An Audio Classification Approach Based On Machine Learning
An Audio Classification Approach Based On Machine Learning
:X 'DQ
&KL)HQJ 8QLYHUVLW\,QQHU 0RQJROLD &KLIHQJ &KLQD
WLQJZDQJ#VLQDFRP
627
6WHS $FFRUGLQJ WR WKH PDMRULW\ YRWLQJ PHWKRG WKH
ILQDO FODVVLILFDWLRQ UHVXOWV DUH REWDLQHG E\ V\QWKHVL]LQJ WKH
UHVXOWV RI LQGLYLGXDO FODVVLILHUV
628
GHWHUPLQLQJ WKH EHVW SDUDPHWHUV GLVFXVVHG LQ WKLV SDSHU VR
WKH WLPH FRQVXPSWLRQ LV JUHDWO\ UHGXFHG
7DEOH ,QWHJUDWHG OHDUQLQJ FODVVLILFDWLRQ UHVXOW E\ %766
$XGLR FODVV
&ODVVLILFDWLRQ $FFXUDF\
7RWDO QXPEHU 0XVL %DFNJURXQG
UHVXOWV 3XUH YRLFH ,PSXUH YRLFH 0XWH
F PXVLF
3XUH YRLFH
,PSXUH YRLFH
0XVLF
%DFNJURXQG PXVLF
0XWH
&RPSDUHG ZLWK Y 95 DQG ELQDU\ WUHH PXOWL >@ 8]XQHU g =KDQJ ; 6LEDQGD 7 0DFKLQH /HDUQLQJ DQG 5XOH
FODVVLILFDWLRQ DOJRULWKPV WKH FRUH SDUDPHWHUV RI WKHVH WKUHH EDVHG $SSURDFKHV WR $VVHUWLRQ &ODVVLILFDWLRQ -RXUQDO RI WKH
WUDGLWLRQDO PXOWLFODVVLILFDWLRQ PHWKRGV DUH REWDLQHG E\ $PHULFDQ 0HGLFDO ,QIRUPDWLFV $VVRFLDWLRQ
FURVVYDOLGDWLRQ 7KH H[SHULPHQWDO UHVXOWV DUH VKRZQ LQ WDEOH >@ 6KDPL 0 9HUKHOVW : $Q HYDOXDWLRQ RI WKH UREXVWQHVV RI
7KH H[SHULPHQWDO UHVXOWV VKRZ WKDW WKH SURSRVHG H[LVWLQJ VXSHUYLVHG PDFKLQH OHDUQLQJ DSSURDFKHV WR WKH
DOJRULWKP KDV WKH KLJKHVW FODVVLILFDWLRQ DFFXUDF\ FRPSDUHG FODVVLILFDWLRQ RI HPRWLRQV LQ VSHHFK 6SHHFK &RPPXQLFDWLRQ
ZLWK WKH WUDGLWLRQDO PXOWLFODVVLILFDWLRQ PHWKRGV Y /95
DQG ELQDU\ WUHH ,W DOVR VKRZV WKDW %766 DOJRULWKP FDQ >@ /\RQV:HLOHU - 3DWHO 6 %KDWWDFKDU\D 6 $ &ODVVLILFDWLRQ
DFKLHYH KLJKHU FODVVLILFDWLRQ DFFXUDF\ WKURXJK HQVHPEOH %DVHG 0DFKLQH /HDUQLQJ $SSURDFK IRU WKH $QDO\VLV RI
OHDUQLQJ VKRZLQJ WKH DGYDQWDJHV RI HQVHPEOH OHDUQLQJ *HQRPH:LGH ([SUHVVLRQ 'DWD *HQRPH 5HVHDUFK
7DEOH &ODVVLILFDWLRQ DFFXUDF\ FRPSDULVRQ ZLWK RWKHU
>@ $OL : 6KDPVXGGLQ 6 0 ,VPDLO $ 6 ,QWHOOLJHQW :HE SUR[\
FODVVLILFDWLRQ DOJRULWKPV
FDFKLQJ DSSURDFKHV EDVHG RQ PDFKLQH OHDUQLQJ WHFKQLTXHV
5HFRJQLWLRQ DFFXUDF\ 'HFLVLRQ 6XSSRUW 6\VWHPV
&ODVV
'DWDVHW %LQDU\ %76 >@ *DVWDOGR 3 3LQQD / 6HPLQDUD / HW DO $ WHQVRUEDVHG
QXPEHU Y YU
WUHH 6 DSSURDFK WR WRXFK PRGDOLW\ FODVVLILFDWLRQ E\ XVLQJ PDFKLQH
9HKLFOH OHDUQLQJ 5RERWLFV $XWRQRPRXV 6\VWHPV
9RZHO 3
*ODVV >@ &$55 $QGUHZ ' /$&++$% HW DO 0DFKLQH OHDUQLQJ
'DUSD DSSURDFK IRU VWUXFWXUHEDVHG ]HROLWH FODVVLILFDWLRQ
0LFURSRURXV 0HVRSRURXV 0DWHULDOV
9 &21&/86,216
>@ (VVLG 6 5LFKDUG * 'DYLG % ,QVWUXPHQW UHFRJQLWLRQ LQ
'XH WR WKH FKDUDFWHULVWLFV RI DXGLR LQIRUPDWLRQ DXGLR SRO\SKRQLF PXVLF EDVHG RQ DXWRPDWLF WD[RQRPLHV ,(((
FODVVLILFDWLRQ WHFKQRORJ\ SOD\V DQ LPSRUWDQW UROH LQ DXGLR 7UDQVDFWLRQV RQ $XGLR 6SHHFK /DQJXDJH 3URFHVVLQJ
VWUXFWXULQJ ZKLFK LV GLUHFWO\ UHODWHG WR WKH GHSWK DQG
DFFXUDF\ RI DXGLR LQIRUPDWLRQ DQDO\VLV DQG WKH H[WUDFWLRQ RI >@ -~QLRU $ + ' 6 &RURQD ) %DUUHWR * $ HW DO 0LQLPDO
DXGLR FRQWHQW VHPDQWLFV $FFRUGLQJ WR WKH KLHUDUFKLFDO /HDUQLQJ 0DFKLQH $ QRYHO VXSHUYLVHG GLVWDQFHEDVHG
VWUXFWXUH RI DXGLR WKLV SDSHU GLVFXVVHV WKH FKDUDFWHULVWLFV RI DSSURDFK IRU UHJUHVVLRQ DQG FODVVLILFDWLRQ 1HXURFRPSXWLQJ
DXGLR IUDPHV DQG DXGLR VHJPHQWV LQ GHWDLO 7KH DSSURDFK
ZLWK H[WUDFWLQJ PHDVXUHV DQG H[WUDFWLQJ WKH IHDWXUHV RI >@ /LX 4 6XQJ $ + 4LDR 0 1HLJKERULQJ MRLQW GHQVLW\EDVHG
-3(* VWHJDQDO\VLV $FP 7UDQVDFWLRQV RQ ,QWHOOLJHQW 6\VWHPV
DXGLR IUDPH DQG DXGLR VHJPHQW DUH SURYLGHG UHVSHFWLYHO\
7HFKQRORJ\
7KH PHWKRG RI HQVHPEOH OHDUQLQJ LV XVHG WR JHW KLJKHU >@ 3KDP % 7 3UDNDVK , %XL ' 7 6SDWLDO SUHGLFWLRQ RI
GLDJQRVWLF DFFXUDF\ E\ LQWHJUDWLQJ PXOWLSOH LQGLYLGXDO ODQGVOLGHV XVLQJ D K\EULG PDFKLQH OHDUQLQJ DSSURDFK EDVHG RQ
FODVVLILHUV 7KH H[SHULPHQWV RQ 8&, GDWDVHWV VKRZ WKDW WKH 5DQGRP 6XEVSDFH DQG &ODVVLILFDWLRQ DQG 5HJUHVVLRQ 7UHHV
SURSRVHG PHWKRG KDV KLJKHU FODVVLILFDWLRQ DFFXUDF\ WKDQ *HRPRUSKRORJ\
WUDGLWLRQDO PXOWLFODVVLILFDWLRQ PHWKRGV VXFK DV Y DQG >@ 1DUGXFFL ) 0XVWR & *HPPLV 0 ' HW DO 793URJUDP
VLQJOH ELQDU\ WUHH 5HWULHYDO DQG &ODVVLILFDWLRQ $ &RPSDULVRQ RI $SSURDFKHV
5()(5(1&(6 EDVHG RQ 0DFKLQH /HDUQLQJ ,QIRUPDWLRQ 6\VWHPV )URQWLHUV
629