ACL-OCL / Base_JSON /prefixR /json /rocling /2019.rocling-1.14.json
Benjamin Aw
Add updated pkl file v3
6fa4bc9
{
"paper_id": "2019",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T14:54:23.448243Z"
},
"title": "Recurrent Neural Network-based Microphone Howling Suppression",
"authors": [
{
"first": "Cheng-Yang",
"middle": [],
"last": "\u6797\u653f\u967d",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "National Taipei University of Technology",
"location": {}
},
"email": "chengyang@speech.ntut.edu.tw"
},
{
"first": "Yuan-Fu",
"middle": [],
"last": "Lin",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "National Taipei University of Technology",
"location": {}
},
"email": ""
},
{
"first": "",
"middle": [],
"last": "Liao",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "National Taipei University of Technology",
"location": {}
},
"email": "yfliao@mail.ntut.edu.tw"
},
{
"first": "Chen-Ming",
"middle": [],
"last": "\u6f58\u632f\u9298",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Telecommunication Laboratories",
"location": {
"settlement": "Chunghwa Telecom",
"country": "Taoyuan Taiwan"
}
},
"email": "chenming@cht.com.tw"
},
{
"first": "Tzu-Hsiu",
"middle": [],
"last": "Pan",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Telecommunication Laboratories",
"location": {
"settlement": "Chunghwa Telecom",
"country": "Taoyuan Taiwan"
}
},
"email": ""
},
{
"first": "",
"middle": [],
"last": "Kuo",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Telecommunication Laboratories",
"location": {
"settlement": "Chunghwa Telecom",
"country": "Taoyuan Taiwan"
}
},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "",
"pdf_parse": {
"paper_id": "2019",
"_pdf_hash": "",
"abstract": [],
"body_text": [
{
"text": "e( )= d( ) \u2212\u0302( )u ( ) \u0302 ( + 1)= \u0302 ( )+\u2016 ( )\u2016 2 u ( ) * ( ) \u5728\u983b\u57df NLMS \u6f14\u7b97\u6cd5\u90e8\u5206\uff0c\u6211\u5011\u6bcf 512 \u9ede\u53d6\u4e00\u500b\u97f3\u6846\uff0c\u5c07\u6b64\u97f3\u6846\u6240\u6709\u9ede\u7531 FFT",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "(t) = f (\u2211 v ji x i (t) n i + b j ) (3.1) \u6b64\u6642\u7684\u7b2c\u4e00\u5c64\u96b1\u85cf\u5c64\u8f38\u51fa\u6703\u8f38\u5165\u5230\u7b2c\u4e00\u5c64\u7684\u56de\u994b\u5c64\uff0c\u56de\u6388\u9023\u7d50\u6b0a\u91cdr 1,jm \u518d\u5230\u7b2c\u4e00\u5c64\u96b1\u85cf \u5c64\uff0c\u4e5f\u5c31\u662f\u4e0a\u4e00\u500b\u6642\u9593\u9ede(t-1)\u795e\u7d93\u5143\u7684\u8f49\u63db\u72c0\u614b\uff0c\u4e0b\u5f0f(3.2)\u70ba\u7d50\u5408\u8f38\u5165x i (t)\u6240\u7522\u751f\u7684\u65b0 \u7684\u8f38\u51fa\u65b9\u7a0b\u5f0f\u3002 y j (t) = f (\u2211 v ji x i (t) n i + \u2211 r 1,jm y i (t \u2212 1) i + b j ) (3.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF1": {
"ref_id": "b1",
"title": "Nonliner Acoustic Echo Cancellstion with 2nd Order Adaptive Volterra Filters, IEEE Int",
"authors": [
{
"first": "A",
"middle": [],
"last": "Stenger",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Trautmann",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Rabenstein",
"suffix": ""
}
],
"year": 1999,
"venue": "Conf. on Acoustics, Speech & Signal Procrssing (ICASSP)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Stenger, A., L. Trautmann, and R. Rabenstein. \"Nonliner Acoustic Echo Cancellstion with 2nd Order Adaptive Volterra Filters, IEEE Int.\" Conf. on Acoustics, Speech & Signal Procrssing (ICASSP). 1999.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "The performance study of NLMS algorithm for acoustic echo cancellation",
"authors": [
{
"first": "Ranbeer",
"middle": [],
"last": "Tyagi",
"suffix": ""
},
{
"first": "Roop",
"middle": [],
"last": "Singh",
"suffix": ""
},
{
"first": "Rahul",
"middle": [],
"last": "Tiwari",
"suffix": ""
}
],
"year": 2017,
"venue": "2017 International Conference on Information",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tyagi, Ranbeer, Roop Singh, and Rahul Tiwari. \"The performance study of NLMS algorithm for acoustic echo cancellation.\" 2017 International Conference on Information, Communication, Instrumentation and Control (ICICIC). IEEE, 2017.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Toward bias minimization in acoustic feedback cancellation systems",
"authors": [
{
"first": "Christos",
"middle": [],
"last": "Boukis",
"suffix": ""
},
{
"first": "Danilo",
"middle": [
"P"
],
"last": "Mandic",
"suffix": ""
},
{
"first": "Anthony",
"middle": [
"G"
],
"last": "Constantinides",
"suffix": ""
}
],
"year": 2007,
"venue": "The Journal of the Acoustical Society of America",
"volume": "121",
"issue": "",
"pages": "1529--1537",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Boukis, Christos, Danilo P. Mandic, and Anthony G. Constantinides. \"Toward bias minimization in acoustic feedback cancellation systems.\" The Journal of the Acoustical Society of America 121.3 (2007): 1529-1537.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Prediction-error-method-based adaptive feedback cancellation in hearing aids using pitch estimation",
"authors": [
{
"first": "Kim",
"middle": [],
"last": "Ngo",
"suffix": ""
}
],
"year": 2010,
"venue": "18th European Signal Processing Conference",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ngo, Kim, et al. \"Prediction-error-method-based adaptive feedback cancellation in hearing aids using pitch estimation.\" 2010 18th European Signal Processing Conference. IEEE, 2010.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Adaptive feedback cancellation for audio applications",
"authors": [
{
"first": "Toon",
"middle": [],
"last": "Van Waterschoot",
"suffix": ""
},
{
"first": "Marc",
"middle": [],
"last": "Moonen",
"suffix": ""
}
],
"year": 2009,
"venue": "Signal Processing",
"volume": "89",
"issue": "",
"pages": "2185--2201",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Van Waterschoot, Toon, and Marc Moonen. \"Adaptive feedback cancellation for audio applications.\" Signal Processing 89.11 (2009): 2185-2201.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Adaptive feedback canceller with howling detection for hearing aids",
"authors": [
{
"first": "Kakeru",
"middle": [],
"last": "Kashima",
"suffix": ""
}
],
"year": 2015,
"venue": "Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kashima, Kakeru, et al. \"Adaptive feedback canceller with howling detection for hearing aids.\" 2015 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA). IEEE, 2015.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Room impulse response generator",
"authors": [
{
"first": "Emanuel",
"middle": [],
"last": "Habets",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Ap",
"suffix": ""
}
],
"year": 2006,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Habets, Emanuel AP. \"Room impulse response generator.\" Technische Universiteit Eindhoven, Tech. Rep 2.2.4 (2006): 1. Available: https://github.com/ehabets/RIR- Generator [Accessed: Jul. 15, 2019]",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "\u7531\u4e0a\u5c64\u96b1\u85cf\u5c64\u8f38\u51fay j (t)\u6703\u8f38\u5165\u7b2c\u4e8c\u5c64\u7684\u96b1\u85cf\u5c64\uff0c\u6b0a\u91cdw kj \u9023\u63a5\u7b2c\u4e8c\u5c64\u96b1\u85cf\u5c64\uff0c\u52a0\u4e0a\u7b2c\u4e8c \u5c64\u96b1\u85cf\u5c64\u504f\u58d3\u503c bk\uff0c\u7d93\u8f49\u63db\u51fd\u6578f(\u22c5)\u8f49\u63db\uff0c\u7522\u51fa\u5728\u7b2c\u4e8c\u500b\u96b1\u85cf\u5c64\u9019\u4e00\u500b\u6642\u9593\u9ede\u7684\u795e\u7d93\u5143 \u8f38\u51fay k (t)\uff0c\u5176\u8f38\u51fa\u5982\u4e0b\u65b9\u7a0b\u5f0f(3.3)\u3002 y k (t) = f (\u2211 w kj y j (t) n j + b k ) (3.3) \u6b64\u6642\u7684\u8f38\u51fa\u4e00\u6a23\u6703\u8f38\u5165\u5230\u7b2c\u4e8c\u5c64\u56de\u994b\u5c64\uff0c\u900f\u904e\u56de\u6388\u9023\u63a5\u6b0a\u91cdr 2,kg \u7b2c\u4e8c\u5c64\u96b1\u85cf\u5c64\uff0c\u4e5f\u662f\u4e0a \u4e00\u500b\u500b\u6642\u9593\u9ede(t-1)\u795e\u7d93\u5143\u8f49\u63db\u72c0\u614b\uff0c\u4e0b\u5f0f(3.4)\u70ba\u7d50\u5408\u8f38\u5165y j (t)\u6240\u7522\u751f\u7684\u65b0\u7684\u8f38\u51fa\u65b9\u7a0b\u5f0f\u3002 y k (t) = f (\u2211 w kj y j (t)"
},
"TABREF0": {
"content": "<table><tr><td>The 2019 Conference on Computational Linguistics and Speech Processing</td></tr><tr><td>ROCLING 2019, pp. 22-36</td></tr><tr><td>\u00a9The Association for Computational Linguistics and Chinese Language Processing</td></tr><tr><td>\u8907\u96dc\u7684\u7a7a\u9593\u74b0\u5883\u97ff\u61c9\uff0c\u6703\u6709\u56b4\u91cd\u7684\u975e\u7dda\u6027\u5931\u771f\uff0c\u4f7f\u5f97\u7dda\u6027\u6f14\u7b97\u6cd5\u7684 NLMS \u7121\u6cd5\u6709\u6548\u7684 \u70ba\u4e86\u6709\u5225\u65bc\u904e\u53bb\u65b9\u6cd5\uff0c\u6211\u5011\u5e0c\u671b\u53ef\u4ee5\u5728\u562f\u53eb\u767c\u751f\u524d\u5c31\u5c07\u562f\u53eb\u6e05\u9664\uff0c\u56e0\u6b64\u4f7f\u7528\u4e86\u9069\u61c9\u6027\u6ffe</td></tr><tr><td>\u89e3\u6c7a\u975e\u7dda\u6027\u554f\u984c\uff0c\u9032\u800c\u63d0\u51fa\u4e86\u57fa\u65bc RNN \u7684\u9032\u968e\u6f14\u7b97\u6cd5\u3002\u7531\u65bc RNN \u662f\u64c1\u6709\u56de\u6388\u529f\u80fd\u7684 \u6ce2\u5668\uff0c\u5728\u562f\u53eb\u767c\u751f\u524d\u5148\u8b93\u6ffe\u6ce2\u5668\u5b78\u7fd2\u74b0\u5883\u97ff\u61c9\u7684\u8def\u5f91\u800c\u6539\u8b8a\u5176\u6b0a\u91cd\uff0c\u7a69\u5b9a\u7684\u6d88\u9664\u591a\u51fa\u4f86</td></tr><tr><td>\u905e\u8ff4\u985e\u795e\u7d93\u7db2\u8def\uff0c\u80fd\u5c07\u4e0a\u500b\u904e\u53bb\u7684\u6642\u9593\u4e4b\u8f38\u51fa\u503c\u5132\u5b58\u4e0b\u4f86\uff0c\u518d\u91cd\u65b0\u5c0e\u56de\u5230\u8f38\u5165\u7aef\uff0c\u4f7f\u5f97 \u7684\u56de\u8072\uff0c\u5176\u4e2d\u6700\u5c0f\u5747\u65b9\u6f14\u7b97\u6cd5(LMS)\u662f\u6700\u6613\u5be6\u73fe\u3001\u4e14\u7a69\u5b9a\u53ca\u8a08\u7b97\u91cf\u5c0f[4]\uff0c\u6642\u81f3\u4eca\u65e5\u4ecd\u53d7</td></tr><tr><td>\u7cfb\u7d71\u80fd\u5920\u6293\u53d6\u9577\u5ea6\u8f03\u9577\u7684\u6642\u9593\u8f38\u5165\u8a0a\u865f\uff0c\u8b93\u7cfb\u7d71\u64c1\u6709\u9f90\u5927\u7684\u904e\u53bb\u8cc7\u6599\u4f86\u5b78\u7fd2\u74b0\u5883\u97ff\u61c9\u7684 \u5230\u8a31\u591a\u4eba\u559c\u611b\u4e14\u5ee3\u6cdb\u5730\u904b\u7528\uff0c\u540c\u6642\u70ba\u4e86\u89e3\u6c7a\u7cfb\u7d71\u6536\u6582\u7de9\u6162\u7684\u7f3a\u9ede\u3002\u4fbf\u5c07\u8f38\u5165\u8a0a\u865f\u6b63\u898f\u5316\uff0c</td></tr><tr><td>\u8def\u5f91\uff0c\u6539\u5584\u975e\u7dda\u6027\u7684\u90e8\u5206\u3002 \u800c\u6f14\u8b8a\u6210 NLMS \u6f14\u7b97\u6cd5\uff0c\u5176\u63a1\u7528\u53ef\u8b8a\u6b65\u9577\u7684\u65b9\u6cd5\u4f86\u7a69\u5b9a\u6536\u6582\u904e\u7a0b\u3002</td></tr><tr><td>\u6b64\u5916\u562f\u53eb\u7684\u767c\u751f\u90fd\u662f\u5373\u6642\u4e14\u7a81\u7136\u591a\u8b8a\u7684\uff0c\u800c\u6642\u57df\u7684\u6bcf\u4e00\u9ede\u8abf\u8b8a\u4e00\u6b21\uff0c\u5728\u9762\u5c0d\u7a81\u7136\u8b8a\u5316\u7684 \u53e6\u5916\u7531\u65bc\u9589\u9396\u8ff4\u8def\u7684\u95dc\u4fc2\uff0c\u8fd1\u7aef\u8a0a\u865f\u548c\u63da\u8072\u5668\u4e4b\u9593\u6709\u56b4\u91cd\u7684\u76f8\u95dc\u6027\u554f\u984c\uff0c\u9020\u6210\u562f\u53eb\u66f4\u5bb9 \u5716\u4e00\u3001\u9ea5\u514b\u98a8\u6b63\u56de\u6388\u562f\u53eb \u97f3\u6a02\u6216\u983b\u7387\u8b8a\u5316\uff0c\u4e0d\u78ba\u5b9a\u662f\u5426\u4ecd\u80fd\u6709\u6548\u7684\u6536\u6582\u8207\u6d88\u9664\uff0c\u4f46\u5728\u983b\u57df\u505a\u6f14\u7b97\u6cd5\u53ef\u9867\u53ca\u5230\u4e0d\u540c \u6613\u767c\u751f\uff0c\u56e0\u6b64\u5f8c\u9762\u6709\u95dc\u65bc\u56de\u8072\u6d88\u9664\u7684\u554f\u984c\u4fbf\u6709\u90e8\u5206\u4eba\u8457\u91cd\u5728\u53bb\u76f8\u95dc\u6027\u4e0a\u9762\uff0c\u6709\u4eba\u5728\u9589\u9396</td></tr><tr><td>\u983b\u6bb5\u7684\u74b0\u5883\u8a0a\u865f\uff0c\u56e0\u6b64\u6211\u5011\u4e5f\u91dd\u5c0d\u6b64\u9ede\u505a\u4e86\u6642\u57df\u3001\u983b\u57df\u6f14\u7b97\u6cd5\u7684\u6bd4\u8f03\uff0c\u89c0\u5bdf\u6642\u57df\u8207\u983b\u57df \u8ff4\u8def\u4e2d\u7684\u524d\u5411\u8def\u5f91\u52a0\u5165\u5168\u901a\u6ffe\u6ce2\u5668\u4f86\u964d\u4f4e\u76f8\u95dc\u6027[5]\uff0c\u6b64\u5916\u4e5f\u6709\u8ad6\u6587\u63d0\u51fa\u4e86\u9810\u6e2c\u8aa4\u5dee\u65b9</td></tr><tr><td>\u5728\u7d30\u81a9\u5ea6\u8207\u6536\u6582\u901f\u5ea6\u4e0a\u662f\u5426\u6709\u8457\u660e\u986f\u5dee\u7570\u3002\u53e6\u5916\u4e5f\u56e0 RNN \u5728\u8a08\u7b97\u91cf\u8207 NLMS \u6709\u8457\u660e\u986f \u6cd5\u9069\u61c9\u6027\u6ffe\u6ce2\u5668(PEM AFC)\u4f86\u6e1b\u5c11\u8fd1\u7aef\u8a0a\u865f\u8207\u63da\u8072\u5668\u4e4b\u9593\u76f8\u95dc\u6027[6] \u3002\u800c\u5728 PEM \u65b9\u6cd5\u4e2d\uff0c \u70ba\u89e3\u6c7a\u9019\u4ef6\u4e8b\u60c5\uff0c\u53ef\u91dd\u5c0d\u786c\u9ad4\u4e0a\u9ea5\u514b\u98a8\u7684\u6307\u5411\u6027\u6536\u97f3\uff0c\u5c07\u7279\u5b9a\u65b9\u5411\u7684\u8072\u97f3\u6536\u9032\u9ea5\u514b\u98a8\uff0c \u7684\u5dee\u8ddd\uff0c\u6211\u5011\u4e5f\u89c0\u5bdf\u5176\u8a08\u7b97\u91cf\u662f\u5426\u6709\u8207\u5176\u6548\u679c\u6210\u6b63\u6bd4\u3002\u56e0\u6b64\u57fa\u65bc\u4e0a\u8ff0\u8003\u91cf\uff0c\u65bc\u672c\u7bc7\u8ad6\u6587 \u662f\u5229\u7528\u53cd\u5411\u8fd1\u7aef\u8a0a\u865f\u6a21\u578b\u5c0d\u9ea5\u514b\u98a8\u548c\u5587\u53ed\u9032\u884c\u9810\u6ffe\u6ce2\uff0c\u7136\u5f8c\u5c07\u9019\u4e9b\u8a0a\u865f\u9001\u81f3\u81ea\u9069\u61c9\u6ffe\u6ce2 \u5716\u56db\u3001\u57fa\u65bc\u905e\u8ff4\u985e\u795e\u7d93\u7db2\u8def\u4e4b\u9069\u61c9\u6027\u6ffe\u6ce2\u5668\u6291\u5236\u562f\u53eb\u7cfb\u7d71\u67b6\u69cb\u5716</td></tr><tr><td>\u76e1\u91cf\u4e0d\u8b93\u9664\u4e86\u8fd1\u7aef\u4eba\u8072\u5916\u7684\u8072\u97f3\u6536\u5165\uff0c\u4ea6\u6216\u8005\u662f\u589e\u52a0\u9ea5\u514b\u98a8\u4e4b\u6536\u97f3\u654f\u611f\u5ea6\uff0c\u53ef\u6e1b\u5c11\u975e\u5fc5 \u8981\u6536\u5165\u9ea5\u514b\u98a8\u7684\u8072\u97f3\uff1b\u5728\u6f14\u7b97\u6cd5\u4e0a\u6709\u4f7f\u7528\u79fb\u983b\u6253\u65b7\u5171\u632f\uff0c\u5c07\u8a0a\u865f\u983b\u7387\u505a\u4e9b\u8a31\u5347\u983b\u6216\u964d\u983b\uff0c \u7834\u58de\u4e86\u562f\u53eb\u7684\u767c\u751f\u689d\u4ef6\uff0c\u9032\u800c\u6291\u5236\u4e86\u562f\u53eb\uff0c\u6216\u662f\u4f7f\u7528\u5e36\u963b\u6ffe\u6ce2\u5668\u5c07\u6703\u767c\u751f\u562f\u53eb\u7684\u983b\u6bb5\u505a \u8870\u6e1b\uff0c\u5982\u679c\u8870\u6e1b\u9019\u4e9b\u904e\u5f37\u7684\u983b\u7387\u5c31\u80fd\u6291\u5236\u4f4f\u562f\u53eb\uff0c\u4f46\u6b64\u5169\u8005\u96d6\u7136\u53ea\u5c0d\u5c0f\u7bc4\u570d\u7684\u983b\u7387\u505a\u4e86 \u8abf\u6574\uff0c\u4f46\u4ecd\u6703\u7834\u58de\u539f\u59cb\u8a0a\u865f\u8072\u97f3\uff0c\u751a\u81f3\u662f\u4eba\u8033\u53ef\u807d\u51fa\u7684\u5dee\u5225\uff0c\u4e14\u562f\u53eb\u4ecd\u662f\u6642\u4e0d\u6642\u7684\u767c\u751f\uff0c \u56e0\u800c\u76ee\u524d\u5728\u5be6\u969b\u63a7\u5236\u562f\u53eb\u767c\u751f\u7684\u4f5c\u6cd5\u4ecd\u662f\u6cbb\u6a19\u4e0d\u6cbb\u672c\u3002 \u4e2d\u6211\u5011\u5c07\u63d0\u51fa\u905e\u8ff4\u985e\u795e\u7d93\u7db2\u8def\u9ea5\u514b\u98a8\u6291\u5236\u562f\u53eb\u7cfb\u7d71\uff0c\u6bd4\u8f03\u50b3\u7d71\u6642\u57df\u3001\u983b\u57df NLMS \u548c\u6642 \u57df RNN \u5728\u4e0d\u540c\u66f2\u98a8\u3001\u4e0d\u540c\u74b0\u5883\u7a7a\u9593\u97ff\u61c9\u60c5\u6cc1\u4e0b\uff0c\u4e0d\u540c\u6f14\u7b97\u6cd5\u7684\u6536\u6582\u901f\u5ea6\u3001\u8a08\u7b97\u91cf\u9700\u6c42 \u8207\u562f\u53eb\u6291\u5236\u6548\u679c\uff0c\u5728\u4ee5\u4e4b\u5f8c\u7ae0\u7bc0\u5c07\u6703\u4ecb\u7d39\u5176\u6a21\u64ec\u65b9\u6cd5\u8207\u67b6\u69cb\u3002 \u4e8c\u3001\u76f8\u95dc\u7814\u7a76 \u6f14\u7b97\u6cd5\u4e2d\uff0c\u6b64\u4fbf\u9054\u5230\u4e86\u964d\u4f4e\u76f8\u95dc\u6027\u7684\u76ee\u6a19\u3002\u800c\u5c0d\u65bc\u8fd1\u7aef\u8a9e\u97f3\u8a0a\u865f\uff0c\u901a\u5e38\u4f7f\u7528\u7dda\u6027\u9810\u4f30 (\u4e00) NLMS (Linear Prediction\uff0cLP)\u4f86\u4f30\u8a08[7]\uff0c\u8a9e\u97f3\u8a0a\u865f\u7531\u65bc\u6642\u9593\u76f8\u8fd1\u7684\u8a0a\u865f\u9ede\u5f7c\u6b64\u6709\u76f8\u95dc\u6027\uff0c\u6bcf\u500b \u5176\u6a21\u64ec\u70ba\u5229\u7528\u9ea5\u514b\u98a8\u6240\u6536\u96c6\u5230\u7684\u8072\u97f3 d \u8207\u900f\u904e\u5587\u53ed\u8f38\u51fa\u7684\u8072\u97f3 x\uff0c\u5229\u7528 NLMS \u9810\u6e2c\u53ef \u8a0a\u865f\u9ede\u53ef\u7531\u76f8\u8fd1\u7684\u8a0a\u865f\u9ede\u85c9\u7531\u7dda\u6027\u7d44\u5408\u52a0\u4ee5\u903c\u8fd1\u6216\u4f30\u6e2c\u3002\u6545\u53ef\u85c9\u7531 LPC \u4f30\u8a08\u53bb\u76f8\u95dc\u9810 \u80fd\u9304\u5230\u7684\u74b0\u5883\u97ff\u61c9\uff0c\u4f7f\u5176\u76f8\u6e1b\u5f8c\u5c07\u56de\u8072\u6d88\u9664\uff0c\u6b64\u6642\u7cfb\u7d71\u7684\u8aa4\u5dee\u8a0a\u865f e \u70ba\u8f38\u51fa\u5f97\u5230\u7684\u6e05\u6670 \u8a13\u7df4\u6ffe\u6ce2\uff0c\u5c07\u8a9e\u97f3\u8a0a\u865f\u4e2d\u5206\u96e2\u6210\u53e3\u8154\u8207\u8072\u5e36\u8a0a\u865f\uff0c\u4ee5\u53bb\u9664\u5587\u53ed\u8f38\u51fa\u8a0a\u865f\u8207\u4eba\u8072\u8a0a\u865f\u5404\u81ea \u7684\u76f8\u95dc\u6027\u3002 \u8a9e\u8005\u8072\u97f3\u3002\u4e0b\u5217\u70ba NLMS \u6f14\u7b97\u6cd5:</td></tr><tr><td>\u5e38\u7528\u7684\u56de\u8072\u6d88\u9664\u65b9\u6cd5\u5305\u62ec\u904e\u53bb\u7684\u5e36\u963b\u6ffe\u6ce2\u5668\u3001\u79fb\u983b\uff0c\u5230\u8fd1\u5e74\u4e3b\u6d41\u7684 NLMS \u548c\u57fa\u65bc\u9810\u6e2c \u9810\u6e2c\u8aa4\u5dee\u65b9\u6cd5\u6b21\u9069\u61c9\u6027\u6ffe\u6ce2\u5668\u80fd\u5920\u5f9e\u671f\u671b\u8a0a\u865f\u4e2d\u53bb\u9664\u76f8\u95dc\u5206\u91cf\uff0c\u56e0\u6b64\u8a72\u53cd\u994b\u6d88\u9664\u5668\u5c0d\u4e0d \u70ba\u4e86\u6539\u5584\u904e\u53bb\u6291\u5236\u562f\u53eb\u7684\u7f3a\u9ede\uff0c\u9019\u88e1\u6211\u5011\u4f7f\u7528\u9069\u61c9\u6027\u6ffe\u6ce2\u5668\u4f86\u63d0\u524d\u6d88\u9664\u56de\u8072\u4ee5\u6291\u5236\u562f\u53eb\uff0c \u8aa4\u5dee\u65b9\u6cd5\u7684\u9069\u61c9\u6027\u6ffe\u6ce2\u5668(Prediction Error Method-based Adaptive Feedback Cancellation\uff0c \u76f8\u95dc\u4fe1\u865f\u548c\u593e\u5e36(entrainment)\u80fd\u5920\u4f7f\u4e4b\u4e0d\u53cd\u61c9\uff0c\u4f46\u4e0d\u5e78\u5730\uff0c\u7576\u671f\u671b\u8a0a\u865f\u662f\u9031\u671f\u6027\u865f\u6642\uff0c \u9996\u5148\u6211\u5011\u5148\u4ecb\u7d39\u57fa\u65bc NLMS \u4e4b\u9069\u61c9\u6027\u6ffe\u6ce2\u5668\u6f14\u7b97\u6cd5\u7684\u56de\u8072\u6d88\u9664\u7cfb\u7d71[1-2]\u5982\u4e0b\u5716\u4e8c\uff0c\u5229 PEM AFC)\u3002\u904e\u53bb\u5728\u4f7f\u7528\u79fb\u983b\u662f\u4f7f\u7528\u4e00\u7a2e\u53ef\u4ee5\u6539\u8b8a\u8072\u97f3\u983b\u7387\u7684\u8a2d\u5099\u79fb\u983b\u5668\uff0c\u5176\u5de5\u4f5c\u539f\u7406 \u9810\u6e2c\u8aa4\u5dee\u65b9\u6cd5\u81ea\u9069\u61c9\u6ffe\u6ce2\u5668\u5c07\u6703\u7d66\u51fa\u96f6\uff0c\u5018\u82e5\u662f\u6b64\u7a2e\u60c5\u5f62\u562f\u53eb\u4fbf\u4e0d\u80fd\u88ab\u6d88\u9664\u4e86\uff0c\u56e0\u70ba\u562f \u7528\u64f4\u5927\u6a5f\u5587\u53ed\u7684\u8f38\u5165\u97f3\u6e90\u4f5c\u70ba\u53c3\u8003\u8a0a\u865f x\uff0c\u81ea\u52d5\u4f30\u7b97\u5728\u4e0d\u540c\u7a7a\u9593\u74b0\u5883\u3001\u4e0d\u540c\u6b4c\u66f2\u3001\u4e0d\u540c \u985e\u4f3c\u8b8a\u8abf\u5668\uff0c\u80fd\u5920\u5c07\u8072\u97f3\u8a0a\u865f\u589e\u52a0\u3001\u6e1b\u5c11 5Hz\uff0c\u7834\u58de\u4e86\u562f\u53eb\u767c\u751f\u7684\u689d\u4ef6\uff0c\u9032\u800c\u6291\u5236\u4e86\u562f \u53eb\u4e5f\u662f\u6709\u9031\u671f\u6027\u7684\uff0c\u56e0\u6b64\u82e5\u767c\u751f\u6b64\u7a2e\u60c5\u5f62\u6216\u8a31\u9700\u8981\u56de\u982d\u4f9d\u9760\u50b3\u7d71\u7cfb\u7d71\uff0c\u56e0\u800c\u4e5f\u6709\u4eba\u5c0d\u6b64 \u8a0a\u96dc\u6bd4\u4e0b\uff0c\u9ea5\u514b\u98a8\u53ef\u80fd\u9304\u5230\u7684\u56de\u6388\u8a0a\u865f\uff0c\u518d\u5c07\u9810\u6e2c\u51fa\u4e4b\u671f\u671b\u4fe1\u865f\u8207\u8f38\u5165\u7684\u9ea5\u514b\u98a8\u8a0a\u865f d \u5c0d\u4eba\u8033\u5df2\u7d93\u6709\u660e\u986f\u7684\u611f\u89ba\u4e86\uff1b\u6b64\u5916\u5e36\u7d44\u6ffe\u6ce2\u5668\u81f3\u4eca\u5728\u61c9\u7528\u4e0a\u4ecd\u662f\u4e3b\u6d41\u6291\u5236\u562f\u53eb\u7684\u5de5\u5177\uff0c \u76f8\u6e1b\uff0c\u4f7f\u56de\u8072\u8a0a\u865f\u589e\u76ca\u524d\u5c31\u5c07\u5176\u6d88\u9664\uff0c\u76f4\u63a5\u5f9e\u6e90\u982d\u6d88\u9664\u562f\u53eb\u767c\u751f\u7684\u53ef\u80fd\u6027\u3002 \u53eb\uff0c\u96d6\u7136\u5c0d\u8072\u97f3\u7684\u7834\u58de\u5f88\u5c0f\uff0c\u4f46\u5176\u5728\u6f14\u5531\u548c\u6a02\u5668\u4e2d\u5c31\u6703\u6709\u660e\u986f\u5dee\u7570\uff0c\u5149 5Hz \u7684\u97f3\u8abf\u8b8a\u5316 \u73fe\u8c61\u505a\u51fa\u4e86\u562f\u53eb\u5075\u6e2c\u5668\u5c0d\u6b64\u60c5\u6cc1\u4f5c\u51fa\u50b3\u7d71\u548c PEM \u767c\u6cd5\u7684\u5207\u63db[8]\u3002</td></tr><tr><td>\u7a7a\u9593\u97ff\u61c9\u60c5\u6cc1\u4e0b\uff0c\u4e0d\u540c\u6f14\u7b97\u6cd5\u7684\u6536\u6582\u901f\u5ea6\u3001\u8a08\u7b97\u91cf\u9700\u6c42\u8207\u562f\u53eb\u6291\u5236\u6548\u679c\u3002\u7531\u5be6\u9a57\u7d50\u679c\u767c \u73fe\uff1a (1)\u5728 \u6642\u57df\u5be6\u73fe\u6536\u6582\u6bd4\u8f03\u5feb\uff0c (2)\u5728\u983b\u7387\u53ef\u5be6\u73fe\u8a08\u7b97\u91cf\u5c0f\u65bc\u6642\u57df\uff0c( 3)RNN \u5728\u6536\u6582 \u5728\u97f3\u97ff\u7cfb\u7d71\u4e2d\u51fa\u73fe\u562f\u53eb\u662f\u7531\u65bc\u6b63\u53cd\u994b\u4f7f\u97f3\u983b\u4fe1\u865f\u4e2d\u7684\u67d0\u4e9b\u983b\u7387\u9ede\u4e0d\u65b7\u88ab\u52a0\u5f37\u800c\u9020\u6210\u7684\uff0c \u56e0\u6b64\u5c07\u9019\u4e9b\u983b\u7387\u9ede\u5207\u9664\u6216\u9032\u884c\u5927\u5e45\u5ea6\u8870\u6e1b\uff0c\u5c31\u53ef\u4ee5\u6709\u6548\u6291\u5236\u8072\u53cd\u994b\uff0c\u5176\u539f\u7406\u5982\u5716\u4e09\uff0c\u4e3b \u4e09\u3001\u9069\u61c9\u6027\u6ffe\u6ce2\u5668\u6291\u5236\u562f\u53eb\u7cfb\u7d71</td></tr><tr><td>\u901f\u5ea6\u53ca\u7a81\u7136\u8b8a\u5316\u7684\u983b\u7387\u6d88\u9664\u4e0a\u512a\u65bc NLMS\u3002 \u8981\u662f\u5229\u7528\u6a5f\u5668\u5feb\u901f\u6383\u63cf\u5c0b\u627e\u51fa\u767c\u751f\u562f\u53eb\u7684\u983b\u6bb5\uff0c\u4e26\u81ea\u52d5\u751f\u6210\u4e00\u7d44\u8207\u9019\u4e9b\u562f\u53eb\u983b\u7387\u76f8\u540c\u7684 \u7531\u65bc\u9ea5\u514b\u98a8\u8207\u5587\u53ed\u7684\u9589\u9396\u8ff4\u8def\u6703\u5728\u5ba4\u5167\u53cd\u5c04\u9020\u6210\u56de\u8072\uff0c\u800c \u56de\u8072\u6642\u9593\u6703\u53d7\u623f\u9593\u5927\u5c0f\u5f71\u97ff\uff0c</td></tr><tr><td>\u7a84\u5e36\u6ffe\u6ce2\u5668\u4f86\u5207\u9664\u562f\u53eb\u983b\u7387\uff0c\u9032\u800c\u9054\u5230\u6d88\u9664\u56de\u8072\u800c\u6291\u5236\u562f\u53eb[3]\u3002 \u4f46\u50b3\u7d71\u56de\u8072\u6d88\u9664\u7cfb\u7d71\u6703\u88ab\u8f38\u5165\u7684\u9577\u5ea6\u56fa\u5b9a\u800c\u9650\u5236\u4f4f\uff0c\u5018\u82e5\u8f38\u5165\u6642\u9593\u904e\u9577\uff0c\u5c07\u5c0e\u81f4\u8a08\u7b97\u91cf \u95dc\u9375\u8a5e\uff1a NLMS\u3001\u905e\u8ff4\u985e\u795e\u7d93\u7db2\u8def RNN\u3001\u9069\u61c9\u6027\u6ffe\u6ce2\u5668\u3001\u9ea5\u514b\u98a8\u562f\u53eb \u904e\u65bc\u9f90\u5927\u800c\u6536\u6582\u8f03\u6162\uff0c\u56e0\u6b64\u5e38\u7121\u6cd5\u6709\u6548\u5730\u63d0\u53d6\u4f7f\u7528\u8005\u9577\u6642\u9593\u7684\u8a0a\u865f\uff0c\u6545 \u672c\u8ad6\u6587\u6539\u7528\u905e\u8ff4</td></tr><tr><td>\u985e\u795e\u7d93\u7db2\u8def\uff0c\u5176\u67b6\u69cb\u5716\u5982\u5716\u56db\uff0c\u5c07\u53ef\u4ee5\u7531\u56de\u6388\u8def\u7dda\u770b\u5230\u904e\u53bb\u8f03\u9577\u6642\u9593\u7684\u8cc7\u6599\uff0c\u4ee5\u5b78\u7fd2\u8f03 \u4e00\u3001\u7c21\u4ecb \u9577\u6642\u9593\u7684\u74b0\u5883\u97ff\u61c9\u8def\u5f91\uff0c\u4e26\u63a8\u6e2c\u4e0b\u4e00\u6642\u9593\u9ede\u7684\u8072\u97f3\uff0c\u4e14\u50c5\u7528\u55ae\u5c64\u7684 RNN \u4fbf\u8db3\u4ee5\u7372\u53d6\u8db3</td></tr><tr><td>\u5728\u4f7f\u7528\u9ea5\u514b\u98a8\u5531\u5361\u62c9 OK \u7684\u7cfb\u7d71\u4e2d\uff0c\u7531\u65bc\u6536\u97f3\u7684\u5587\u53ed\u8207\u9ea5\u514b\u98a8\u4e26\u672a\u9694\u96e2\u5728\u4e0d\u540c\u5340\u57df\uff0c\u7576 \u5920\u7684\u9577\u6642\u9593\u8cc7\u8a0a\uff0c\u56e0\u6b64\u591a\u5c64\u7684 RNN \u5c31\u80fd\u770b\u5f97\u66f4\u6df1\u66f4\u5ee3\u6cdb\uff0c\u6293\u53d6\u66f4\u591a\u7684\u8a0a\u606f\u3002\u4e0b\u5217\u5148\u8aaa</td></tr><tr><td>\u5587\u53ed\u767c\u51fa\u4e4b\u8072\u97f3\u900f\u904e\u7a7a\u9593\u50b3\u5230\u9ea5\u514b\u98a8\uff0c\u7531\u65bc\u653e\u5927\u96fb\u8def\u589e\u76ca\u904e\u9ad8\u800c\u5c0e\u81f4\u6b63\u56de\u6388\u53cd\u994b\u5982\u5716\u4e00\uff0c \u660e\u6642\u57df\u8207\u983b\u57df\u7684 NLMS \u505a\u6cd5\uff0c\u5f8c\u9762\u518d\u4ecb\u7d39\u905e\u8ff4\u985e\u795e\u7d93\u7db2\u8def\u7684\u9069\u61c9\u6027\u6ffe\u6ce2\u5668\u6d88\u9664\u6f14\u7b97\u6cd5\u3002</td></tr><tr><td>\u4e0d\u65b7\u5c07\u653e\u51fa\u4e4b\u8072\u97f3\u91cd\u8907\u6536\u5165\u9032\u800c\u767c\u751f\u562f\u53eb\u3002\u6b64\u7a2e\u9589\u9396\u8ff4\u8def\u7684\u562f\u53eb\u4e3b\u8981\u539f\u56e0\u70ba\u6574\u500b\u96fb\u8def\u8207</td></tr><tr><td>\u74b0\u5883\u5c0d\u67d0\u4e9b\u5171\u632f\u983b\u7387\u7684\u589e\u76ca\u904e\u5927\uff0c\u7576\u63d0\u5347\u5587\u53ed\u901a\u9053\u4e4b\u589e\u76ca\u6642\uff0c\u9019\u4e9b\u589e\u76ca\u904e\u5927\u7684\u5171\u632f\u983b\u7387</td></tr><tr><td>\u5148\u9054\u5230\u8072\u5b78\u53cd\u994b\u6240\u9700\u7684\u5f37\u5ea6\u689d\u4ef6\uff0c\u82e5\u6b64\u983b\u7387\u7684\u53cd\u994b\u985e\u578b\u525b\u597d\u70ba\u6b63\u53cd\u994b\uff0c\u5247\u5fc5\u5b9a\u5728\u6b64\u983b\u7387</td></tr><tr><td>\u4e0a\u7522\u751f\u81ea\u6fc0\u9707\u76ea\u73fe\u8c61\uff0c\u4fbf\u662f\u6211\u5011\u6240\u8aaa\u7684\u562f\u53eb\u3002 \u5716\u4e8c\u3001\u50b3\u7d71\u8072\u5b78\u56de\u8072\u6d88\u9664\u7cfb\u7d71\u67b6\u69cb\u5716</td></tr><tr><td>\u7136\u800c\u5728\u4f7f\u7528\u5361\u62c9 OK \u9ea5\u514b\u98a8\u6642\uff0c\u6642\u5e38\u56e0\u6709\u80fd\u91cf\u975e\u5e38\u5927\u7684\u64f4\u5927\u6a5f\u5587\u53ed\u64a5\u51fa\u4e4b\u8072\u97f3\uff0c\u8207\u5ba4\u5167 \u5716\u4e09\u3001\u5e36\u7d44\u6ffe\u6ce2\u5668\u67b6\u69cb</td></tr></table>",
"num": null,
"type_str": "table",
"html": null,
"text": "\u6458\u8981 \u5728\u4f7f\u7528\u5361\u62c9 OK \u7cfb\u7d71\u5531\u6b4c\u6642\uff0c\u5e38\u6703\u56e0\u9ea5\u514b\u98a8\u62ff\u96e2\u5587\u53ed\u592a\u8fd1\uff0c\u6216\u662f\u64f4\u5927\u6a5f\u529f\u7387\u958b\u592a\u5927\uff0c\u7522 \u751f\u6b63\u56de\u6388\u800c\u5c0e\u81f4\u562f\u53eb\uff0c\u9020\u6210\u6b4c\u8005\u8ddf\u807d\u773e\u90fd\u975e\u5e38\u4e0d\u8212\u670d\u3002\u4e00\u822c\u8655\u7406\u9ea5\u514b\u98a8\u562f\u53eb\uff0c\u5e38\u662f\u5229\u7528 \u79fb\u983b\u6253\u65b7\u5171\u632f\uff0c\u6216\u662f\u7528\u5e36\u963b\u6ffe\u6ce2\u5668\u505a\u4e8b\u5f8c\u88dc\u6551\uff0c\u4f46\u6709\u53ef\u80fd\u6703\u9020\u6210\u97f3\u8cea\u7834\u58de\u3002\u56e0\u6b64\u6211\u5011\u60f3 \u6539\u7528\u9069\u61c9\u6027\u56de\u6388\u6d88\u9664\u6f14\u7b97\u6cd5\uff0c\u5229\u7528\u64f4\u5927\u6a5f\u5587\u53ed\u7684\u8f38\u5165\u97f3\u6e90\u7576\u53c3\u8003\u8a0a\u865f\uff0c\u4f86\u81ea\u52d5\u4f30\u7b97\u5728\u4e0d \u540c\u7a7a\u9593\u74b0\u5883\u3001\u4e0d\u540c\u6b4c\u66f2\u3001\u4e0d\u540c\u8a0a\u96dc\u6bd4\u4e0b\uff0c\u9ea5\u514b\u98a8\u53ef\u80fd\u9304\u5230\u7684\u56de\u6388\u8a0a\u865f\uff0c\u4e26\u5728\u505a\u8a0a\u865f\u589e\u76ca \u524d\u5148\u5c07\u5176\u6d88\u9664\uff0c\u4ee5\u76f4\u63a5\u5f9e\u6e90\u982d\u6d88\u9664\u562f\u53eb\u767c\u751f\u7684\u53ef\u80fd\u6027\u3002\u57fa\u65bc\u4ee5\u4e0a\u60f3\u6cd5\uff0c\u5728\u672c\u8ad6\u6587\u4e2d\uff0c\u5be6 \u73fe\u4e86 normalized least mean square(NLMS)\u7684\u562f\u53eb\u6d88\u9664\u6f14\u7b97\u6cd5\uff0c\u5c24\u5176\u662f\u9032\u4e00\u6b65\u8003\u616e\u64f4\u97f3 \u7cfb\u7d71\u7684\u975e\u7dda\u6027\u5931\u771f\uff0c\u63d0\u51fa\u57fa\u65bc recurrent neural network(RNN)\u7684\u9032\u968e\u6f14\u7b97\u6cd5\u3002\u4e26\u5728\u5be6 \u9a57\u6642\u5206\u5225\u6e2c\u8a66\u5728\u6642\u57df\u6216\u662f\u983b\u57df\u8655\u7406\uff0c\u8207\u4f7f\u7528 NLMS \u6216\u662f RNN\uff0c\u5c0d\u4e0d\u540c\u66f2\u98a8\u3001\u4e0d\u540c\u74b0\u5883"
},
"TABREF2": {
"content": "<table><tr><td>4) \u6b4c\u4e2d\u7684\u67d0\u4e9b\u983b\u6bb5\u525b\u597d\u8207\u562f\u53eb\u767c\u751f\u983b\u6bb5\u7b26\u5408\uff0c\u56e0\u6b64\u5feb\u6b4c\u5728\u6d88\u9664\u6548\u679c\u4e0a\u6c92\u6162\u6b4c\u4f86\u7684\u597d\u3002 \u8868\u4e8c\u3001\u74b0\u5883\u97ff\u61c9\u53c3\u6578\u8a2d\u7f6e(\u55ae\u4f4d:\u516c\u5c3a) \u6536\u6582\u901f\u5ea6\u6703\u76f8\u8f03\u6162\u7684\u3002 \u4e94\u3001\u7d50\u8ad6</td></tr><tr><td>\u64da\u6642\u9593\u524d\u5f8c\u9806\u5e8f\u4f86\u8abf\u6574\u6b0a\u91cd\u503c\uff0c\u56e0\u6b64\u8abf\u6574\u6b0a\u91cd\u6703\u7d93\u7531\u4e0d\u540c\u6642\u9593\u9ede\u7684\u96b1\u85cf\u5c64\u8cc7\u8a0a\u9032\u884c\uff0c\u7531 \u6700\u5f8c\u6642\u9593\u9ede\u5c0d\u6210\u672c\u51fd\u6578 (cost function) \u4f5c\u504f\u5fae\u5206\uff0c\u5f80\u524d\u4f30\u7b97\u51fa\u4e00\u958b\u59cb\u6642\u9593\u9ede\u7684\u504f\u5fae\u5206\u503c\uff0c \u76f4\u5230\u6574\u500b\u6b0a\u91cd\u4f5c\u51fa\u8abf\u6574\u5b8c\u5f8c\uff0c\u518d\u4ee3\u56de\u7db2\u8def\u6c42\u65b0\u8aa4\u5dee\u503c\uff0c\u4f7f MSE \u63a5\u8fd1\u6700\u5c0f\u503c\uff0c\u8b93\u8f38\u51fa\u63a5 \u8fd1\u671f\u671b\u7684\u503c\u3002 \u56db\u3001\u5be6\u9a57\u7d50\u679c \u672c\u8ad6\u6587\u4f7f\u7528\u4e4b\u97f3\u6a94\u70ba\u81ea\u884c\u9304\u88fd\u771f\u5be6\u4eba\u8072(A Cappella)\u8207\u4eba\u8072\u540c\u4e00\u66f2\u76ee\u7684\u4f34\u5531\u97f3\u6a94\uff0c\u6bd4\u8f03 \u6642\u57df NLMS\u3001\u983b\u57df NLMS\u3001\u6642\u57df RNN \u7b49\u4e09\u500b\u6f14\u7b97\u6cd5\u3002 \u70ba\u907f\u514d\u9ea5\u514b\u98a8\u4e00\u6253\u958b\u6536\u97f3\u5c31\u63a5\u6536\u5927\u91cf\u80fd\u91cf\uff0c\u4ee5\u81f4\u9069\u61c9\u6027\u6ffe\u6ce2\u5668\u5c1a\u672a\u6536\u6582\u4fbf\u5f15\u767c\u4e0d\u53ef\u6536\u62fe \u7684\u562f\u53eb\uff0c\u56e0\u6b64\u5c07\u6e2c\u8a66\u97f3\u6a94\u7686\u7de8\u8f2f\u6210:\u7b2c\u4e00\u6bb5(\u97f3\u6a02)\u3001\u7b2c\u4e8c\u6bb5(\u97f3\u6a02)\u3001\u7b2c\u4e09\u6bb5(\u4eba\u8072)\u3001\u7b2c\u56db\u6bb5 (\u4eba\u8072)\u3001\u7b2c\u4e94\u6bb5(\u4eba\u8072)\u3001\u7b2c\u516d\u6bb5(\u97f3\u6a02)\u4e4b\u9577\u5ea6\uff0c\u800c\u7b2c\u4e00\u6bb5(\u97f3\u6a02)\u70ba\u7d14\u4f34\u5531\u97f3\u6a02\u4e14\u672a\u5c07\u6ffe\u6ce2 \u5668\u6d88\u9664\u7d50\u679c\u8f38\u51fa\u81f3\u5587\u53ed\uff0c\u70ba\u55ae\u7d14\u5b78\u7fd2\u74b0\u5883\u97ff\u61c9\u3001\u8abf\u6574\u6ffe\u6ce2\u5668\u6b0a\u91cd\uff1b\u7b2c\u4e8c\u6bb5(\u97f3\u6a02)\u624d\u5c07\u6ffe \u6ce2\u5668\u6d88\u9664\u7d50\u679c\u52a0\u5165\u5587\u53ed\u8f38\u51fa\uff0c\u5230\u7b2c\u4e09\u6bb5(\u4eba\u8072)\u624d\u6b63\u5f0f\u5c07\u8fd1\u7aef\u4eba\u8072\u52a0\u5165\u7cfb\u7d71\u4e4b\u4e2d\u3002\u6b64\u6a21\u64ec \u985e\u4f3c\u70ba google \u4e4b Google Home \u8207 apple \u4e4b HomePod\uff0c\u5728\u958b\u6a5f\u6642\u7d66\u4e88\u4e00\u500b\u63d0\u793a\u97f3\uff0c\u4f7f\u7cfb \u7d71\u5b78\u7fd2\u74b0\u5883\u97ff\u61c9\u7684\u60c5\u6cc1\uff0c\u907f\u514d\u6bcf\u6b21\u958b\u6a5f\u7686\u662f\u4e0d\u540c\u74b0\u5883\u7684\u60c5\u5f62\u3002 (\u4e00) \u97f3\u6a94\u8aaa\u660e \u8868\u4e00\u3001\u97f3\u6a94\u683c\u5f0f \u4f4d\u5143\u7387 \u8072\u9053\u6578\u91cf \u53d6\u6a23\u983b\u7387 \u9577\u5ea6(\u79d2) \u66f2\u98a8 \u7709\u98db\u8272\u821e 16-bit \u55ae\u8072\u9053 16000 Hz 33.5 \u5feb\u6b4c \u5929\u9ed1\u9ed1 16-bit \u55ae\u8072\u9053 16000 Hz 33.5 \u6162\u6b4c (\u4e8c) \u74b0\u5883\u97ff\u61c9\u8aaa\u660e(\u623f\u9593\u5927\u5c0f\u3001\u56de\u97f3\u9577\u5ea6) \u4ee5\u4e0b\u74b0\u5883\u97ff\u61c9\u70ba RIR-Generator \u751f\u6210\u4e4b\u74b0\u5883\u97ff\u61c9\u97f3\u6a94[9]\uff0c\u6211\u5011\u5c07\u6b64\u97f3\u6a94\u647a\u7a4d\u8f38\u51fa\u97f3\u6a94\u4ee5 \u4e4b\u4f86\u6a21\u64ec\u771f\u5be6\u74b0\u5883\u4e0b\u9ea5\u514b\u98a8\u6240\u63a5\u6536\u5230\u7684\u74b0\u5883\u97ff\u61c9\u56de\u8072\u3002\u4ee5\u4e0b\u8868\u4e8c\u70ba\u74b0\u5883\u97ff\u61c9\u53c3\u6578\u8a2d\u7f6e\uff0c \u7686\u70ba\u8072\u901f(Sound velocity): 340 (m/s)\u3001\u53d6\u6a23\u983b\u7387(Sample frequency) = 16000 (sample/s)\u3002 \u5927\u623f\u9593(\u79ae\u5802\u5927\u5c0f) \u4e2d\u623f\u9593(\u6703\u8b70\u5ba4\u5927\u5c0f) \u5c0f\u623f\u9593(\u8eca\u5167\u5927\u5c0f) Source position [10.0 4.0 2.0] [2.5 1.0 1.2] [1.7 0.2 0.2] Receiver position [6.0 4.0 1.5] [2.5 2.0 1.6] [1.7 0.8 0.8] Room dimensions [15.5 8.5 6.0] [5.0 4.0 3.0] [2.3 1.7 1.2] Beta(\u6b98\u97ff\u9577\u5ea6)(s) 1.8 0.45 0.1 (\u4e09)\u6291\u5236\u8a55\u4f30-MSE \u56de\u8072\u6d88\u9664\u6548\u679c\u9664\u4e86\u4e3b\u89c0\u7531\u8033\u6735\u807d\u8072\u97f3\u4e4b\u5916\uff0c\u6211\u5011\u5c07\u4f7f\u7528\u5e73\u5747\u8aa4\u5dee\u503c(Mean Squared Error\uff0c MSE)\u4f86\u6578\u503c\u5316\u5176\u56de\u8072\u6d88\u9664\u7684\u6548\u679c\u4e0a\uff0c\u5176\u65b9\u7a0b\u5f0f\u5982\u4e0b\u6240\u793a: MSE = 1 \u2211 ( ( ) \u2212 ( )) 2 =1 \u5176\u4e2d\uff0cs(n)\u70ba\u8fd1\u7aef\u4eba\u8072\u4e4b\u539f\u59cb\u97f3\u6a94\u3002e(n)\u70ba\u5269\u9918\u8a0a\u865f:\u5373\u7d93\u56de\u8072\u6d88\u9664\u7cfb\u7d71\u6d88\u9664\u74b0\u5883\u97ff\u61c9\u5f8c \u5f97\u5230\u7684\u5269\u9918\u4eba\u8072\u3002\u501f\u7531\u539f\u59cb\u7684\u8fd1\u7aef\u4eba\u8072 s(n)\u8207\u7d93\u56de\u8072\u6d88\u9664\u7684\u5269\u9918\u4eba\u8072 e(n)\uff0c\u5169\u8005\u76f8\u76f8\u6e1b \u53d6\u5e73\u65b9\uff0c\u7576 MSE \u503c\u6108\u5c0f\uff0c\u8868\u793a\u6d88\u9664\u6548\u679c\u6108\u597d\uff0c\u4ee3\u8868\u4e5f\u5f97\u5230\u6108\u4e7e\u6de8\u7684\u8fd1\u7aef\u4eba\u8072\u3002 (\u56db)\u5be6\u9a57\u7d50\u679c \u6bcf\u500b\u5be6\u9a57\u7686\u542b (\u6642\u57df NLMS\u3001\u983b\u57df NLMS\u3001RNN) 1.\u5be6\u9a57\u4e00\uff0c\u540c\u4e00\u9996\u6b4c_\u4e0d\u540c\u74b0\u5883(\u623f\u9593\u5927\u5c0f\u3001\u56de\u97f3\u7a0b\u5ea6) \u4e0b\u5716\u516d\u4e2d\uff0c\u5728\u4e0d\u540c\u74b0\u5883\u4e0b\u6bcf\u500b\u6f14\u7b97\u6cd5\u7684\u6548\u679c\u7686\u662f\u7a69\u5b9a\u7684\uff0c\u6574\u7406\u4f86\u770b\u662f\u6642\u57df\u6703\u597d\u904e\u983b\u57df\uff0c \u4e0d\u904e\u4ee5\u6f14\u7b97\u6cd5\u4f86\u8aaa RNN \u8207 NLMS \u5f7c\u6b64\u7684\u6548\u679c\u537b\u662f\u76f8\u5dee\u4e0d\u591a\u7684\u3002 \u5716\u516d\u3001\u5be6\u9a57\u4e00\u6bd4\u8f03\u5716 2.\u5be6\u9a57\u4e8c\uff0c\u540c\u4e00\u74b0\u5883_\u4e0d\u540c\u9996\u6b4c \u4e0b\u5716\u4e03\u4e2d\uff0c\u5728\u4e0d\u540c\u9996\u6b4c\u4e0b\u6162\u6b4c\u6bd4\u5feb\u6b4c\u6548\u679c\u4f86\u7684\u66f4\u597d\uff0c\u9019\u6216\u8a31\u8ddf\u562f\u53eb\u767c\u751f\u983b\u6bb5\u6709\u95dc\uff0c\u5728\u5feb \u5716\u4e03\u3001\u5be6\u9a57\u4e8c\u6bd4\u8f03\u5716 3.\u6642\u57df NLMS\u3001\u983b\u57df NLMS\u3001\u6642\u57df RNN \u6bd4\u8f03(\u540c\u4e00\u9996\u6b4c\u3001\u540c\u4e00\u74b0\u5883) \u4e0b\u5716\u516b\u3001\u4e5d\u3001\u5341\u3001\u5341\u4e00\uff0c\u70ba\u6642\u57df NLMS\u3001\u6642\u57df RNN\u3001\u983b\u57df NLMS \u7684\u6642\u57df\u5716\u3001\u983b\u8b5c\u5716 \u6bd4\u8f03\uff0c\u6bd4\u8f03\u4e0b\u4f86\u6642\u57df NLMS \u8207\u6642\u57df RNN \u6548\u679c\u76f8\u5dee\u4e0d\u591a\uff0c\u800c\u983b\u57df NLMS \u7531\u65bc\u6536\u6582\u6bd4\u8f03 \u6162\u7684\u95dc\u4fc2\uff0c\u6240\u4ee5\u6548\u679c\u7565\u905c\u8272\u65bc\u5176\u4ed6\u5169\u8005\uff0c\u63a5\u4e0b\u4f86\u6703\u6bd4\u8f03\u7b2c\u4e00\u3001\u4e8c\u6bb5\u8207\u6700\u5f8c\u4e00\u6bb5\u7684\u6548\u679c \u8207\u6536\u6582\u901f\u5ea6\u3002 \u2022 \u5716\u4e5d\u3001\u6642\u57df NLMS \u5269\u9918\u4eba\u8072\u983b\u57df\u5716 \u5716\u5341\u3001\u6642\u57df RNN \u5269\u9918\u4eba\u8072\u983b\u57df\u5716 \u5716\u5341\u4e00\u3001\u983b\u57df NLMS \u5269\u9918\u4eba\u8072\u983b\u57df\u5716 \u7531\u5716\u5341\u4e8c\u3001\u5341\u4e09\u3001\u5341\u56db\u53ef\u77e5\uff0c\u7b2c\u4e00\u6bb5\u3001\u7b2c\u4e8c\u6bb5\u6536\u6582\u904e\u7a0b\u4e2d\uff0c\u6642\u57df RNN \u7684\u6d88\u9664\u6548\u679c\u8f03\u597d \u4e14\u8f03\u5feb\uff0c\u63a5\u8457\u662f\u6642\u57df NLMS\uff0c\u4f46\u5f9e\u6f14\u7b97\u8907\u96dc\u5ea6\u4f86\u770b\uff0c\u6642\u57df\u662f\u6bcf\u4e00\u9ede\u90fd\u8a08\u7b97\u4e00\u6b21\uff0c\u4e5f\u5c31\u662f \u8aaa\u7b2c\u4e00\u3001\u4e8c\u6bb5 67 \u79d2\u7684\u6b4c\u66f2\u4e2d\u4fbf\u6f14\u7b97\u4e86 NLMS 1072 \u5343\u6b21\uff0c\u6bcf\u4e00\u9ede\u90fd\u8abf\u6574\u4e00\u6b21\u6b0a\u91cd\uff1b\u800c\u983b \u57df\u7531\u65bc\u662f\u6bcf 256 \u9ede\u624d\u53d6\u4e00\u6b21 512 \u9ede\u6578\u505a\u904b\u7b97\uff0c512 \u9ede\u6bcf\u4e00\u9ede\u90fd\u7528\u4e00\u6a23\u7684\u8abf\u6574\u91cf\uff0c\u983b\u57df\u7684 \u5716\u5341\u4e8c\u3001\u6642\u57df NLMS \u7b2c\u4e00\u3001\u4e8c\u6bb5\u983b\u57df\u5716 \u5716\u5341\u4e09\u3001\u6642\u57df RNN \u7b2c\u4e00\u3001\u4e8c\u6bb5\u983b\u57df\u5716 \u5716\u5341\u56db\u3001\u983b\u57df NLMS \u7b2c\u4e00\u3001\u4e8c\u6bb5\u983b\u57df\u5716 \u7d93\u904e\u4e2d\u9593\u7b2c\u4e09\u3001\u56db\u3001\u4e94\u6bb5\u6709\u52a0\u5165\u4eba\u8072\u5f8c\uff0c\u56de\u5230\u6700\u5f8c\u4e00\u6bb5\u50c5\u6709\u97f3\u6a02\u7684\u7b2c\u516d\u6bb5\uff0c\u5982\u5716\u5341\u4e94\u7d05 \u6846\u90e8\u5206\u6240\u793a\uff0c\u6b64\u6642\u5f7c\u6b64\u90fd\u5df2\u7d93\u6536\u6582\u5f97\u5dee\u4e0d\u591a\u4e86\uff0c\u4f46\u6642\u57df\u90e8\u5206\u6b64\u6642\u5df2\u7d93\u6f14\u7b97\u4e86 3216 \u5343\u6b21 \u95dc\u4fc2\uff0c\u6240\u4ee5\u7a81\u7136\u7684\u8b8a\u5316\uff0c\u4ecd\u53ef\u4ee5\u7a69\u5b9a\u6d88\u9664\u56de\u8072\uff1b\u800c RNN \u96d6\u7136\u4e5f\u70ba\u6642\u57df\u904b\u7b97\uff0c\u4f46\u56e0\u70ba\u5176 \u70ba\u975e\u7dda\u6027\u7684\u6f14\u7b97\u6cd5\uff0c\u5728\u9019\u7a2e\u60c5\u5f62\u4e0b\u4fbf\u4ecd\u6709\u826f\u597d\u7684\u6548\u679c\u3002\u5728\u505a\u66f4\u4e45\u7684\u6f14\u7b97\u4e0b\uff0c\u6642\u57df\u6216\u8a31\u4ecd \u56e0\u6bcf\u4e00\u9ede\u90fd\u904b\u7b97\u4e00\u6b21\u800c\u6bd4\u983b\u57df\u6548\u679c\u4f86\u7684\u66f4\u597d\uff0c\u4f46\u7531\u65bc\u5f7c\u6b64\u90fd\u6536\u6582\u7684\u60c5\u5f62\u4e0b\uff0c\u6548\u679c\u4e26\u4e0d\u6703 \u76f8\u5dee\u592a\u591a\uff0c\u4f46\u4e09\u8005\u7684\u6f14\u7b97\u91cf\u8207\u7a69\u5b9a\u5ea6\u5728\u65e5\u5f8c\u7684\u61c9\u7528\u4e0a\uff0c\u9700\u8981\u5728\u6548\u679c\u8207\u8a08\u7b97\u91cf\u4e0a\u505a\u53d6\u6368\u4e86\u3002 \u5716\u5341\u4e94\u3001\u6642\u57df NLMS\u3001\u6642\u57df RNN\u3001\u983b\u57df NLMS \u7b2c\u516d\u6bb5\u6642\u57df\u5716 \u5716\u5341\u4e03\u3001\u6642\u57df RNN \u7b2c\u516d\u6bb5\u983b\u57df\u5716 \u5716\u5341\u516b\u3001\u983b\u57df NLMS \u7b2c\u516d\u6bb5\u983b\u57df\u5716 \u5728\u672c\u5be6\u9a57\u4e2d\uff0c\u6a21\u64ec\u4e86\u771f\u5be6\u74b0\u5883\u4e2d\u9ea5\u514b\u98a8\u5728\u4e0d\u540c\u7a7a\u9593\u4e0b\u7684\u6536\u97f3\u8207\u74b0\u5883\u97ff\u61c9\u60c5\u5f62\uff0c\u5206\u5225\u6a21\u64ec \u4e86\u5927\u623f\u9593\u3001\u4e2d\u623f\u9593\u3001\u5c0f\u623f\u9593\u7684\u74b0\u5883\u97ff\u61c9\uff0c\u8a9e\u6599\u90e8\u5206\u4e5f\u5206\u4e86\u4e0d\u540c\u66f2\u98a8\u7684\u6b4c\u66f2\u4f86\u505a\u5be6\u9a57\uff0c\u63a5 \u8457\u5206\u5225\u6e2c\u8a66\u4e86\u7dda\u6027\u7684\u6642\u57df NLMS\u3001\u983b\u57df NLMS \u4ee5\u53ca\u975e\u7dda\u6027\u6ffe\u6ce2\u6f14\u7b97\u6cd5 RNN\uff0c\u4f5c\u9ea5\u514b\u98a8 \u562f\u53eb\u6291\u5236\u5be6\u9a57\u3002\u5728\u6642\u57df NLMS \u8207\u983b\u57df NLMS \u4e0a\uff0c\u5728\u4e00\u958b\u59cb\u6642\u57df\u7684\u7531\u65bc\u662f\u6bcf\u4e00\u9ede\u5c31\u505a\u4e00 \u6b21\uff0c\u5176\u6536\u6582\u6548\u679c\u6703\u6bd4\u983b\u57df\u4f86\u7684\u66f4\u5feb\uff0c\u751a\u81f3\u6d88\u9664\u6548\u679c\u66f4\u597d\uff0c\u4f46\u5728\u5169\u8005\u90fd\u6f14\u7b97\u4e86\u4e00\u6bb5\u6642\u9593\u5f8c\uff0c \u5f7c\u6b64\u90fd\u5df2\u9054\u5230\u4e86\u6536\u6582\uff0c\u6548\u679c\u5176\u5be6\u662f\u5dee\u4e0d\u591a\u7684\uff0c\u4f46\u983b\u57df\u5728\u67d0\u4e9b\u7a81\u5982\u7684\u9ad8\u983b\u6216\u4f4e\u983b\u653e\u9762\u6703\u6bd4 \u6642\u57df\u4f86\u7684\u6548\u679c\u66f4\u597d\uff0c\u8a08\u7b97\u8907\u96dc\u5ea6\u4e0a\uff0c\u983b\u57df\u7684\u647a\u7a4d\u6bd4\u6642\u57df\u4f86\u7684\u7c21\u55ae\uff0c\u4e14 256 \u9ede\u624d\u505a\u4e00\u6b21\u904b \u7b97\uff0c\u4f46\u6bcf\u6b21\u90fd\u662f\u4e00\u6b21\u8abf\u6574 512 \u9ede\uff0c\u56e0\u6b64\u8a08\u7b97\u91cf\u662f\u76f8\u5dee\u4e0d\u591a\u7684\u3002\u6b64\u5916\u6642\u57df RNN \u4e0a\u7531\u65bc\u975e \u7dda\u6027\u8207\u6709\u6642\u9593\u7684\u8a18\u61b6\uff0c\u5728\u67d0\u4e9b\u90e8\u5206\u6d88\u9664\u7684\u6548\u679c\u5176\u5be6\u662f\u6700\u597d\u7684\uff0c\u4f46\u7531\u65bc\u5176\u904b\u7b97\u91cf\u9f90\u5927\uff0c\u65e5 \u7684\u904b\u7b97\u91cf\u4e86\uff0c\u800c\u983b\u57df\u6f14\u7b97\u6cd5\u624d\u505a\u4e86 12.56 \u63db\u7684\u60c5\u5f62\u4e0b\uff0c\u5c31\u9700\u8981\u4e00\u6bb5\u6642\u9593\u624d\u80fd\u91cd\u65b0\u6536\u6582\uff0c\u4f46\u983b\u57df NLMS \u56e0\u70ba\u6709\u7167\u9867\u5230\u6bcf\u500b\u983b\u6bb5\u7684 \u5716\u5341\u516d\u3001\u6642\u57df NLMS \u7b2c\u516d\u6bb5\u983b\u57df\u5716 \u5f8c\u82e5\u6709\u61c9\u7528\uff0c\u5728\u9019\u4e09\u8005\u4e4b\u9593\uff0c\u8a08\u7b97\u91cf\u8207\u6548\u679c\u597d\u58de\u7684\u53d6\u6368\u4fbf\u7aef\u958b\u4f7f\u7528\u7684\u60c5\u6cc1\u3002</td></tr><tr><td>\u5716\u516b\u3001\u6642\u57df NLMS\u3001\u6642\u57df RNN\u3001\u983b\u57df NLMS \u6642\u57df\u6bd4\u8f03\u5716</td></tr></table>",
"num": null,
"type_str": "table",
"html": null,
"text": "\u6700\u5f8c\u518d\u7531\u6b0a\u91cd upk \u9023\u63a5\u7b2c\u4e8c\u5c64\u96b1\u85cf\u5c64\u8f38\u51fay k (t)\u5230\u8f38\u51fa\u5c64y p (t)\uff0c\u6b64\u6642y p (t)\u5c31\u662f\u6df1\u5c64\u905e\u8ff4 \u5f0f\u7db2\u8def\u6700\u7d42\u8f38\u51fa\u3002 \u6709\u5225\u65bc\u4e00\u822c RNN \u7684\u53cd\u5411\u50b3\u64ad\u6f14\u7b97\u6cd5(Backpropagation Through Time\uff0cBPTT)\uff0c\u6211\u5011\u6703\u6839 \u5343\u6b21\u5de6\u53f3\u7684\u8a08\u7b97\u8207 FFT\uff0c\u5f7c\u6b64\u5728 NLMS \u8a08\u7b97\u91cf \u4ecd\u76f8\u5dee\u4e86 256 \u500d\uff0c\u800c\u6642\u57df RNN \u662f\u5176\u4e2d\u6f14\u7b97\u6700\u8907\u96dc\u7684\u3002\u6b64\u5916\u5f9e\u5716\u5341\u516d\u3001\u5341\u4e03\u3001\u5341\u516b\u6bd4\u8f03 \u53ef\u5f97\u77e5\uff0c\u983b\u57df NLMS \u5728 500 Hz \u7684\u5730\u65b9\u4ecd\u6709\u4e9b\u8a31\u6c92\u6d88\u4e7e\u6de8\uff0c\u800c\u6642\u57df NLMS \u5728\u97f3\u6a94\u7a81\u7136\u8b8a"
}
}
}
}