{
    "paper_id": "O04-1001",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T08:00:47.153400Z"
    },
    "title": "",
    "authors": [],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "",
    "pdf_parse": {
        "paper_id": "O04-1001",
        "_pdf_hash": "",
        "abstract": [],
        "body_text": [
            {
                "text": "\u96a8\u8457\u6c7d\uf902\u5c0e\u822a\u7cfb\u7d71\u7684\u65e5\u6f38\u666e\u53ca\uff0c\u9664\uf9ba\u63d0\u4f9b\u6c7d\uf902\ufa08\uf902\u8cc7\uf9be\u53ca\u5a1b\uf914\u5916\uff0c\u85c9\u7531\u7d50\u5408\ufa08\u52d5\u96fb\u8a71\u7684\u7121\u7dda\u901a\u8a0a\u529f\u80fd\uff0c\uf901\u8b93",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u524d\u8a00",
                "sec_num": "1."
            },
            {
                "text": "Model PWPD ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u524d\u8a00",
                "sec_num": "1."
            },
            {
                "text": "(b) \u5716\u4e8c\uff1a4KHz \u5167(a)\u4eba\u8033\u7684\u807d\u89ba\u7684\u5df4\u514b\u983b\u8b5c\u53ca(b)\u95dc\u9375\u983b\u5bec\u66f2\u7dda\u5716\u3002 Decomposition Level w 5,1 w 5,2 w 5,3 w 5,4 w 5,7 w 5,8 w 5,6 w 5,5 0 0.5 1.0 1.5 2.0 3.5 3.0 2.5 4.0 w 4,9 w 4,10 w 4,11 w 4,12 w 4,13 w 4,14 w 3,15 w 3,16 w 3,17 Frequency (kHz) 5 4 3 2 1 0 (a) w 5,1 w 5,2 w 5,3 w 5,4 w 5,7 w 5,8 w 5,6 w 5,5 0 0.5 1.0 1.5 2.0 3.5 3.0 2.5 4.0 w 4,9 w 4,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u524d\u8a00",
                "sec_num": "1."
            },
            {
                "text": "\u2211 = = M m m m V s y 1 K M < , (1) \u5176\u4e2d M s s ..., , 1 \u70ba\u5e73\u5747\u503c\u70ba\uf9b2\u7684\u96a8\u6a5f\u8b8a\uf969\uff0c\u800c M V V ..., , 1 \u70ba\u57fa\u5e95\u3002\u8a0a\u865f\u5206\u4f48\u7684\u5411\uf97e\u7a7a\u9593\u5176\u7dad\ufa01\u70ba K\uff0c\u800c\u8a9e\u97f3\u8a0a \u865f\u5206\u4f48\u7684\u5b50\u7a7a\u9593\u5176\u7dad\ufa01\u70ba M\uff0c K M < \u3002\u5f0f\u5b50(1)\u53ef\u4ee5\u8868\u793a\u70ba Vs y = \uff0c ] ,..., [ 1 M V V V \u2261 \u70ba M K \u00d7 \u7684\u77e9\u9663\uff0c \u5176\u79e9(Rank)\u70ba M\uff0c\u4e14 s \u70ba\u4e00\ufa08\u5411\uf97e\u8868\u793a\u70ba T M s s s ) ,..., ( 1 \u2261 \u3002\u8a9e\u97f3\u8a0a\u865f\u7684\u5171\u8b8a\uf962\uf969\u77e9\u9663 T s y V VR R = \u5176\u79e9\u70ba M\uff0c s R \u70ba s \u7684\u5171\u8b8a\uf962\uf969\u77e9\u9663\u4e26\u5047\u8a2d\u5176\u70ba\u6b63\u5b9a\u77e9\u9663(Positive Definite Matrix)\u3002 K M < \u7684\u6027\u8cea\u4f7f\u5f97\u5728 K \u7dad\u8a9e\u97f3\u8a0a\u865f y \u4e2d\uff0cR y \u6709 M K \u2212 \u500b\u7279\u5fb5\u503c\u70ba\uf9b2\uff0c\u9019\u500b\u5c0d\u65bc\u5728\u4ee5\u5b50\u7a7a\u9593\u6f14\u7b97\u6cd5\u505a\u8a9e\u97f3\u5f37\u5065\u4e2d\u6975\u70ba\u91cd\u8981\u3002 \uf9a8 w \u70ba K \u7dad\u5411\uf97e\u8868\u793a\u80cc\u666f\u767d\u8272\u566a\u97f3\uff0c\u5176\u5e73\u5747\u503c\u70ba\uf9b2\u3002\u5176\u5171\u8b8a\uf962\uf969\u77e9\u9663 I ww E R w T w 2 } { \u03c3 = = \u3002\u767d\u8272\u566a\u97f3 \u7684\u5171\u8b8a\uf962\uf969\u77e9\u9663\u5176\u79e9\u70ba K \uff0c\u4e5f\u5c31\u662f\uf96f\u5b83\u6703\u4f48\u6eff\u6574\u500b\u6b50\u5f0f\u7a7a\u9593 R K \u4e2d\u3002\u56e0\u6b64\uff0c\u5c0d\u65bc\u80cc\u666f\u70ba\u767d\u8272\u566a\u97f3\u7684\u96dc\u8a0a\u8a9e\u97f3 \u8a0a\u865f\uff0c\u6574\u500b K \u7dad\u7684\u5411\uf97e\u7a7a\u9593\u7531 M \u7dad\u7684\u8a0a\u865f\u5b50\u7a7a\u9593\u53ca K \u7dad\u7684\u566a\u97f3\u5b50\u7a7a\u9593\u6240\u7d44\u5408\u800c\u6210\uff0c\u5176\u4e2d\u53ef\u4ee5\u5c07 M K \u2212 \u7684 \u7279\u5fb5\u503c\u70ba\uf9b2\u6240\u5c0d\u61c9\u7684\u5b50\u7a7a\u9593\u53bb\u9664\u6389\uff0c\u800c\u5269\u4e0b\u7684 M \u7dad\u7684\u96dc\u8a0a\u5b50\u7a7a\u9593\uff0c\u53ef\u4ee5\u7528\u4e00\u7dda\u6027\u9810\u4f30\u5668\u5c07\u5176\u4e7e\u6de8\u8a9e\u97f3\u7cb9\u53d6 \u51fa\uf92d\u3002 \u5e95\u4e0b\u5c07\uf96f\u660e\u7dda\u6027\u9810\u4f30\u5668\u7684\u6c42\u53d6\uff0c\u5047\u8a2d\u96dc\u8a0a\u8a9e\u97f3\u70ba ) ( ) ( ) ( n W n Y n Z + = \uff0c ) (n W \u70ba K \u7dad\u7684\u80cc\u666f\u566a\u97f3\u5411\uf97e\uff0c ) (n Y \u70ba K \u7dad\u7684\u8a9e\u97f3\u5411\uf97e\uff0cn \u70ba\u8a0a\u865f\u97f3\u6846\u7684\uf96a\u5f15\u3002\uf9a8 ) (n H \u70ba\u4e00 K K \u00d7 \u7684\u4e7e\u6de8\u8a9e\u97f3\u4e4b\u7dda\u6027\u9810\u4f30\u5668\u4ea6\u5373 ) ( ) ( ) ( n Z n H n Y = (2) \u5247\u5176\u9810\u4f30\u932f\u8aa4\u8a0a\u865f\u5247\u70ba ) ( ) ( ) ( ) ( ) ( ) ) ( ( ) ( ) ( ) ( y n n n W n H n Y I n H n Y n Y n w \u03b5 \u03b5 \u03b5 + = + \u2212 = \u2212 = (3) ) ( ) ) ( ( ) ( n Y I n H n y \u2212 \u2261 \u03b5 \u4ee3\u8868\u8a0a\u865f\u7684\u5931\u771f\uf97e\uff0cI \u70ba\u55ae\u4f4d\u77e9\u9663(Identity Matrix)\uff0c ) ( ) ( ) ( n W n H n w \u2261 \u03b5 \u4ee3\u8868\u566a\u97f3 \u7684\u6b98\u9918\uf97e\u3002\u5b9a\u7fa9\u8a0a\u865f\u5931\u771f\u80fd\uf97e\u53ca\u566a\u97f3\u6b98\u9918\u80fd\uf97e\u5206\u5225\u70ba ) ( 2 n y \u03b5 \u3001 ) ( 2 n w \u03b5 \u3002\u5247\u8a0a\u865f\u5931\u771f\u80fd\uf97e\u8868\u793a\u70ba ) ) ) ( )( ( ) ) ( (( )]), ( ) ( [ ( ) ( 2 T y T y y y I n H n R I n H tr n n E tr n \u2212 \u2212 = = \u03b5 \u03b5 \u03b5 (4) \u4e14\u566a\u97f3\u7684\u6b98\u9918\u80fd\uf97e\u70ba ) ) ( ) ( ) ( ( )]), ( ) ( [ ( ) ( 2 T w T w w w n H n R n H tr n n E tr n = = \u03b5 \u03b5 \u03b5 (5) ) (n R y \u53ca ) (n R w \u5206\u5225\u70ba\u4e7e\u6de8\u8a9e\u97f3\u8a0a\u865f\u53ca\u566a\u97f3\u8a0a\u865f\u7684\u5171\u8b8a\uf962\uf969\u77e9\u9663\u3002\u56e0\u8981\u5176\u8a0a\u865f\u5931\u771f\u80fd\uf97e\u6700\u5c0f\u5316\uff0c\u800c\u6700\u5c0f\u5316 \u7684\u60c5\u6cc1\u8981\u9650\u5236\u5728\u566a\u97f3\u80fd\uf97e\u5c0f\u65bc\u4e00\u500b\u5f88\u5c0f\u7684\u5e38\uf969\uff0c\u56e0\u6b64\u5176\u6700\u4f73\u7684\u7dda\u6027\u9810\u4f30\u5668\u5b9a\u7fa9\u5982\u4e0b 2 2 1 2 ) ( ) ( : Subject to ), ( min arg ) ( \u03c3 \u03b5 \u03b5 \u2264 \u2261 n n n H w K y n H opt (6) \u5176\u4e2d 2 \u03c3 \u662f\u4e00\u500b\u6b63\u7684\u5e38\uf969\u503c\u3002\u6c42\u89e3\u5f0f\u5b50(6)\uff0c\u53ef\uf9dd\u7528\uf925\u6c0f\u4e58\u5b50\u6cd5(Lagrange Multiplier)\uff0c\u5f97\u5230 ) ) ( ( ) ( ) ), ( ( 2 2 2 \u03c3 \u03b5 \u00b5 \u03b5 \u00b5 K n n n H L w y \u2212 + \u2261 (7) \u53ca 0 , 0 ) ) ( ( 2 2 \u2265 = \u2212 \u00b5 \u03c3 \u03b5 K n w (8) \u800c \u00b5 \u70ba\uf925\u6c0f\u4e58\u5b50\u3002\u5c0d\u5f0f\u5b50(7)\u53d6\u68af\ufa01\u904b\u7b97(gradient)\uff0c\uf9a8\u5176\u70ba\uf9b2\u6c42\u89e3\uff0c\u5247\u7dda\u6027\u9810\u4f30\u5668\u53ef\u5f97\u5230\u70ba 1 )) ( ) ( )( ( ) ( \u2212 + = n R n R n R n H w y y opt \u00b5 (9) \u7531\u7279\u5fb5\u503c\u5206\u89e3\uff0c\u5f0f\u5b50(9)\u53ef\u8868\u70ba ) ( )) ( ) ( )( ( ) ( ) ( 1 n U n n n n U n H T w y y opt \u2212 \u039b + \u039b \u039b = \u00b5 (10) \u7279\u5fb5\u5206\u89e3 ) ( ) ( ) ( ) ( n U n n U n R T y y \u039b = \uff0c ) (n U \u70ba\u7279\u5fb5\u5411\uf97e\u7684\u77e9\u9663\uff0c ) (n y \u039b \u70ba\u7279\u5fb5\u503c\u77e9\u9663\uff0c ) (n w \u039b \u70ba ) (n R w \u7684\u7279\u5fb5\u503c\u77e9\u9663\u3002\uf9a8 1 )) ( ) ( )( ( ) ( \u2212 \u039b + \u039b \u039b = n n n n G w y y \u00b5 \uff0c\u5247 ) ( ) ( ) ( ) ( n U n G n U n H T opt = (11) 2.3. \u5b50\u7a7a\u9593\u8ffd\u8e64\u6f14\u7b97\u6cd5 \u5b50\u7a7a\u9593\u8a9e\u97f3\u5f37\u5065\u6700\u5f8c\u7684\u7dda\u6027\u9810\u4f30\u5668\u9808\u8981\u96dc\u8a0a\u8a9e\u97f3\u7684\u7279\u5fb5\u5206\u89e3\uff0c\u5176\u904b\u7b97\u8907\u96dc\ufa01\u9ad8\u3002\u6240\u4ee5\u5728\u672c\uf941\u6587\u4e2d\u63a1\u7528\u8ffd\u8e64\u758a \u4ee3\u7684\u65b9\u5f0f\uf92d\u903c\u8fd1\u7279\u5fb5\u503c\uff0c\u9019\u500b\u6f14\u7b97\u6cd5\u7a31\u70ba PAST (Projection Approximation Subspace Tracking) [2]\u3002PAST \u7684\u6f14 \u7b97\u6cd5\u7528\uf92d\u8ffd\u8e64\u5b50\u7a7a\u9593\u7684\u7279\u5fb5\u503c\u5728\u8a31\u591a\u6587\u737b\u4e2d\u88ab\u8b49\u660e\u662f\u6e96\u78ba\u4e14\u8a08\u7b97\u8907\u96dc\ufa01\u4f4e\u7684\u3002\uf974\u4ee5\u5b50\u7a7a\u9593\u6f14\u7b97\u6cd5\uff0c\u5176\u904b\u7b97\u8907 \u96dc\ufa01\u70ba ) ( 3 n O \uff0c n \u70ba\u8f38\u5165\u5411\uf97e\u7684\u7dad\ufa01\uff0c\uf974\u5b50\u7a7a\u9593\u8ffd\u8e64\u65b9\u6cd5\uf92d\u505a\u8a08\u7b97\uff0c\u5176\u904b\u7b97\u8907\u96dc\ufa01\u53ef\u4ee5\u6e1b\u5c11\u81f3 ) (nr O \uff0c\u5176 \u4e2d n \u70ba\u8f38\u5165\u5411\uf97e\u7684\u7dad\ufa01\uff0c r \u70ba\u6211\u5011\u9700\u8981\u7684\u7279\u5fb5\u503c\u66a8\u7279\u5fb5\u5411\uf97e\u7684\u7684\uf969\u76ee\u3002 PAST \u6f14\u7b97\u6cd5\u5176\u539f\uf9e4\u70ba\u5c0d\u6240\u7d66\u5b9a\u7684\u6210\u672c\u51fd\uf969(Cost Function)\u6c42\u53d6\u6700\u5c0f\u503c\uff0c\u6210\u672c\u51fd\uf969\u7684\u6c7a\u5b9a\u8207\u5171\u8b8a\uf962\uf969\u77e9\u9663\u6709 \u95dc\uff0c 2 1 ) ( ) ( ) ( ) ( )) ( ( \u2211 = \u2212 \u2212 = n i T i n i Z n u n u i Z n u J \u03b2 (11) \u5176\u4e2d ) (n u \u70ba K \u7dad\u7684\u5411\uf97e\uff0c\u4e14 1 0 \u2264 \u2264 \u03b2 \u70ba\u6d88\u6563\u4fc2\uf969(Forgetting Factor)\uff0c \u03b2 \u53ef\u4ee5\u4f7f\u6210\u672c\u51fd\uf969\u7684\u6975\u503c\u6240\u5728\u6703\u662f\u4e0b \u9762\u5b9a\u7fa9\u7684\u76f8\u95dc\u77e9\u9663 ) (n R z \u4e2d\u7684\u7279\u5fb5\u5411\uf97e\u4e4b\u4e00\u3002 \u2211 = \u2212 = n i T i n Z i Z i Z n R 1 ) ( ) ( ) ( \u03b2 (12) \u5b9a\u7fa9\u4e00\u500b )) ( ( n u J \u2032 \u5982\u4e0b 2 1 ) ( ) 1 ( ) ( ) ( )) ( ( \u2211 = \u2212 \u2212 \u2212 = \u2032 n i T i n i Z i u n u i Z n u J \u03b2 (13) )) ( ( n u J \u548c )) ( ( n u J \u2032 \uf967\u540c\u5728\u65bc\u4f7f\u7528\uf9ba ) 1 ( \u2212 i u T \u4ee3\u66ff ) (n u T \uff0c\u76f4\u89ba\u7684\u89c0\u5bdf\uff0c )) ( ( n u J \u2032 \u53ef\u4ee5\u88ab\u7528\uf92d\u8fd1\u4f3c )) ( ( n u J \u2032 \u3002 \u56e0 \u70ba \u8a9e \u97f3 \u8a0a \u865f \u7684 \u7d71 \u8a08 \u7279 \u6027 \u70ba \u7a69 \u614b (stationary) \uff0c \u4e5f \u5c31 \u662f \u67d0 \u500b \u6642 \u9593 \u5340 \u6bb5 \u5b83 \u8b8a \u5316 \u7684 \u5f88 \u6162 \u6240 \u4ee5 ) ( ) 1 ( n u i u T T \u2248 \u2212 \u3002\u7576 n i << \u6642\uff0c i n\u2212 \u03b2 \u6703\u8b8a\u5f97\u5f88\u5c0f\u4f7f\u5f97\u6700\u5f8c )) ( ( )) ( ( n u J n u J \u2248 \u2032 \u3002\u6211\u5011\u53ef\u4ee5\u7528\u9069\u61c9\u6027\u68af\ufa01 \u6f14\u7b97\u6cd5\u5c07 )) ( ( n u J \u2032 \u53d6\u68af\ufa01\u904b\u7b97\u8fed\u4ee3\u6c42\u51fa\u7279\u5fb5\u5411\uf97e\u3002\u5176 PAST \u6f14\u7b97\u6cd5\u5982\u4e0b\u6240\u793a\u3002 \u8868\u4e8c: PAST \u6f14\u7b97\u6cd5\u3002 )] ( | | ) ( | ) ( [ ) ( \u8f38\u51fa\uff1a end end ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) 1 ( ) ( ) ( ) 1 ( ) ( ) ( ) ( ) 1 ( ) ( ) ( ) 1 ( ) ( do , 2, , 1 For do 2, , 1 For )] 0 ( | | ) 0 ( | ) 0 ( [ ) 0 ( 95 . 0 , 0 ) 0 ( \u521d\u59cb\u5316\uff1a 2 1 1 2 1 2 1 n u n u n u n U n v n u n Z n Z n d n v n E n T n u n u n v n u n Z n E n v n d n d n Z n u n vi k i Z(n) (n) Z n I u u u U d k i i i i i i i i i i i i i i i i i T i",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u524d\u8a00",
                "sec_num": "1."
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "= = + \u03b2 \u03b2 \u5176\u4e2d ) (n d i \u70ba\u5b50\u7a7a\u9593\u7684\u7279\u5fb5\u503c\u3002\u5c0d\u65bc\u4e7e\u6de8\u8a9e\u97f3\u4ee5\u53ca\u566a\u97f3\u70ba\u7121\u76f8\u95dc\u7684\u6240\u4ee5\u5728\u7279\u5fb5\u503c\u4e0a\u7684\u5206\u4f48\u53ef\u4ee5\u5beb\u6210 ) ( ) ( ) ( n n n w y z \u039b + \u039b = \u039b (14) \u6240\u4ee5\u53ea\u8981\u6c42\u96dc\u8a0a\u8a9e\u97f3\u7684\u7279\u5fb5\u503c\u4e14 ) (n w \u039b \u70ba\u767d\u8272\u566a\u97f3\u5176\u7d71\u8a08\u7279\u6027\u5df2\u77e5\uff0c\u53ef\u4ee5\u5c07 ) ( ) ( n n w z \u039b \u2212 \u039b \u5f97\u5230\u4e7e\u6de8\u8a9e\u97f3 \u7684 ) (n y \u039b \u3002\u4e00\u822c\uf92d\uf96f\u566a\u97f3\u7684\u7d71\u8a08\u7279\u6027\uf967\u662f\u7a69\u614b\u7684\u3002\u6240\u4ee5\u8981\u5b8c\u6210\u4f30\u7b97 ) (n w \u039b \uff0c\u901a\u5e38\ufa26\u5047\u8a2d\u566a\u97f3\u5728\u67d0\u6bb5\u6642\u9593\u5167 \u8b8a\u5316\u5f88\u6162\uff0c\u56e0\u70ba\u5be6\u969b\u4e00\u822c\u74b0\u5883\u4e2d\u566a\u97f3\u5f97\u8b8a\u5316\u5176\u5be6\uf967\u5927(\u5982\u6c7d\uf902\u5167\u3001\u5ba4\u5167\uf92e\u6c23\u8072\u2026\u7b49)\u3002\u6240\u4ee5\u5728\u96dc\u8a0a\u8a9e\u97f3\u4e2d\uff0c\u524d \u4e00\u6bb5\u7684\u566a\u97f3\u8cc7\u8a0a\u53ef\u4ee5\u5b58\u8d77\uf92d\u7d66\u4e0b\u4e00\u500b\u8a9e\u97f3\u6bb5\u4f7f\u7528\u4ee5\u6c42\u51fa ) (n Y \u039b \u3002\u6240\u4ee5\u5728\u7528\u8ffd\u8e64\u6f14\u7b97\u6cd5\u4f30\u7b97 ) (n w \u039b \u6642\u7528\u6307\uf969 \u5f0f\u7684\u65b9\u5f0f\uf92d\u758a\u52a0\u5e73\u5747\uff0c\u4ee5\u9054\u5230\u6b63\u78ba\u771f\u5be6\u7684 ) (n w \u039b \uff0c\u5176\u6307\uf969\u5f0f\u7684\u65b9\u5f0f\uf92d\u758a\u52a0\u5e73\u5747\u5982\u4e0b ) ( ) ( ) 1 ( ) ( n W n U n n T w w + \u2212 \u039b = \u039b \u03b2",
                        "eq_num": "("
                    }
                ],
                "section": "\u524d\u8a00",
                "sec_num": "1."
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "A signal subspace approach for speech enhancement",
                "authors": [
                    {
                        "first": "Y",
                        "middle": [],
                        "last": "Ephraim",
                        "suffix": ""
                    },
                    {
                        "first": "H",
                        "middle": [
                            "L"
                        ],
                        "last": "Van-Trees",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "IEEE Trans. Speech Audio Processing",
                "volume": "3",
                "issue": "",
                "pages": "251--266",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Y. Ephraim and H. L. Van-Trees, \"A signal subspace approach for speech enhancement,\" IEEE Trans. Speech Audio Processing, vol. 3, pp. 251-266, July 1995.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Projection approximation subspace tracking",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Yang",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "IEEE Trans. Signal Processing",
                "volume": "43",
                "issue": "",
                "pages": "95--107",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "B. Yang, \"Projection approximation subspace tracking,\" IEEE Trans. Signal Processing, vol. 43, pp. 95-107, Jan. 1995.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "Speech Enhancement Using Perceptual Wavelet Packet Decomposition and Teager Energy Operator",
                "authors": [
                    {
                        "first": "Jhing-Fa",
                        "middle": [],
                        "last": "Shi-Huang Chen",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "The Journal of VLSI Signal Processing Systems, Special Issue on Real World Speech Processing",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Shi-Huang Chen and Jhing-Fa Wang, \"Speech Enhancement Using Perceptual Wavelet Packet Decomposition and Teager Energy Operator,\" accepted to appear in The Journal of VLSI Signal Processing Systems, Special Issue on Real World Speech Processing.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Auditory model and human performance in tasks related to speech coding and speech recognition",
                "authors": [
                    {
                        "first": "O",
                        "middle": [],
                        "last": "Ghitza",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "IEEE Trans. Speech and Audio Processing",
                "volume": "2",
                "issue": "",
                "pages": "115--132",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "O. Ghitza, \"Auditory model and human performance in tasks related to speech coding and speech recognition,\" IEEE Trans. Speech and Audio Processing, vol. 2, pp. 115-132, 1994.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Fundamentals of Speech Recognition",
                "authors": [
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Rabiner",
                        "suffix": ""
                    },
                    {
                        "first": "Biing-Hwang",
                        "middle": [],
                        "last": "Juang",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lawrence Rabiner and Biing-Hwang Juang, Fundamentals of Speech Recognition. Englewood Cliffs, NJ: Prentice-Hall, 1993",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "TAICAR -A Collection of In-Car Mandarin Speech Database in Taiwan",
                "authors": [
                    {
                        "first": "Jhing-Fa",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Hsien-Chang",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Chung-Hsien",
                        "middle": [],
                        "last": "Yang",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "O-COCOSDA2003 / PACLIC17",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Jhing-Fa Wang, Hsien-Chang Wang and Chung-Hsien Yang, \"TAICAR -A Collection of In-Car Mandarin Speech Database in Taiwan,\" O-COCOSDA2003 / PACLIC17, Singapore.",
                "links": null
            }
        },
        "ref_entries": {
            "TABREF0": {
                "text": "Van-Trees \u65bc 1995 \uf98e\u63d0\u51fa\u4e00\u5957\u57fa\u65bc\u8a0a\u865f\u5b50\u7a7a\u9593\u5206\u89e3\u7684\u8a9e\u97f3\u5f37\u5316\u7cfb\u7d71 [1]\uff0c\uf9dd\u7528\u566a\u97f3\u80fd\uf97e\u662f\u5747\u52fb \u5206\u4f48\u65bc\u8a0a\u865f\u6240\u5728\u7684\u5411\uf97e\u7a7a\u9593\u800c\u8a9e\u97f3\u8a0a\u865f\u80fd\uf97e\u5247\u662f\u5206\u4f48\u65bc\u67d0\u4e00\u5b50\u7a7a\u9593\u7684\u7279\u6027\uff0c\u85c9\u7531\u7279\u5fb5\u5206\u89e3\uf92d\u5206\u6790\u51fa\u8a9e\u97f3\u8a0a\u865f \u53ca\u80cc\u666f\u566a\u97f3\uff0c\u4e26\u9032\u4e00\u6b65\u7528\u4e00\u7dda\u6027\u4f30\u6e2c\u5668\uf92d\u8655\uf9e4\u5f97\u5230\u5f37\u5316\u5f8c\u7684\u8a9e\u97f3\u3002\u7531\u65bc\u7279\u5fb5\u5206\u89e3\u7684\u904b\u7b97\u8907\u96dc\ufa01\u9ad8\uff0c\u5728\u672c\uf941\u6587 \u4e2d\u63a1\u7528\u5b50\u7a7a\u9593\u8ffd\u8e64(Subspace Tracking)\u7684\u65b9\u5f0f\uf92d\u505a\u7279\u5fb5\u5206\u89e3\uff0c\u9019\u500b\u6f14\u7b97\u6cd5\u7a31\u70ba PAST (Projection Approximation Subspace Tracking, PAST) [2]\uff0c\u4ee5\u671f\u80fd\u7b26\u5408\u5373\u6642(Real-Time)\u7684\u61c9\u7528\u3002\u800c\u5728\uf902\u5167\u566a\u97f3\u74b0\u5883\u4e2d\uff0c\u566a\u97f3\u80fd\uf97e\u7684\u5206\u4f48",
                "num": null,
                "content": "<table><tr><td colspan=\"11\">\u5f37\u5316\u65b9\u6cd5\uf92d\u514b\u670d\u6b64\u4e00\u554f\u984c\u3002\u6b64\u4e00\u807d\u89ba\u5206\u983b\u8655\uf9e4\u4fc2\uf9dd\u7528\u5c0f\u6ce2\u8f49\u63db(Wavelet Transform)\uf92d\u5be6\u73fe\uff0c\u85c9\u7531\u5c0f\u6ce2\u5c07\u8072\u97f3 \u6b64\u4e8c\u66f2\u7dda\u8a2d\u8a08\uff0c\u5716\u4e8c(a)\u53ca\u5716\u4e8c(b)\u5167\u4ea6\u6a19\u793a\uf9ba\uf9dd\u7528\u5c0f\u8f49\u63db\u903c\u8fd1\u5df4\u514b\u983b\u8b5c\u53ca\u95dc\u9375\u983b\u5bec\u7684\u66f2\u7dda\u5716\u3002</td></tr><tr><td colspan=\"11\">\u5206\u89e3\u6210\u591a\u500b\u983b\u5e36\uff0c\u800c\u5404\u500b\u5b50\u983b\u5e36\u7684\u5206\u4f48\u5247\u7b26\u5408\u4eba\u8033\u807d\u89ba\u97ff\u61c9\u7684\u7279\u6027\uff0c\u5404\u5b50\u983b\u5e36\u7684\u8a0a\u865f\u518d\u7d93\u7531\u5b50\u7a7a\u9593\u65b9\u6cd5\u9032\ufa08 \u7531\u5c0f\u8f49\u63db\u903c\u8fd1\u5df4\u514b\u983b\u8b5c\u53ca\u95dc\u9375\u983b\u5bec\u662f\u85c9\u7531\u8abf\u6574\u5c0f\u6ce2\u8f49\u63db\u7684\u6a39\uf9fa\u7d50\u69cb\uf92d\u9054\u6210\u3002\u4f9d\u64da\u8868\u4e00\u7684\u95dc\u9375\u983b\u5bec\u5206\u4f48\u60c5</td></tr><tr><td colspan=\"11\">\u566a\u97f3\u6d88\u9664\uff0c\u518d\u7531\u5c0f\u6ce2\u53cd\u8f49\u63db\u5408\u6210\u5404\u5b50\u983b\u5e36\u7684\u8a0a\u865f\uff0c\u9032\u800c\u5f97\u5230\u5f37\u5316\u5f8c\u7684\u8a9e\u97f3\u3002\u5be6\u9a57\u7684\u9a57\u8b49\uff0c\u5247\u662f\u63a1\u7528 TAICAR</td></tr><tr><td colspan=\"11\">\uf902\u5167\u8a9e\u97f3\u8cc7\uf9be\u5eab\uf92d\u9032\ufa08\uff0c\u5be6\u9a57\u7d50\u679c\uf96f\u660e\u672c\u6587\u6240\u63d0\u51fa\u7684\u65b9\u6cd5\u6bd4\u8d77\u50b3\u7d71\u8a0a\u865f\u5b50\u7a7a\u9593\u5f37\u5316\u6cd5\uff0c\uf901\u9069\u7528\u65bc\uf902\u5167\u566a\u97f3\u7684 \u5f62\uff0c\u9069\u7576\u5c0d\u8a0a\u865f\u505a\u9ad8\u4f4e\u983b\u7684\u5206\u89e3\uff0c\u4f7f\u5f97\u5b50\u983b\u5e36\u8a0a\u865f\u7684\u983b\uf961\u5206\u4f48\u8ddf\u95dc\u9375\u983b\u5bec\u8fd1\u4f3c\u3002\u5716\u4e09\u70ba\u6240\u4f7f\u7528\u7684\u5177\u807d\u89ba\u611f\u77e5</td></tr><tr><td colspan=\"10\">\u6d88\u9664\uff0c\u4f4e\u983b\u566a\u97f3\u7684\u6d88\u9664\u4e5f\uf901\u660e\u986f\u3002 \u7684\u5c0f\u6ce2\u8f49\u63db\u5206\u89e3\u67b6\u69cb\u5716\uff0c\u5176\u4e2d\u8f38\u5165\u8a0a\u865f\u7d93\u4e94\u500b\u968e\u6bb5\uff0c\u5171 16 \u6b21\u7684\u9ad8\u4f4e\u983b\u5206\u89e3\u3002</td></tr><tr><td colspan=\"11\">\u672c\uf941\u6587\u7684\u7ae0\u7bc0\u7d50\u69cb\u5982\u4e0b\uff1a\u7b2c\u4e8c\u7bc0\u662f\u6240\u63d0\u51fa\uf92d\u7684\u566a\u97f3\u6d88\u9664\u7cfb\u7d71\u67b6\u69cb\uff0c\u5305\u542b\u5c0f\u6ce2\u807d\u89ba\u5206\u983b\u8655\uf9e4\u3001\u8a0a\u865f\u5b50\u7a7a\u9593\u8a9e</td></tr><tr><td colspan=\"11\">\u97f3\u5f37\u5316\u4ee5\u53ca\u5b50\u7a7a\u9593\u8ffd\u8e64\u6cd5\u4e4b\u63cf\u8ff0\uff1b\u7b2c\u4e09\u7bc0\uf96f\u660e\u5be6\u9a57\u7d50\u679c\uff0c\u5305\u542b TAICAR \uf902\u5167\u8a9e\u97f3\u8cc7\uf9be\u5eab\u7684\u4ecb\u7d39\u4ee5\u53ca\u672c\u6587\u6240\u63d0 \u8868\u4e00\uff1a\u95dc\u9375\u983b\u5bec\u5206\u4f48\u3002</td></tr><tr><td colspan=\"10\">\u51fa\u4e4b\u65b9\u6cd5\u8ddf\u5176\u5b83\u8a0a\u865f\u5b50\u7a7a\u9593\u8a9e\u97f3\u5f37\u5316\u6cd5\u4e4b\u6bd4\u8f03\uff1b\u6700\u5f8c\uff0c\u7b2c\u56db\u7bc0\u5247\u662f\u7d50\uf941\u8207\u8a0e\uf941\u3002 Critical Band Center Frequency CBW Lower Cutoff</td><td>Upper Cutoff</td></tr><tr><td colspan=\"2\">Number</td><td>(Hz)</td><td/><td/><td/><td/><td/><td colspan=\"2\">frequency (Hz)</td><td>Frequency (Hz)</td></tr><tr><td/><td>1</td><td>50</td><td/><td/><td/><td/><td>-</td><td/><td>-</td><td>100</td></tr><tr><td>2. \u7cfb\u7d71\u67b6\u69cb</td><td>2 3</td><td>150 250</td><td/><td/><td/><td/><td>100 100</td><td/><td>100 200</td><td>200 300</td></tr><tr><td/><td>4</td><td>350</td><td/><td/><td/><td/><td>100</td><td/><td>300</td><td>400</td></tr><tr><td colspan=\"11\">\u672c\uf941\u6587\u6240\u63d0\u51fa\u7684\uf902\u5167\u566a\u97f3\u6d88\u9664\u7cfb\u7d71\uff0c\u5982\u5716\u4e00\u6240\u793a\u3002\u5728\u7cfb\u7d71\u524d\u7aef\uff0c\u9ea5\u514b\u98a8\u6240\uf93f\u5230\u7684\u96dc\u8a0a\u8a9e\u97f3\uff0c\u7d93\u7531\u5c0f\u6ce2\u807d\u89ba\uf984 5 450 110 400 510</td></tr><tr><td colspan=\"11\">6 \u6ce2\u7d44(Perceptual Wavelet Filterbank)\u5206\u6210\uf969\u500b\u5b50\u983b\u5e36\u8a0a\u865f\uff0c\u5404\u500b\u5b50\u983b\u5e36\u5247\u7531\u8a0a\u865f\u5b50\u7a7a\u9593\u8a9e\u97f3\u5f37\u5316\uf92d\u9032\ufa08\u566a\u97f3\u6d88 570 120 510 630 7 700 140 630 770</td></tr><tr><td colspan=\"11\">\u9664\u7684\u8655\uf9e4\uff0c\u800c\u8a0a\u865f\u5b50\u7a7a\u9593\u7684\u62c6\u89e3\u5247\u662f\u7531\u5b50\u7a7a\u9593\u8ffd\u8e64\u6cd5\uf92d\u5b8c\u6210\u3002\u7531\u5b50\u7a7a\u9593\u8ffd\u8e64\u6cd5\u6240\u4f30\u7b97\u51fa\uf92d\u7684\u7279\u5fb5\u503c 8 840 150 770 920</td></tr><tr><td colspan=\"11\">9 (Eigenvalue)\uff0c\u5247\u7528\u4ee5\u8a08\u7b97\u5404\u500b\u5b50\u983b\u5e36\u8a0a\u865f\u7684\u589e\ufa17\u503c\u3002\u8a9e\u97f3\u5f37\u5316\u7684\u8655\uf9e4\u70ba\u5c07\u5b50\u983b\u5e36\u8a0a\u865f\u7d93\u904e\u7279\u5fb5\u5411\uf97e 1000 160 920 1080 10 1170 190 1080 1270</td></tr><tr><td colspan=\"11\">(Eigenvector)\u6295\u5f71\u8f49\u63db\u5f8c\uff0c\u7531\u589e\ufa17\u503c\uf92d\u8abf\u6574\u5176\u8a0a\u865f\u5927\u5c0f\uff0c\u518d\u7d93\u904e\u53cd\u8f49\u63db\uf92d\u5f97\u5230\u5f37\u5316\u5f8c\u7684\u8a9e\u97f3\u8a0a\u865f\u3002\u4ee5\u4e0b\u5404\u5c0f 11 1370 210 1270 1480</td></tr><tr><td colspan=\"10\">12 \u7bc0\u5247\u5c0d\u5c0f\u6ce2\u807d\u89ba\u5206\u983b\u8655\uf9e4\u3001\u8a0a\u865f\u5b50\u7a7a\u9593\u8a9e\u97f3\u5f37\u5316\u4ee5\u53ca\u5b50\u7a7a\u9593\u8ffd\u8e64\u6cd5\u505a\u4e00\u63cf\u8ff0\u3002 1600 240 1480 13 1850 280 1720</td><td>1720 2000</td></tr><tr><td colspan=\"2\">14</td><td>2150</td><td/><td/><td/><td/><td>320</td><td/><td>2000</td><td>2320</td></tr><tr><td colspan=\"2\">15 16 17</td><td>NLMS eigenvectors filter output 2500 2900 3400</td><td>*</td><td>+</td><td>+</td><td>-</td><td>380 450 Eigenvector Projection 550</td><td>Gain G 1 Adaptation</td><td>2320 2700 Projection Inverse 3150</td><td>2700 3150 3700</td></tr><tr><td colspan=\"11\">\u6c7d\uf902\u513c\u7136\u5df2\u7d93\u8b8a\u6210\u96a8\u6642\u53ef\u7372\u77e5\u5404\u7a2e\u751f\u6d3b\u8cc7\u8a0a\u7684\ufa08\u52d5\u4e2d\u5fc3\u3002\u5728\u6c7d\uf902\u5167\u50b3\u7d71\u7684\u4eba\u6a5f\u4ecb\u9762\u662f\u63a1\u7528\u89f8\u78b0\u5f0f\u87a2\u5e55\uff0c\u5728\ufa08 \uf902\u7684\uf9fa\u6cc1\u4e0b\uff0c\u9019\u6a23\u7684\u4ecb\u9762\u662f\uf967\u5920\u5b89\u5168\u7684\uff0c\u800c\u96a8\u8457\u5373\u6642\u8a9e\u97f3\u8fa8\uf9fc\u6280\u8853\u7684\u65e5\u8da8\u6210\u719f\uff0c\u4eba\u6a5f\u4ecb\u9762\u5fc5\u5b9a\u662f\u671d\u8457\u8a9e\u97f3\u5c0d Noisy Speech Enhanced Speech NLMS filter output + * + -Gain Adaptation G 2 Eigenvector Projection eigenvectors Inverse Perceptual Perceptual Projection Filterbank Filterbank \u8a71\u7684\u64cd\u63a7\u65b9\u5f0f\u6539\u9032\u3002\u5728\ufa08\uf902\u74b0\u5883\u4e2d\uff0c\u5145\u65a5\u5404\u7a2e\u566a\u97f3\uff0c\u5c0d\u65bc\u8a9e\u97f3\u8fa8\uf9fc\u7cfb\u7d71\u800c\u8a00\uff0c\u9019\u4e9b\u80cc\u666f\u566a\u97f3\u6703\u56b4\u91cd\u5730\u5f71\u97ff\u8fa8 \uf9fc\u7d50\u679c\u3002\u56e0\u6b64\uff0c\u4e00\u822c\u7684\u8fa8\uf9fc\u7cfb\u7d71\ufa26\u9700\u4f7f\u7528\u624b\u6301\u5f0f\u6216\u982d\u6234\u5f0f\u9ea5\u514b\u98a8\uff0c\uf92d\u4fc3\u6210\u8fd1\u8ddd\uf9ea\u7684\uf93f\u97f3\uff0c\u4ee5\u907f\u514d\u80cc\u666f\u566a\u97f3\u7684 (Analysis) (Synthesis) ... ... ...</td></tr><tr><td colspan=\"11\">\u5e72\u64fe\u3002\u7136\u800c\uff0c\u4f7f\u7528\u9019\u4e9b\uf93f\u97f3\u8a2d\u5099\u6703\u5c0d\u99d5\u99db\u6216\u8005\u4e58\u5ba2\u9020\u6210\uf967\uf965\uff0c\u6240\u4ee5\u63d0\u4f9b\u4e00\u500b\u5728\ufa08\uf902\u74b0\u5883\u4e0b\u80fd\u5be6\ufa08\u9060\u8ddd\uf9ea\uf93f\u97f3</td></tr><tr><td colspan=\"11\">\u4e26\u5177\u6709\u6297\u566a\u97f3\u80fd\uf98a\u7684\u9ea5\u514b\u98a8\u7cfb\u7d71\uff0c\u662f\u6709\u5176\u9700\u6c42\u3002\u672c\uf941\u6587\u63d0\u51fa\u4e00\u500b\uf9dd\u7528\u5c0f\u6ce2\u807d\u89ba\u5206\u983b\u8655\uf9e4\u8207\u8a0a\u865f\u5b50\u7a7a\u9593\u5206\u89e3\uf92d \u9054\u6210\uf902\u5167\u80cc\u666f\u566a\u97f3\u6d88\u9664\u7684\u76ee\u7684\u3002 NLMS filter output + * + -Gain Eigenvector eigenvectors Inverse Adaptation G i Projection Projection \u5716\u4e00\uff1a\uf902\u5167\u566a\u97f3\u6d88\u9664\u7cfb\u7d71\u67b6\u69cb\u3002 2.1. \u5c0f\u6ce2\u807d\u89ba\u5206\u983b\u8655\uf9e4 Ephraim \u548c \u5728\u4f4e\u983b\u5e36\u70ba\u6700\u591a\u5ef6\u4f38\u5230\u9ad8\u983b\u5247\u9010\u6f38\u8f03\u5c11\uff0c\u5728\u5be6\u9a57\u904e\u7a0b\u4e2d\u767c\u73fe\uff0c\u55ae\u4e00\u7684\u8a0a\u865f\u5b50\u7a7a\u9593\u7684\u8a9e\u97f3\u5f37\u5316\u65b9\u6cd5\u5df2\uf967\u80fd\uf901\u6709 \u5177\u807d\u89ba\u611f\u77e5\u7684\u5c0f\u6ce2\u8f49\u63db(Perceptual Wavelet Packet Transform, PWPT)\u662f\u6539\uf97c\u81ea\u50b3\u7d71\u5c0f\u6ce2\u8f49\u63db\uff0c\u4f7f\u8a9e\u97f3\u4fe1\u865f\u7d93 \u6548\u7684\u6d88\u9664\u4f4d\u5728\u4f4e\u983b\u5e36\u7684\u80cc\u666f\u566a\u97f3\u3002\u56e0\u6b64\uff0c\u672c\uf941\u6587\u63d0\u51fa\u4e00\u500b\u57fa\u65bc\u4eba\u8033\u807d\u89ba\u7279\u6027\u7684\u5206\u983b\u8655\uf9e4\uff0c\u4e26\u7d50\u5408\u8a0a\u865f\u5b50\u7a7a\u9593 PWPT \u5206\u89e3\u5f8c\u7684\u5404\u500b\u5b50\u983b\u5e36\u4fe1\u865f\u7684\u983b\u5bec\u63a5\u8fd1\u4eba\u8033\u7684\u807d\u89ba\u97ff\u61c9 [3]\uff0c\u63cf\u8ff0\u4eba\u8033\u807d\u89ba\u97ff\u61c9\u7684\uf96b\uf969\u4e3b\u8981\u6709\u5df4\u514b\u983b\u8b5c</td></tr><tr><td colspan=\"11\">(Bark)\u4ee5\u53ca\u95dc\u9375\u983b\u5bec(Critical Bandwidth) \uff0c\u8868\u4e00\u70ba\u4eba\u8033\u807d\u89ba\u95dc\u9375\u983b\u5bec\u7684\u5206\u4f48\u60c5\u5f62\u3002\u5716\u4e8c(a)\u53ca\u5716\u4e8c(b)\u5206\u5225</td></tr><tr><td colspan=\"11\">\u662f\u5728 4KHz \u5167\uff0c\u4eba\u8033\u7684\u807d\u89ba\u7684\u5df4\u514b\u983b\u8b5c\u53ca\u95dc\u9375\u983b\u5bec\u66f2\u7dda\u5716 [4, 5]\u3002\u56e0\u6b64\uff0c\u672c\uf941\u6587\u6240\u8a2d\u8a08\u7684\u807d\u89ba\u5206\u983b\u8655\uf9e4\u5373\u671d</td></tr></table>",
                "type_str": "table",
                "html": null
            },
            "TABREF2": {
                "text": "\u6548\u80fd\u8a55\u4f30 \u5be6\u9a57\u7684\u9a57\u8b49\u4ee5 TAICAR \u97f3\u6a94\uf92d\u505a\u6e2c\u8a66\uff0c\u5c0d\u9032\ufa08\u904e\u566a\u97f3\u6d88\u9664\u5f8c\u4e4b\u97f3\u6a94\u9032\ufa08\u8a55\u5206\u3002\u8a55\u5206\u65b9\u5f0f\u63a1\u4eba\u8033\u8a66\u807d\u70ba\u4e4b(Mean Opinion Score, MOS)\uff0c\u7d66\u5206\u7b49\u7d1a\u70ba\uff1a5 \u70ba\u512a\uff0c4 \u70ba\u597d\uff0c3 \u70ba\u5c1a\u53ef\uff0c2 \u70ba\uf976\u5dee\uff0c1 \u70ba\uf967\u597d\u3002\u4ee5\u672c\u6587\u6240\u63d0\u4e4b\u65b9\u6cd5\u8207 \u53e6\u5916\uf978\u7a2e\u5b50\u7a7a\u9593\u5206\u89e3\u65b9\u6cd5\u4f5c\u6bd4\u8f03\uff0c\u5176\u70ba\u5b50\u7a7a\u9593\u5206\u89e3\u63a1\u7528\uf9ea\u6563\u9918\u5f26\u8f49\u63db(Discrete Cosine Transform, DCT)\u53ca\u63a1\u7528 KL \u8f49\u63db(Karhunen-Loeve Transform, KLT)\u3002\u8a08\u6709\u4e8c\u5341\u4f4d\u8a66\u807d\u8005\u7d66\u5206\uff0c\u7d66\u5206\u7d50\u679c\u5982\u8868\u4e09\u6240\u793a\u3002 \u8868\u4e09: MOS \u6e2c\u8a66\u8a55\u5206\u3002",
                "num": null,
                "content": "<table><tr><td colspan=\"6\">3.1. TAICAR \uf902\u5167\u8a9e\u97f3\u8cc7\uf9be\u5eab \u5728\u570b\u5916\u6709\u5f88\u591a\u7684\u8a9e\u97f3\u8cc7\uf9be\u5eab\u6536\u96c6\u4e4b\u65b9\u6cd5\uff0c\uf9b5\u5982\uff1a\u65e5\u672c\u7684 CIAIR\u3001\u6b50\u6d32\u7684 SpeechDat \u7b49\u7b49\uff0c\u7136\u800c\u5728\u6c7d\uf902\u74b0\u5883\u4e0b \u7684\u8a9e\uf9be\u6536\u96c6\u537b\u662f\u5f88\u5c11\ufa0a\uff0cTAICAR \u8cc7\uf9be\u5eab\u76ee\u6a19\u5c31\u5728\u65bc\u6536\u96c6\u6c7d\uf902\u74b0\u5883\u4e0b\u7684\u8a9e\uf9be\u4ee5\u65b9\uf965\u5404\u7a2e\u8a9e\u97f3\u8655\uf9e4\u6280\u8853\u4e4b\u958b \u767c\u3002\uf9b5\u5982\uff1a\u566a\u97f3\u88dc\u511f\u6280\u8853\u3001\u566a\u97f3\u4e0b\u52d5\u614b\u8a9e\u97f3\u5075\u6e2c\u6280\u8853\u3001\u5f37\u5065\u578b\u8a9e\u97f3\u8fa8\uf9fc\u6280\u8853\u3001\u8a9e\u97f3\u8abf\u9069\u6280\u8853\u7b49\u3002\u8a9e\uf9be\u7684\uf93f\u97f3 \uf96b\u8003\u570b\u5167\u57f7\ufa08\u904e\u7684\u5927\u578b\u8a08\u756b\u300cMAT \u8a9e\uf9be\u6536\u96c6\u300d\u4e4b\u4f5c\u6cd5\uff0c\u5148\u7531\u7a0b\u5f0f\u5f9e 100 \u842c\u5b57\u7684\u6587\u5b57\u5eab\u4e2d\u6311\u9078\u51fa\u80fd\u5920\u6db5\u84cb\u6240 \u6709\u570b\u8a9e\u57fa\u672c\u97f3\u7bc0\u7684\u77ed\u8a5e\u3001\u55ae\u5b57\u7b49\uff0c\u4e26\u52a0\u4e0a\u82f1\u6587\u3001\uf969\u5b57\u90e8\u5206\uff0c\u7e3d\u5171\u9019\u6a23\u7684\u8a9e\uf9be\u6709 360 \u4efd\u3002\u70ba\uf9ba\u5be6\u969b\u8a18\uf93f\u5404\u7a2e\uf967 \u540c\uf937\u6cc1\uff0c\uf93f\u97f3\u6642\u5206\uf978\u7a2e\uf937\u6bb5\uff1a\u5e02\u5340\uf937\u6bb5\u4ee5\u53ca\u5feb\u901f\u9053\uf937\uf937\u6bb5\u3002\u5e02\u5340\uf937\u6bb5\u4e0b\uff0c\u6642\u901f\u70ba 0\uff5e50 \u516c\uf9e9\uff1b\u5feb\u901f\u9053\uf937\u5247\u9700 \u7dad\u6301\u5728 70\uff5e100 \u516c\uf9e9\u3002 4. \u7d50\uf941 \u672c\uf941\u6587\u63d0\u51fa\u4e00\u500b\u57fa\u65bc\u4eba\u8033\u807d\u89ba\u7279\u6027\u7684\u5206\u983b\u8655\uf9e4\uff0c\u4e26\u7d50\u5408\u8a0a\u865f\u5b50\u7a7a\u9593\u5f37\u5316\u65b9\u6cd5\uf92d\u6d88\u9664\uf902\u5167\u80cc\u666f\u566a\u97f3\u3002\u6b64\u4e00\u807d\u89ba \u5206\u983b\u8655\uf9e4\u4fc2\uf9dd\u7528\u5c0f\u6ce2\u8f49\u63db\uf92d\u5be6\u73fe\uff0c\u85c9\u7531\u5c0f\u6ce2\u5c07\u8072\u97f3\u5206\u89e3\u6210\u591a\u500b\u983b\u5e36\uff0c\u800c\u5404\u500b\u5b50\u983b\u5e36\u7684\u5206\u4f48\u5247\u7b26\u5408\u4eba\u8033\u807d\u89ba\u97ff \u61c9\u7684\u7279\u6027\uff0c\u5404\u5b50\u983b\u5e36\u7684\u8a0a\u865f\u518d\u7d93\u7531\u5b50\u7a7a\u9593\u65b9\u6cd5\u9032\ufa08\u566a\u97f3\u6d88\u9664\uff0c\u518d\u7531\u5c0f\u6ce2\u53cd\u8f49\u63db\u5408\u6210\u5404\u5b50\u983b\u5e36\u7684\u8a0a\u865f\uff0c\u9032\u800c\u5f97 \u5230\u5f37\u5316\u5f8c\u7684\u8a9e\u97f3\u3002\u5be6\u9a57\u7684\u9a57\u8b49\uff0c\u5247\u662f\u63a1\u7528 TAICAR \uf902\u5167\u8a9e\u97f3\u8cc7\uf9be\u5eab\uf92d\u9032\ufa08\uff0c\u7531 MOS \u8a55\u5206\u53ca\u6642\u9593\u6ce2\u5f62\u548c\u983b\u8b5c\u5716 \uf92d\u770b\uff0c\u672c\u6587\u6240\u63d0\u51fa\u7684\u65b9\u6cd5\u6bd4\u8d77\u50b3\u7d71\u8a0a\u865f\u5b50\u7a7a\u9593\u63a1\u7528 DCT \u53ca KLT \u7b49\u65b9\u6cd5\uff0c\uf901\u9069\u7528\u65bc\uf902\u5167\u566a\u97f3\u7684\u6d88\u9664\uff0c\u4f4e\u983b\u566a\u97f3 3.2. TAICAR \u97f3\u6a94 \u7684\u6d88\u9664\u4e5f\uf901\u660e\u986f\u3002</td></tr><tr><td colspan=\"6\">\u5728\uf902\u5167\uf93f\u97f3\u9700\u8003\u616e\u5230\uf965\uf9dd\u6027\uff0c\u56e0\u6b64\u4ee5\u7b46\u8a18\u578b\u96fb\u8166\u70ba\uf93f\u97f3\u7684\u5e73\u53f0\uff0c\u914d\u5408\u4e0a\u7279\u6b8a\u786c\u9ad4\uf92d\u9032\ufa08\uf93f\u97f3\u3002\u6240\u7528\uf93f\u97f3\u5668 \u6750\u8a08\u6709\uff1a \u5f85\u901f \u5e02\u5340\uf937\u6bb5 \u5feb\u901f\u9053\uf937\uf937\u6bb5 DCT 2.6 2.1 1.9 5. \uf96b\u8003\u6587\u737b</td></tr><tr><td/><td colspan=\"5\">\u7b46\u8a18\u578b\u96fb\u8166: \u8ca0\u8cac\u4e3b\u8981\u7684\uf93f\u97f3\u7a0b\u5f0f\u4e4b\u9032\ufa08 KLT 4.4</td><td>4.1</td><td>3.8</td></tr><tr><td/><td colspan=\"5\">PCMCIA \u4ecb\u9762\u4e4b\u591a\u983b\u9053\u4fe1\u865f\u64f7\u53d6\u5361\uff1a\u8ca0\u8cac\u64f7\u53d6\u591a\u983b\u9053\u7684\u8a9e\u97f3\u8a0a\u865f \u672c\uf941\u6587\u6240\u63d0\u65b9\u6cd5 4.2 4.0</td><td>3.9</td></tr><tr><td/><td colspan=\"5\">\u9ea5\u514b\u98a8\uff1a1 \u652f\u6307\u5411\u6027(\u982d\u6234\u5f0f\uff0c\u6536\uf93f\u4e7e\u6de8\u8a9e\u97f3)\uff0b5 \u652f\u5168\u5411\u6027(\u6536\uf93f\u96dc\u8a0a\u8a9e\u97f3)\uff1a\u8ca0\u8cac\u8a9e\u97f3\u8a0a\u865f</td></tr><tr><td colspan=\"6\">\u7684\u8f38\u5165 \u5716\uf9d1\u5247\u70ba\u96dc\u8a0a\u97f3\u6a94\u7d93\u7531\u4e0a\u8ff0\u4e09\u7a2e\u65b9\u6cd5\u9032\ufa08\u566a\u97f3\u6d88\u9664\u5f8c\u4e4b\u6ce2\u5f62\u53ca\u983b\u8b5c\u5716\u3002\u5f9e\u5716\uf9d1\u89c0\u5bdf\u566a\u97f3\u6291\u5236\u7d50\u679c\uff0c\u4ee5 KLT</td></tr><tr><td colspan=\"6\">\uf902\u8f1b\uff1a\u4efb\u610f \u53ca\u672c\u6587\u6240\u63d0\u65b9\u6cd5\u7686\u512a\u65bc DCT \u7684\u6548\u679c\uff0c\u518d\u5f9e\u4f4e\u983b\u5e36\u7684\u566a\u97f3\u6d88\u9664\uf92d\u770b\uff0c\u5247\u662f\u4ee5\u672c\u6587\u6240\u63d0\u7684\u65b9\u6cd5\u70ba\u6700\u597d\u3002</td></tr><tr><td colspan=\"6\">\uf902\u5167\u7684\uf93f\u97f3\u8edf\u9ad4\uff0c\u53ef\u540c\u6642\u9032\ufa08 6 \u500b channel \uf93f\u97f3\uff0c\u55ae\u97f3\u53d6\u6a23: 16KHz , 16bits\u3002\u5728\uf93f\u97f3\u4e4b\u540c\u6642\u53ef\u6a19\u8a18\uf937\u6cc1\u3001\uf902\u901f\u3001</td></tr><tr><td colspan=\"6\">\u8a9e\u8005\u6027\u5225\u3001\u57fa\u672c\u8cc7\uf9be\u7b49 [6]\u3002\u5716\u56db\u70ba TAICAR \uf902\u5167\u97f3\u6a94\uf93f\u97f3\u60c5\u6cc1\uff0c\u5716\u4e94\u5247\u70ba\u6240\uf93f\u5f97\u97f3\u6a94\u4e4b\u6642\u9593\u6ce2\u5f62\u3002</td></tr><tr><td/><td/><td/><td colspan=\"3\">(a)\u96dc\u8a0a\u97f3\u6a94\u6ce2\u5f62\u53ca\u983b\u8b5c\u5716</td></tr><tr><td/><td>(a)</td><td/><td/><td/><td>(b)</td></tr><tr><td/><td colspan=\"5\">\u5716\u56db\uff1a(a) \uf902\u5167\u591a\u9ea5\u514b\u98a8\u914d\u7f6e\u53ca(b)\u8a9e\u8005\u8207\uf93f\u97f3\u8a2d\u5099\u3002</td></tr><tr><td/><td/><td/><td/><td/><td>15)</td></tr><tr><td colspan=\"2\">\u03b2 \u70ba\u4e00\u5e73\uf904\u4fc2\uf969(smoothing factor)\u4e14</td><td>W</td><td>(n</td><td>)</td><td>\u7684\u503c\u5f9e\u4e4b\u524d\u7684\u975c\u97f3\u5340\u6bb5\u9078\u64c7\u4ee3\u7528\u3002\u6240\u4ee5\u53ef\u4ee5\u8a08\u7b97\u566a\u97f3\u7684\u80fd\uf97e\u5728\u6bcf</td></tr><tr><td colspan=\"6\">\u500b\u8a9e\u97f3\u6bb5\u4e4b\u9593\uff0c\u5176\u8a08\u7b97\u51fa\uf92d\u7684\u566a\u97f3\u80fd\uf97e\u8cc7\u8a0a\u76f4\u63a5\u7d66\u4e0b\u500b\u8a9e\u97f3\u6bb5\u4f7f\u7528\u3002</td></tr><tr><td>3. \u5be6\u9a57</td><td>(b) DCT</td><td/><td/><td/><td>(c) KLT</td><td>(d)\u672c\u6587\u6240\u63d0\u65b9\u6cd5\u4e4b\u7d50\u679c</td></tr><tr><td/><td colspan=\"5\">\u5716\uf9d1\uff1a\u566a\u97f3\u6d88\u9664\u7d50\u679c\u6bd4\u8f03\u4e4b\u6ce2\u5f62\u53ca\u983b\u8b5c\u5716\u3002</td></tr><tr><td colspan=\"6\">\u5c0d\u65bc\u672c\u6587\u6240\u63d0\u51fa\u7684\u65b9\u6cd5\uff0c\u5247\u662f\u63a1\u7528 TAICAR \uf902\u5167\u8a9e\u97f3\u8cc7\uf9be\u5eab\uf92d\u9032\ufa08\u5be6\u9a57\u7684\u9a57\u8b49\u3002\u4ee5\u4e0b\u5c31\u5c0d TAICAR \u8cc7\uf9be\u5eab</td></tr><tr><td colspan=\"6\">\u505a\u4e00\u4ecb\u7d39\uff0c\u63a5\u8457\u5c0d\uf902\u5167\u6240\u8490\u96c6\u7684\u96dc\u8a0a\u97f3\u6a94\u9032\ufa08\u566a\u97f3\u6d88\u9664\u7684\u5be6\u9a57\u3002</td></tr><tr><td/><td colspan=\"5\">\u5716\u4e94\uff1a\uf902\u5167\uf9d1\u500b\u9ea5\u514b\u98a8\u6240\uf93f\u4e4b\u6642\u9593\u6ce2\u5f62\u3002</td></tr></table>",
                "type_str": "table",
                "html": null
            }
        }
    }
}