e0b <code>IW-RIR<\/code> \u548c <code>IW-Algebraic<\/code> \u4ecd\u4fdd\u6301\u7ade\u4e89\u529b\uff0c\u8bf4\u660e\u7269\u7406\u5148\u9a8c\u5728\u6570\u636e\u4e0d\u8db3\u65f6\u66f4\u6709\u4ef7\u503c\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u7ed3\u8bba<\/h2>\n\n\n\n<p>\u672c\u6587\u63d0\u51fa\u4e86\u4e00\u79cd\u5c06\u7269\u7406\u7a7a\u95f4\u5148\u9a8c\u5f15\u5165 SELD \u8bad\u7ec3\u8fc7\u7a0b\u7684\u65b9\u6cd5\u3002\u4e0e\u76f4\u63a5\u4fee\u6539\u7f51\u7edc\u7ed3\u6784\u4e0d\u540c\uff0c\u8be5\u65b9\u6cd5\u901a\u8fc7\u7a7a\u95f4\u534f\u65b9\u5dee\u77e9\u9635\u6b63\u5219\u9879\u7ea6\u675f\u6a21\u578b\u8f93\u51fa\uff0c\u4f7f\u9884\u6d4b\u7ed3\u679c\u4e0d\u4ec5\u63a5\u8fd1\u6807\u7b7e\uff0c\u4e5f\u66f4\u7b26\u5408\u58f0\u5b66\u4f20\u64ad\u89c4\u5f8b\u3002\u7531\u4e8e\u6b63\u5219\u9879\u53ea\u5728\u8bad\u7ec3\u65f6\u4f7f\u7528\uff0c\u63a8\u7406\u9636\u6bb5\u4e0d\u4f1a\u589e\u52a0\u989d\u5916\u8ba1\u7b97\u6210\u672c\u3002\u5bf9\u4e8e\u5f53\u524d SELD \u7814\u7a76\u6765\u8bf4\uff0c\u8fd9\u7bc7\u8bba\u6587\u63d0\u4f9b\u4e86\u4e00\u4e2a\u5f88\u597d\u7684\u65b9\u5411\uff1a\u7269\u7406\u5148\u9a8c\u7684\u4ef7\u503c\u4e0d\u4ec5\u53d6\u51b3\u4e8e\u5148\u9a8c\u672c\u8eab\uff0c\u8fd8\u53d6\u51b3\u4e8e\u6ce8\u5165\u4f4d\u7f6e\u3002\u8f93\u5165\u5c42\u3001\u8868\u793a\u5c42\u3001\u6ce8\u610f\u529b\u5c42\u3001\u5377\u79ef\u7279\u5f81\u5c42\u548c\u8bad\u7ec3\u76ee\u6807\u5c42\u90fd\u53ef\u4ee5\u6210\u4e3a\u7269\u7406\u5148\u9a8c\u8fdb\u5165\u6a21\u578b\u7684\u8def\u5f84\u3002\u4e0d\u540c\u8def\u5f84\u53ef\u80fd\u9002\u5408\u4e0d\u540c\u6a21\u578b\u7ed3\u6784\u548c\u6570\u636e\u6761\u4ef6\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u4f5c\u8005\uff1aHaocheng Liu, Diego Di Carlo, Aditya Arie Nugraha, K [&hellip;]<\/p>\n","protected":false},"author":72,"featured_media":0,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-5359","post","type-post","status-publish","format-standard","hentry","category-events"],"_links":{"self":[{"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/posts\/5359","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/users\/72"}],"replies":[{"embeddable":true,"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/comments?post=5359"}],"version-history":[{"count":2,"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/posts\/5359\/revisions"}],"predecessor-version":[{"id":5367,"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/posts\/5359\/revisions\/5367"}],"wp:attachment":[{"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/media?parent=5359"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/categories?post=5359"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.15zhi.net\/blog\/wp-json\/wp\/v2\/tags?post=5359"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}