æ¬è¨äºã¯ãå½ç¤¾ãªã¦ã³ãã¡ãã£ã¢ãDoorsãã«ç§»è»¢ãã¾ããã
ç´5ç§å¾ã«èªåçã«ãªãã¤ã¬ã¯ããã¾ãã
ããã«ã¡ã¯ãã¢ããªãã£ã¯ã¹ãµã¼ãã¹æ¬é¨ã®äºåºã§ãã
ä»åã®ããã°ã¯ãé³å£°ãã¼ã¿ããã¼ãã¨ãã¦åãä¸ãã
- é³å£°ãã¼ã¿ã®ç¹å¾´é
- 深層å¦ç¿ã«ãã楽æ²ã®ã¢ã¼ãã£ã¹ãåé¡
ã«ã¤ãã¦ãç´¹ä»ãã¾ãã
ãã¬ã¤ã³ãããã§ã¯ã深層å¦ç¿ã®æè¡ãé§ä½¿ããæ´»ç¨äºä¾ãå¢ãã¦ãã¦ãã¾ããç¹ã«ãç»åèªèã®åéã«ãããæ´»ç¨äºä¾ã¯å¤ããå½ç¤¾ã®å ¬éããã¦ããäºä¾ã ãã§ã以ä¸ã®ãããªãã®ãããã¾ãã
- ãã¬ã¤ã³ãããããã¦ã¼ãã¼ã®é£åå·¥å ´ã«ãããä¸è¯åã®æ¤ç¥ããã£ã¼ãã©ã¼ãã³ã°ã«ããç»å解æã§æ¯æ´
- å «å代ã¨ã³ã¸ãã¤ãªã³ã°ã¨ãã¬ã¤ã³ãããã洪水ãå®å ¨ã«æµãå½¹å²ãæ ãæ²³å·ã®ã³ã³ã¯ãªã¼ãè·å²¸ã®å£åãAIã§èªåå¤å®ãããµã¼ãã¹ãGoganGoããå ±åéçº
ä¸æ¹ã§ãé³å£°ãã¼ã¿ãæ±ãäºä¾ã¯ãªããªãçããå°è±¡ã§ãã
çè ã¯é³æ¥½ãèããã¨ã好ãã ã¨ãããã¨ã¨ãé³å£°ãã¼ã¿ããã¼ãã¨ãã深層å¦ç¿ã«ä»¥åããèå³ããããåå¼·ã®ä¸ç°ã¨ãã¦èªå® ã§æ¥½æ²åé¡ããã¼ãã¨ãã¦åãçµãã§ãã¾ãããä»åã¯ãã®çµæã«ã¤ãã¦ãç´¹ä»ããã¦ããã ãã¾ãã
ã¾ããé³å£°ãã¼ã¿ã«ã¤ãã¦ä¸æ £ããªäººãå¤ãã¨æãã¾ãã®ã§ãããã°ã®ååã§åºç¤çãªå 容ããç´¹ä»ãã¦ãã¾ãã
åºç¤çãªå 容ã¯å¿ è¦ç¡ãã¨ããæ¹ã¯ãå¾åããèªãã§ããã ããã°å¹¸ãã§ãã
â ç®æ¬¡
- â ç®æ¬¡
- â é³æ¥½ãã¼ã¿ã«ã¤ãã¦
- â ã¡ã«å°ºåº¦å¨æ³¢æ°
- â ãã¼ãªã¨å¤æ
- â ã¡ã«å¨æ³¢æ°ã¹ãã¯ããã°ã©ã
- â 楽æ²ã®ã¢ã¼ãã£ã¹ãåé¡ããã£ã¦ã¿ã
- â ã¾ã¨ã
- â åèæç®
â é³æ¥½ãã¼ã¿ã«ã¤ãã¦
é³ã¯æ³¢ã¨ãã¦èãããã¨ãã§ãã¾ããæ®æ®µãç§ãã¡ãèãã¦ããé³ã¨ããã®ã¯çºçããé³ã空æ°ä¸ã«æ³¢ã¨ãã¦ä¼ããããã®æ³¢ã人éã®è³ã®é¼èãæºãããã¨ã§é³ã¨ãã¦æç¥ãã¾ãããã®é³ã®æ³¢ãæ¯ããå¹ ï¼æ¯å¹ ï¼ã大ããã»ã©ç§ãã¡ã«ã¯å¤§ããé³ã¨ãã¦èãããæºãã¦ããã¹ãã¼ãï¼æ³¢é·ï¼ãæ©ãã»ã©ãé«ãé³ã¨ãã¦èããã¾ãã é³ã®æ³¢ãè¦ããå½¢ã§è¡¨ãã¨ãæ¦ãå¨æçãªä»¥ä¸ã®ãããªæ³¢å½¢ã¨ãªãã¾ãã
ãããããã®é³ã®æ³¢ã¯é£ç¶ä¿¡å·ï¼ã¢ããã°ï¼ã§ãããããã³ã³ãã¥ã¼ã¿ã§ã¯ãã®é³ã®æ³¢ããã®ã¾ã¾æ±ããã¨ã¯ã§ãã¾ãããã³ã³ãã¥ã¼ã¿ã§é³ã®æ³¢ããã¼ã¿ã¨ãã¦æ±ãããã«ã¯ãä¸å®ã®ééã§æ¸¬å®ãããã¨ã«ããé¢æ£ä¿¡å·ï¼ãã¸ã¿ã«ï¼ã«å¤æããå¿ è¦ãããã¾ãããããã¢ããã°-ãã¸ã¿ã«ï¼ADï¼å¤æã¨ããã¾ãã
é³å£°ãADå¤æããããã«ã¯ãæ¨æ¬åï¼ãµã³ããªã³ã°ï¼ã¨éååãå¿ è¦ã¨ãªãã¾ããæ¨æ¬åï¼ãµã³ããªã³ã°ï¼ã¨ã¯æéçã«é£ç¶ããä¿¡å·ãé¢æ£åãããã¨ã§ãããéååã¨ã¯ä¿¡å·ã®å¤§ãããé¢æ£åãããã¨ãããã¾ãã å ç¨ã®é³ã®æ³¢ãä¾ã¨ããã¨ãä¸å³ã«ãããæ¨æ¬åã®çµæã赤ç¹ãæ¨æ¬åã«å ãã¦éååãè¡ã£ãçµæãéç¹ã«è©²å½ãã¾ãã
以ä¸ã®ããã«ãã¦å¾ãããé³å£°ãã¼ã¿ã¯ãæ£å¼¦æ³¢ã¨ãã¦æ°å¼åãããã¨ãã§ãã¾ãã
ã¯æéã¤ã³ããã¯ã¹ãã¯æ¯å¹ ãã¯ãµã³ããªã³ã°å¨æï¼ãµã³ãã«ã®åå¾ééï¼ãã¯åæä½ç¸ã表ãã¾ããã¾ããã¯è§å¨æ³¢æ°ã§ãããæ³¢é· ã¨æ³¢ã®ä½ç¸é度ããæ±ãããã¾ãã
ããã§ã¯ãæéãé¢æ£åããããæéã¤ã³ããã¯ã¹ã«ç½®ãæãã¦ãã¾ãããã¨æãåããããã¨ã§æéãå¾ããã¾ããã¾ããã®éæ°ã¯ãµã³ããªã³ã°å¨æ³¢æ°ï¼1ç§éã®ãµã³ãã«æ°ï¼ã§ãã
ä¸è¬çã«CDé³æºã¯ãµã³ããªã³ã°å¨æ³¢æ°ã44,100Hzã§ããã¤ã¾ãã1ç§éã«44,100åã®é³ãè¨é²ãã¦ãã¾ããä¸æ¹ããããããã¤ã¬ã¾é³æºã¯ããã®ãµã³ããªã³ã°å¨æ³¢æ°ã48,000Hz以ä¸ã§ããå¿ è¦ãããã¾ãã
ã¾ããé³ãéååããéã«ãã¼ã¿ãä½æ®µéã®æ°å¤ã§è¡¨ç¾ãããã示ãå¤ãéååãããæ°ã¨ããã¾ããä¸è¬çã«CDé³æºã¯ãã®éååãããæ°ã16bitã§ãããé³ã2ã®16ä¹=65,536段éã«å解ãã¦è¨é²ãã¦ãã¾ãããã¤ã¬ã¾é³æºã¯éååãããæ°ã16bit以ä¸ã§ããå¿ è¦ãããã¾ãã
ä¸è¨ã®2ã¤ã®æ¡ä»¶ãæºãããã¼ã¿ããã¤ã¬ã¾é³æºã¨ãã¦JEITAãå®ç¾©ãã¦ãã¾ãã
ãã¤ã¬ã¾é³æºã¯CDé³æºã¨æ¯è¼ãã¦ããµã³ããªã³ã°å¨æ³¢æ°ã大ãããããã¨ã§é³åã®å¹ ãåºãããéååãããæ°ã大ãããããã¨ã§é³ãç´°ããèãããããã«ãªãã¾ãããããã®å¹æã«ãã£ã¦ãã¹ã¿ã¸ãªã®åæ²ã®åç¾åº¦ãé«ããªãã¨ãããã¨ã§ããã
â ã¡ã«å°ºåº¦å¨æ³¢æ°
人éã®è³ã«ã¯å®éã®å¨æ³¢æ°ã®é³ããã®ã¾ã¾èãããããã§ã¯ãªãããããçãããããã§ããã¡ãªã¿ã«ã人éã®é³é«ç¥è¦ã«ã¤ãã¦ã¯ããã£ã¦ããªãã¨ãããå¤ããããªã®ã§ãããå¯è´åã®ä¸éã«è¿ãé³ã¯å®éã®é³ãããä½ãã«èãããããã§ãã
ãã®ããã人éã®ç¥è¦ããé³ã®é«ãã測ã尺度ã«èª¿æ´ããå¨æ³¢æ°ãã¡ã«å°ºåº¦å¨æ³¢æ°ã¨ããã¾ããã¡ã«å°ºåº¦ã§ã¯ã1000Hzãåºæºã¨ãã¦ã以ä¸ã®æ°å¼ã§å®ç¾©ããã¦ãã¾ããã¾ãã ã¯å¨æ³¢æ°ã§ãã
å¨æ³¢æ°ã¨ã¡ã«å°ºåº¦å¨æ³¢æ°ã®é¢ä¿ãå³ã«è¡¨ãã¾ãã
ãã®å³ã«ããã横軸ã¯å¨æ³¢æ°ã縦軸ã¯ã¡ã«å°ºåº¦å¨æ³¢æ°ã§ãããããããããããã«ä½é³åã§ã¯æ¥æ¿ã«ãé«é³åã§ã¯ç·©æ ¢ã«ã¡ã«å°ºåº¦å¨æ³¢æ°ãå¢å ãã¦ãããã¨ãåããã¾ããããã¯äººéãä½é³åã§ã®é³é«ã®å¤åã«ã¯ææã§ãããé«é³åã§ã®é³é«ã®å¤åã«ã¯éæã§ãããã¨ãæå³ãã¦ãã¾ãã
â ãã¼ãªã¨å¤æ
å®éã®é³å£°ãã¼ã¿ã¯ãé常ãè¤æ°ã®å¨æ³¢æ°ã®æ³¢ãéãªã£ããã®ã§ããããã®ãããªæ³¢å½¢ã¨ãªãã¾ãã
ãã®ããã«æ¯å¹ ã¨æ³¢é·ããã©ãã©ã§è¤éãªæ³¢å½¢ããè¤æ°ã®å¨æ³¢æ°ã®åç´ãªæ³¢ã«å解ããæ¹æ³ã¨ãã¦ããã¼ãªã¨å¤æãããã¾ãã
ãã¼ãªã¨å¤æã¯ãæéã®é¢æ°ããè§å¨æ³¢æ°ã®é¢æ°ã¸ã®å¤æã表ãã以ä¸ã®æ°å¼ã§è¡¨ãã¾ããï¼å³å¯ãªé¢æ£ãã¼ãªã¨å¤æã®å®ç¾©ã¨ã¯ç°ãªãã¾ãï¼
ã¯é³å£°ãã¼ã¿ã表ãé¢æ°ãã¯ä»»æã®è§å¨æ³¢æ°ãã¯èæ°åä½ã§ãã
ä¸è¨ã®è¤éãªæ³¢ã¯3ã¤ã®æ³¢, , ãéãããã®ã§ããã¨ããã¨ãåç´ãªæ£å¼¦æ³¢ãã¨ãã¦è¡¨ããããããã¼ãªã¨å¤æã§ã¯ãã®åç´ãªæ³¢ã®æ¯å¹ ã¨è§å¨æ³¢æ°ãåãåºããã¨ãã§ãã¾ãã ãã®ããããã®è¤éãªæ³¢ãåç´ãªæ å ±ã§ãããã¯ã¼ã¹ãã¯ãã«ã¨ãã¦ã以ä¸ã®ããã«ã¾ã¨ãããã¨ãã§ãã¾ãããã¼ãªã¨å¤æã«ãã£ã¦é³å£°ãã¼ã¿ãåç´åãããã¨ã§ã解æãåç´ã«ã§ããããã«ãªããã¨ãå¬ããã¨ããã§ãã
â ã¡ã«å¨æ³¢æ°ã¹ãã¯ããã°ã©ã
é³å£°ãã¼ã¿ã®ç¹å¾´éã¨ãã¦ãã¡ã«å¨æ³¢æ°ã¹ãã¯ããã°ã©ã ã¨ãããã®ãããã¾ãã
ã¡ã«å¨æ³¢æ°ã¹ãã¯ããã°ã©ã ã¯ã以ä¸ã®æé ã§ç®åºãããã¨ãã§ãã¾ãã
- é³å£°ãã¬ã¼ã ã«çªé¢æ°ãé©å¿ããé¢æ£ãã¼ãªã¨å¤æãã¦ãã¯ã¼ã¹ãã¯ãã«ãå¾ã
- ã¡ã«å¨æ³¢æ°ã§åçã«ãªãããã«ã¹ã ã¼ã¸ã³ã°ãã
é³å£°ãã¬ã¼ã ã«çªé¢æ°ãé©å¿ããé¢æ£ãã¼ãªã¨å¤æãã¦ãã¯ã¼ã¹ãã¯ãã«ãå¾ã
é³å£°ãã¼ã¿ã解æããå ´åãä¸è¬ã«ä¸å®ã®æéééã§ç¹å¾´éãåºåãã¾ãããã®éãééãã¨ã«åå²ããã対象ãé³å£°ãã¬ã¼ã ã¨å¼ã³ããã®ééããã¬ã¼ã å¨æã¨å¼ã³ã¾ãããã¬ã¼ã å¨æã¨ä¿¡å·ã®å¨æãä¸è´ããªããã¤ã¾ãã両端ãä¸é£ç¶ãªæ³¢å½¢ã¯ããã®ã¾ã¾å¨æ³¢æ°åæãããã¨é«å¨æ³¢ã«éé³ãã®ãã¾ããããã§ã波形ã«å¯¾ãããã®æå»ãä¸å¿ã«ä¸¡ç«¯ãæ¸è¡°ããããã«çªé¢æ°ãããã¾ãã 代表çãªçªé¢æ°ã§ããããã³ã°çªã¯ä»¥ä¸ã®ããã«è¡¨ããã¾ããã¯çªå ã®ãµã³ãã«æ°ã§ãã
å®éã«é³å£°ãã¼ã¿ã«å¯¾ãã¦çªé¢æ°ãé©ç¨ããéã¯ãé£ç¶ããé³å£°ãã¬ã¼ã ã«å¯¾ãã¦çªé¢æ°ãå°ããã¤ããããªããããã®é½åº¦ãé¢æ£ãã¼ãªã¨å¤æãé©ç¨ãã¦ãããã¨ã«ãªãã¾ããã¾ãããã®ä¸é£ã®å¦çãçæéãã¼ãªã¨å¤æã¨ããã¾ããããã¯ãéã¿ä»ã移åå¹³åãã¨ã£ã¦ãããã¨ã«ç¸å½ããç¹å¾´éã®æ¥æ¿ãªå¤åãæããä¸æ¹ã§ãé£ç¶ãããã¬ã¼ã éã®ç¸é¢ã¯ãã大ãããªãã¾ãã ããã«ãããé³å£°ãã¼ã¿ããä¸å®ã®æéééã§ãã¯ã¼ã¹ãã¯ãã«ãæ½åºãããã¨ãã§ãã¾ãã
ã¡ã«å¨æ³¢æ°ã§åçã«ãªãããã«ã¹ã ã¼ã¸ã³ã°ãã
次ã«ãçæéãã¼ãªã¨å¤æã«ããå¾ãããçµæã«å¯¾ãã¦ã¡ã«ãã£ã«ã¿ãã³ã¯ãé©ç¨ãã¾ããã¡ã«ãã£ã«ã¿ãã³ã¯ã¨ã¯ãã¡ã«å°ºåº¦ã§çééãªç¹å®ã®å¨æ³¢æ°å¸¯ã®ã¿ãæ½åºãããã£ã«ã¿ã®ãã¨ã§ãã
ã¡ã«ãã£ã«ã¿ãã³ã¯ã®ä½ææé ã¯ä»¥ä¸ã®ããã«ãªãã¾ãã
- ä¸éå¨æ³¢æ°ã¨ä¸éå¨æ³¢æ°ãã¡ã«å°ºåº¦å¨æ³¢æ°ã«å¤æãã
- ä¸è¨ã®å¤ãçééã§åå²ããåã®å¤ãç®åºãã
- 2ã®å¤ãã¡ã«å°ºåº¦å¨æ³¢æ°ããå¨æ³¢æ°ã«æ»ãï¼å¾ããããµã¤ãºã®å¨æ³¢æ°ãªã¹ããã¨ããï¼
- åãã£ã«ã¿ã以ä¸ã®ããã«å®ç¾©ããï¼ã¯ã®ãµã³ãã«æ°ï¼
ãã¯ã¼ã¹ãã¯ãã«ã«å¯¾ãã¦å¾ãããã¡ã«ãã£ã«ã¿ãã³ã¯ã®åãã£ã«ã¿ãããããã£ã«ã¿å¾ã®ãã¯ã¼ã足ãåããã¦å¯¾æ°ãåããã¨ã§ãé³å§ã¬ãã«ï¼dBï¼ã¹ã±ã¼ã«ã®æ¬¡å ã®ãã¼ã¿ã¨ãã¦ã¹ã ã¼ã¸ã³ã°ãã¾ããããã«ããä»»æã®æéãã¨ã«å¾ããã人éã®é³é«ç¥è¦ã«èª¿æ´ããç¹å¾´éãã¡ã«å¨æ³¢æ°ã¹ãã¯ããã°ã©ã ã¨ããã¾ãã
ã¾ããä»åã¯ä½¿ç¨ãã¾ãããããã®ã¡ã«å¨æ³¢æ°ã¹ãã¯ããã°ã©ã ãé¢æ£ã³ãµã¤ã³å¤æããçµæããä½æ¬¡å æåãåãåºãããã®ããã¡ã«å¨æ³¢æ°ã±ãã¹ãã©ã ä¿æ°ï¼MFCCï¼ã¨ããã¾ãã
â 楽æ²ã®ã¢ã¼ãã£ã¹ãåé¡ããã£ã¦ã¿ã
深層å¦ç¿ã使ã£ã¦é³å£°ãã¼ã¿ãæ´»ç¨ãã¦ã¿ãããã楽æ²åé¡ã«åãçµã¿ã¾ãããå 容ã¨ãã¦ã¯ã楽æ²ã®ç¹å¾´ã表ãã¡ã«å¨æ³¢æ°ã¹ãã¯ããã°ã©ã ãç¨ãã¦ããã®æ¥½æ²ã®ã¢ã¼ãã£ã¹ããäºæ¸¬ããã¨ããå 容ã§ãã
å¦ç¿ãã深層å¦ç¿ã¢ãã«ãç¨ãã¦2ã¤ã®å®é¨ãå®æ½ãã¾ããã
- 楽æ²ã®ã¢ã¼ãã£ã¹ãåé¡ç²¾åº¦ã®æ¤è¨¼
- é¡ä¼¼ã¢ã¼ãã£ã¹ãåé¡ã®æ¤è¨¼
ãããããä¸è¨ã®åãçµã¿ã«é¢ãã¦ãç´¹ä»ãã¾ãã
ãã¼ã¿ã®æºå
ä»åã¯ã©ã³ãã ã«10çµã®ã¢ã¼ãã£ã¹ããé¸å¥ãã楽æ²åé¡ã®å¯¾è±¡ã¨ãã¾ããã ãããã®æ¥½æ²ããã¼ã¿ã¨ãã¦ã¢ãã«ãå¦ç¿ãããã¨ã«ãªãã®ã§ããã楽æ²ã«ã¯éãããããä¸çªå¤ãã¦ãRed Hot Chili Peppersã®179æ²ã¨ãå¦ç¿ãã¼ã¿ã¨ãã¦ã¯å°ãªãã§ãã
ãã®ããã楽æ²ã1åãã¨ã«åºåããã¨ã§ãã¼ã¿ã®æ°´å¢ãããã¾ãããï¼ããã§ããã¼ã¿ã¨ãã¦ã¯å°ãªãã§ãã・・・ï¼
ã¢ãã«ã«å¦ç¿ãããè¨ç·´ç¨ãã¼ã¿ã¨ã¢ãã«ã®ç²¾åº¦ãæ¤è¨¼ããè©ä¾¡ç¨ãã¼ã¿ãç¨æããéã«ã¯ãçºè¡¨ãããå¹´ãå¤ãã¢ã«ãã ã®æ¥½æ²ãè¨ç·´ç¨ãã¼ã¿ãæ°ããã¢ã«ãã ã®æ¥½æ²ãè©ä¾¡ç¨ãã¼ã¿ã¨ãã¾ããã使ç¨ãããã¼ã¿æ°ã¯ä»¥ä¸ã®ã¨ããã§ãã
é³æ¥½ãã¼ã¿ç¹å¾´éã®å¯è¦å
å®éã«æ¥½æ²ãããã³ã½ã³/ã¹ããããã®ç¹å¾´éãæåã®60ç§åã ãå¯è¦åãã¦ã¿ã¾ããããµã³ããªã³ã°å¨æ³¢æ°ã«ãããã®ã§ããããã¼ã¿ã®æ³¢å½¢ã¯é常ã«ç´°ãããªã£ã¦ãã¾ããããã ãã§ã¯ãä½ããããã¾ãããã
import librosa.display
# Load a flac file from 0(s) to 60(s) and resample to 4.41 KHz
filename = 'ããã³ã½ã³.flac'
y, sr = librosa.load(filename, sr=4410, offset=0.0, duration=60.0)
librosa.display.waveplot(y=y, sr=sr)
ããã³ã½ã³ / ã¹ããã
次ã«ããã®ãã¼ã¿ã®ã¡ã«å¨æ³¢æ°ã¹ãã¯ããã°ã©ã ãå¯è¦åãã¾ãããå種ãã©ã¡ã¼ã¿ã¯é©å½ã«æ±ºãã¦ãã¾ããããã¡ãã¯åæé帯ã«ãããå¨æ³¢æ°ãã¨ã®é³ã®å¼·ããæ¿æ·¡ã¨ãã¦è¡¨ç¾ããã¦ãããè¦è¦çã«ã楽æ²ã®ç¹å¾´ã表ãã¦ããæãããã¾ããã
# n_mels is number of Mel bands to generate
n_mels=128
# hop_length is number of samples between successive frames.
hop_length=2068
# n_fft is length of the FFT window
n_fft=2048
# Passing through arguments to the Mel filters
S = librosa.feature.melspectrogram(y=y, sr=sr, n_mels=n_mels, hop_length=hop_length, n_fft=n_fft)
log_S = librosa.power_to_db(S, ref_power=np.max)
print(log_S.shape)
plt.figure(figsize=(12, 4))
librosa.display.specshow(data=log_S, sr=sr, hop_length=hop_length, x_axis='time', y_axis='mel')
plt.colorbar(format='%+2.0f dB')
plt.title('Mel spectrogram')
plt.tight_layout()
(128, 128)
ã¢ãã«
ä¸è¨ã®ããã«ãå¯è¦åãããç»åèªä½ããã®æ¥½æ²ã®ç¹å¾´ã表ãã¦ããã¨ããã¾ããããã¯ãæç³»åã¨å¨æ³¢æ°ãã¨ã«é³å§ã¬ãã«ï¼dBï¼ã並ãã§ããããã§ããã¤ã¾ããé³å£°ãã¼ã¿ã®åé¡ã¯ãç»åã®åé¡åé¡ã¨ä¼¼ãæãæ¹ãã§ãã¾ãããã®ãããä»å使ç¨ããã¢ãã«ã¯ç»ååé¡ã®åéã§æåãåãã¦ãããCNNï¼Convolutional Neural Networkï¼ãæ¡ç¨ãã¾ããã
é³å£°ãã¼ã¿ãã楽å¨ãåé¡ããã¨ããè«æã§ããCNNãç¨ãããã¦ãããé«ã精度ãå¾ããã¦ãã¾ããä»åã®ã¢ãã«ã§ãæ¬è«æã®ããã«ä¸è¨ã®ãããªã·ã³ãã«ãªã¢ã¼ããã¯ãã£ã¨ãã¾ããã
æ¬ã¢ãã«ã®åºå層ã§ã¯ãsoftmaxãç¨ãã¦ãããã¨ãããå ¥åããé³å£°ãã¼ã¿ã«å¯¾ããã¢ã¼ãã£ã¹ããã¨ã®ç¢ºçãåºåããã¾ããã¢ãã«ããã楽æ²ã®ã¢ã¼ãã£ã¹ããåé¡ããéã«ã¯ãï¼ååºåãããåé³å£°ãã¼ã¿ã«å¯¾ãã¦ã確çãç®åºãããããã®ç¢ºçã®å¹³åãåããã¨ã§ãæãé«ã確çã¨ãªãã¢ã¼ãã£ã¹ããäºæ¸¬çµæã¨ãã¾ããã
楽æ²ã®ã¢ã¼ãã£ã¹ãåé¡
è¨ç·´ç¨ãã¼ã¿ã§å¦ç¿ããã¢ãã«ã®ç²¾åº¦ãè©ä¾¡ç¨ãã¼ã¿ï¼æ¥½æ²ï¼ã«å¯¾ããé©åçãåç¾çãFå¤ã§è©ä¾¡ãã¾ãããçµæã¯ä»¥ä¸ã®ããã«ãªãã¾ããã
Få¤ã¯å¹³åã§0.77ã¨ããçµæãå¾ããã¾ããã
ç¹ã«ã¹ãããã¨Metallicaã¯ã©ã¡ããåç¾ç1.0ã§ãããFå¤ã0.9以ä¸ã¨ããçµæãå¾ããã¾ãããä¸æ¹ã§ãColdplayã¯åç¾çã¯1.0ã§ãã£ãã«ãé¢ãããé©åçã0.41ã¨ä½ããFå¤ã0.59ã¨æãä½ãçµæã¨ãªãã¾ãããã¤ã¾ããColdplayã®æ¥½æ²ã¯å ¨ã¦æ£ããäºæ¸¬ãããã¨ãåºæ¥ã¾ãããããã以å¤ã®ã¢ã¼ãã£ã¹ãã®æ¥½æ²ãColdplayã¨èª¤ã£ã¦äºæ¸¬ãã¦ãã¾ã£ãã¨ãããã¨ã«ãªãã¾ãã
ããã¯çè ã®ä¸»è¦³çãªèå¯ã«ãªãã®ã§ãããColdplayã®åæã¯ã¢ã³ã¼ã¹ãã£ãã¯ã®ã¿ã¼ã使ã£ãããã¯ãµã¦ã³ãã®æ²ãå¤ãå°è±¡ã§ãããè¿å¹´ã¯å¤å½©ãªé³ãå ¥ãè¾¼ãã 壮大ãªããããµã¦ã³ãã®æ²ãå¢ãã¦ãã¦ãã¾ãããã®ãããã¢ãã«ãå¦ç¿ãã楽æ²ã®ä¸ã«ã¯ãããã®ã¸ã£ã³ã«ãæ··åãã¦ãããã¨ãããããã¯ãµã¦ã³ããªæ²ãããããµã¦ã³ããªæ²ã¯æ¯è¼çãColdplayã¨äºæ¸¬ãããããã¢ãã«ã«ãªã£ã¦ããã®ããããã¾ãããå®éãColdplayã¨äºæ¸¬ããã楽æ²ã®ã¢ã¼ãã£ã¹ãã確èªããã¨ãRed Hot Chili PeppersãOasisãªã©ãããã¯ãã³ãã®æ¥½æ²ã誤åé¡ããå¾åãããã¾ããã
é¡ä¼¼ã¢ã¼ãã£ã¹ãåé¡
次ã«ãå¦ç¿ã§ä½¿ç¨ãã¦ããªãã¢ã¼ãã£ã¹ãã®æ¥½æ²ãç¨æããä¸è¨ã®å¦ç¿æ¸ã¿ã¢ãã«ã«ããã¢ã¼ãã£ã¹ãåé¡ãè¡ã£ã¦ã¿ã¾ãããã¤ã¾ããã¢ãã«ã¯ãå ¥åããããã楽æ²ãé¸å¥ãã10çµã®ã¢ã¼ãã£ã¹ãã®ãã¡ãã©ãã«è¿ãã®ããåºåããã¨ãããã¨ã«ãªãã¾ãããã®å®é¨ã§ã¯ãå¦ç¿ã§ä½¿ç¨ãã10çµã®ã¢ã¼ãã£ã¹ãã®ããããã¨ãèè ãä¼¼ã¦ããã¨æãã¦ããã¢ã¼ãã£ã¹ãï¼çµãé¸å¥ãã¦è©ä¾¡ãã¾ãããé¸å¥ããã¢ã¼ãã£ã¹ãã¨é¸å¥çç±ã¯ä»¥ä¸ã®ããã«ãªãã¾ãã
å¾ãããçµæã¯ä»¥ä¸ã®ããã«ãªãã¾ããã
ããã§ã¯ç¸¦è»¸ãäºæ¸¬å¯¾è±¡ã®æ¥½æ²ã®ã¢ã¼ãã£ã¹ãã横軸ãã¢ãã«ã§ç¨æãã10çµã®ã¢ã¼ãã£ã¹ãã®idã表ãã¦ãããããããã®å¤ã¯äºæ¸¬ããã楽æ²æ°ã§ãã 赤æ ã§å²ã£ãç®æããå人çã«ä¼¼ã¦ããã¨æãã¦ããã¢ã¼ãã£ã¹ãã®çµã¿åããã«ãªãã®ã§ãããæ¦ãæ³å®éãã®çµæãå¾ããã¦ãã¾ãã ç¹ã«The Birthdayã¯43.6%ã®æ¥½æ²ããã¼ã«ã«ãåãTHEE MICHELLEGE GUN ELEPHANTã¨åé¡ããããã¨ããã声ã®ç¹å¾´ãæãã¦ãããã¨ãåããã¾ãã
BABYMETALã¯å¯ä¸ã®å¥³æ§ãã¼ã«ã«ã§ããæ±äº¬äºå¤ã¨äºæ¸¬ãããããã¨ãèãã¦ããã®ã§ãããå®éã¯2æ²ã®ã¿ãäºæ¸¬ãããã ãã§ãããä¸æ¹ã§ãåãã¸ã£ã³ã«ã§ããMetallicaã¨äºæ¸¬ããã楽æ²ã9æ²ã¨æãå¤ãã£ããã¨ããã声ã®ç¹å¾´ã ãã§ãªããæ¼å¥ã®ç¹å¾´ãèæ ®ããäºæ¸¬ãããã¦ãããã¨ããããã¾ãã
æ¬ã¢ãã«ã®å¿ç¨ã«ã¤ãã¦
å¦ç¿ãã深層å¦ç¿ã¢ãã«ã¯ã楽æ²ã®ã¢ã¼ãã£ã¹ãæ¤ç´¢çã®ããã«ãã¢ã¼ãã£ã¹ããç´æ¥äºæ³ãããã¿ã¹ã¯ã¸ã®å¿ç¨ãèããã¨ç²¾åº¦ã¯ä½ãå°è±¡ã§ãã
ä¸æ¹ã§ãé¡ä¼¼ã¢ã¼ãã£ã¹ãã®åé¡ã«ããã¦ã¯ãæ¦ãæ³å®éãã®çµæãå¾ããã¦ãããããé¡ä¼¼ã¢ã¼ãã£ã¹ãã®ã¬ã³ã¡ã³ãçã®ãããªãé«ã精度ãæ±ãããã¦ããªãã¿ã¹ã¯ã«å¯¾ãã¦ã¯å¿ç¨ãã§ãããã§ãã
ã¾ããä»åã¯ã¢ã¼ãã£ã¹ãã®åé¡ã¨ããã¢ãã«ãä½æããã®ã§ããããããé³æ¥½ã¸ã£ã³ã«ã®åé¡ã«å¤æ´ãããã¨ã§ãç¨æã§ããå¦ç¿ãã¼ã¿æ°ãå¢ãã精度ã®åä¸ãæå¾ ã§ãã¾ããããã«ããã楽æ²ã¸ã®é³æ¥½ã¸ã£ã³ã«ã®èªåä»ä¸ãªã©ã«ãå¿ç¨ã§ãããã§ãã
楽æ²ã«éãããä¾ãã°ãæ©æ¢°ãçºããé³ããã¼ã¿ã¨ãã¦æ±ããã¨ã§ãæ©æ¢°ã®æ éã«ç¹ããç°å¸¸é³ãæ¤ç¥ãããªã©ããã¸ãã¹ã®ãã¾ãã¾ãªèª²é¡ã«å¯¾ãã¦ãå¿ç¨ãæå¾ ã§ãããã§ããã
â ã¾ã¨ã
ä»åã¯é³å£°ã®ãã¼ã¿ã¨ãã¦ã®æ±ãæ¹ã¨ã深層å¦ç¿ã¢ãã«ã«ãã楽æ²ã®ã¢ã¼ãã£ã¹ãåé¡ã¸ã®å人çãªåãçµã¿å 容ã«ã¤ãã¦ãç´¹ä»ããã¦ããã ãã¾ããã
ãã¬ã¤ã³ãããã§ã¯ããã¼ã¿ã®æ´»ç¨ãæ¨é²ãã¦ãã¾ãããã¼ã¿ã¨ãã£ã¦ãããªã¬ã¼ã·ã§ãã«ãã¼ã¿ãã¼ã¹ãªã©ã§æ±ã表形å¼ã®æ§é åãã¼ã¿ã«éãããä»ååãæ±ã£ãé³å£°ããç»åãè¨èªçã®ããã«æ§ã ãªç¨®é¡ã®ãã¼ã¿ãããã¾ãã ããã¦ããªãã¨ãªãéãã¦ããã ãã®ãã¼ã¿ã§ãã£ã¦ãããããæ´»ç¨ãããã¨ã§ä½ããã®èª²é¡ã®è§£æ±ºã«ç¹ããããããã¾ããã
æ¬ããã°ã§ããã¼ã¿åæã®åãçµã¿ã«ã¤ãã¦èå³ãæã£ã¦ããã ããã幸ãã§ãï¼
â åèæç®
- ç¯ ç° æµ©ä¸, ãé³å£°èªèã(æ©æ¢°å¦ç¿ãããã§ãã·ã§ãã«ã·ãªã¼ãº), è¬è«ç¤¾ (2017).
- JEITA
- Beth Logan, ''Mel Frequency Cepstral Coefficients for Music Modeling'' (2000).
- http://practicalcryptography.com/miscellaneous/machine-learning/guide-mel-frequency-cepstral-coefficients-mfccs/
- LosTanlen.V and Cella.C, ''Deep convolutional networks on the pitch spiral for musical instrument recognition'', ISMIR (2016).
- Keras
- Librosa
å½ç¤¾ã§ã¯ããã¸ãã¹æ¹åã«ã¤ãªãã深層å¦ç¿ãªã©ãæå 端ã®åãçµã¿ãç©æ¥µçã«å®æ½ãã¦ãã¾ããå®éã®ãã¸ãã¹ã§ãæå 端ã®æè¡ãæ´»ç¨ãã¦ã¿ããã¨ããæ¹ã¯ããã²ã¨ã³ããªã¼ãã ããï¼