2018å¹´4æ20æ¥ãDeep Learning Labã主å¬ããã¤ãã³ããé³å£°ã»è¨èªãã¤ãããéå¬ããã¾ãããChainerãæä¾ããPreferred Networksã¨ãAzureã¯ã©ã¦ããæä¾ããMicrosoftã«ãããã¨ã³ã¸ãã¢ã³ãã¥ããã£Deep Learning Labãä»åã¯ãèªç¶è¨èªå¦çãåæé³å£°ãªã©ãé³å£°ã»è¨èªÃ深層å¦ç¿ã®ææ°äºä¾ãç¥è¦ãçºè¡¨ãã¾ããããã¬ã¼ã³ãã¼ã·ã§ã³ãMicrosoft Imagine Cupã¨æ·±å±¤å¦ç¿ãç¨ããé³æºåé¢æè¡ã«ã¤ãã¦ãã«ç»å ´ããã®ã¯ãä½è¤é¦å½¦æ°ã深層å¦ç¿ãç¨ããã人ã®å£°ãé³æ¥½ãªã©ãåé¢ããããç¹å®ã®é³å£°ã®ã¿ãæ½åºããæè¡ãç´¹ä»ãã¾ããã å¦çã³ã³ãã¹ãä¸ç大ä¼åºå ´ã®çµæ´ ä½è¤é¦å½¦æ°ï¼ä»¥ä¸ãä½è¤ï¼ï¼ãããããé¡ããã¾ãããMicrosoft Imagine Cupã¨æ·±å±¤å¦ç¿ãç¨ããé³æºåé¢æè¡ã«ã¤ãã¦ãã¨é¡ãã¦ãä½è¤é¦å½¦ãçºè¡¨ãã¾ããã
Speech processing plays an important role in any speech system whether its Automatic Speech Recognition (ASR) or speaker recognition or something else. Mel-Frequency Cepstral Coefficients (MFCCs) were very popular features for a long time; but more recently, filter banks are becoming increasingly popular. In this post, I will discuss filter banks and MFCCs and why are filter banks becoming increas
readme.md Python ã§é³é¿ä¿¡å·å¦ç spectrum.py 2016-07-16 Takuya Nishimoto (@24motz) åºå³¶çIoTç¸ãã®åå¼·ä¼! IoTLTåºå³¶ vol.2 http://iotlt.connpass.com/event/33441/ $ sudo apt-get install portaudio19-dev python-dev $ curl https://bootstrap.pypa.io/get-pip.py | sudo python $ sudo pip install pyaudio $ sudo pip install numpy Raspberry Pi 3 (Raspbian) 㧠USB Audio Interface ãæ¥ç¶ããã¤ã¯ãã¤ãªã python spectrum.py ãå®è¡ ãªã¼ãã£ãªãããµã¼ã§å ¥åã²ã¤ã³
çµ±è¨ç声質å¤æ (6) 声質å¤æã¢ãã«ã®å¦ç¿ã®ç¶ããä»åãçµ±è¨ç声質å¤æã·ãªã¼ãºã®æçµåã§ãã ä»åã¯ãååå¦ç¿ãã声質å¤æã¢ãã«ã使ã£ã¦æ¬å½ã«å£°ãå¤æã§ããã試ãã¦ã¿ãããååãã£ãGMMã®å¦ç¿ã§ ã«ãããåã³ã³ãã¼ãã³ã 㮠㨠㨠ãå¦ç¿ãã¼ã¿ããæ¨å®ãããç¶æ ã§ããã 㨠ã¯ã ã®ããã«åå²ã§ããããã®å¦ç¿çµæã¯ã clb_slt.gmm clb_slt.gmm_01.npy clb_slt.gmm_02.npy clb_slt.gmm_03.npyã®4ã¤ã®ãã¡ã¤ã«ã«ãã³ãããã¦ããã GMMã«ãã声質å¤æ GMMã«ãã声質å¤æã¯ããä¸ããããã¨ãã®ã®æå¾ å¤ãæ±ãããã¨ã§è¡ãã ããã§ã ã§ãããä»åã¯å°åºã¯çãã¦ãè«æï¼PDFï¼ã®çµæããã®ã¾ã¾åç¨ãããã ãã¨ãã°ãAããã®å£°ãBããã®å£°ã«å¤æãããã¨ãèãããAããã®å£°ããæ½åºããã¡ã«ã±ãã¹ãã©ã ãã©ã¡ã¼ã¿ ã使ã£ã¦ä¸ã®å¼ã§å¤
çµ±è¨ç声質å¤æ (5) scikit-learnã®GMMã®ä½¿ãæ¹ï¼2015/3/22ï¼ã®ç¶ãã ä»åã¯ããããã声質å¤æã¢ãã«ãæ··åã¬ã¦ã¹ã¢ãã«ï¼GMMï¼ã§å¦ç¿ãããï¼ç¬¬3åç®ï¼2015/3/4ï¼ã§è¿°ã¹ãããã«clbããã®å£°ãsltããã®å£°ã«å¤æãããã¨ãåæã«é²ããã ä»ã¾ã§ã声質å¤æã¢ãã«ãGMMã§å¦ç¿ãããã¨æ¸ãã¦ããããåèã«ãã¦ãããã¥ã¼ããªã¢ã«ãèªãã§ãä½ãGMMã§è¡¨ãã®ãããã¾ãã¡ãã³ã¨ããªãã£ããããã§ãå è«æï¼PDFï¼ãå½ãã£ãã¨ããå¼ (6) ãè¦ã¦ããããç解ã§ããã ãã®å¼ãã ãGMMã§ã¢ãã«åããã¦ãããã¨ãããããããã¦ããã® ã¯ãå¤æå 話è ã® t ãã¬ã¼ã ç®ã®ç¹å¾´é ã¨å¤æå 話è ã® t ãã¬ã¼ã ç®ã®ç¹å¾´é ã®çµåãã¯ãã« ã§ãããä»åã¯26次å ã®ã¡ã«ã±ãã¹ãã©ã ç¹å¾´éã使ããã ã¯2人ã®ã¡ã«ã±ãã¹ãã©ã ç¹å¾´éãçµåãã52次å ãã¯ãã«ã«ãªããã¤ã¾ããå¦ç¿ãã
çµ±è¨ç声質å¤æ (4) ãã©ã¬ã«ãã¼ã¿ã®ä½æï¼2015/3/10ï¼ã®ç¶ãã ä»åã¯ããããã声質å¤æã¢ãã«ãæ··åã¬ã¦ã¹ã¢ãã«ã§å¦ç¿ãããï¼ã¨æã£ãã®ã ããã©ããã®åã«scikit-learnã®GMMã©ã¤ãã©ãªã®ä½¿ãæ¹ãç°¡åã«ã¾ã¨ãããã¨ã«ãããæ¬æ ¼çã«ä½¿ãåã«ç°¡åãªãã¼ã¿ã§ä½¿ãæ¹ã確èªãã¦ãããã¨ããã®ã趣æ¨ãscikit-learnã¯æè¿ä½¿ãå§ããã®ã§ä½¿ã£ããã¨ããªãæ©è½ãã¾ã ããããããã æ··åã¬ã¦ã¹ã¢ãã«ï¼GMMï¼ GMMã¯ããã¼ã¿xã®çæããã確çãè¤æ°ã®ã¬ã¦ã¹åå¸ã®éã¿ä»ãåã§è¡¨ãã¢ãã«ã§ããã ããã§ãKã¯ä½¿ç¨ããã¬ã¦ã¹åå¸ã®åæ°ãã¯kçªç®ã®ã¬ã¦ã¹åå¸ã®éã¿ï¼æ··åä¿æ°ï¼ãã¯ãkçªç®ã®ã¬ã¦ã¹åå¸ã®å¹³åãã¯ãã«ãã¯ãkçªç®ã®ã¬ã¦ã¹åå¸ã®å ±åæ£è¡åãæ··åä¿æ°ã¯ãã¹ã¦ã®kã«ã¤ãã¦è¶³ãåãããã¨1ã«ãªãã GMMã®å¦ç¿ã¯ããã¼ã¿ã»ããXãç¨ãã¦ã尤度ããã£ã¨ãé«ããªãæ··åä¿æ°ã¨å¹³åãã¯ã
çµ±è¨ç声質å¤æ (3) ã¡ã«ã±ãã¹ãã©ã ã®æ½åºï¼2015/3/4ï¼ã®ç¶ãã ååã¯å¤æå ã®clbããã¨å¤æå ã®sltããã®ã¡ã«ã±ãã¹ãã©ã ãä¸æ¬æ½åºãããååã®æå¾ã®çµæãè¦ãã¨ãäºäººã®ããã¹ãéããéãããã¡ã«ã±ãã¹ãã©ã ãæéæ¹åã«ããã¦ãããã¨ãããã£ãããã¨ãã°ãä¸ã®å³ã¯éè²ãclbããã®ã¡ã«ã±ãã¹ãã©ã ç³»åãç·è²ãsltããã®ã¡ã«ã±ãã¹ãã©ã ç³»åã表ãã¦ããã赤ã®ç¢å°ã®å ´æã§å½¢ç¶ãä¼¼ã¦ãããä½ç½®ãããã¦ãããã¨ããããã ãã®ããã¯ã¡ã«ã±ãã¹ãã©ã éã®å¤æã¢ãã«ãå¦ç¿ããã¨ãã«åé¡ã«ãªãããæéåæãåãããã®æéåæãåã£ããã¼ã¿ããã©ã¬ã«ãã¼ã¿ã¨å¼ã¶ã DTW (Dynamic Time Warping: åçæé伸縮æ³ï¼ ãã®äºã¤ã®æç³»åãã¼ã¿ã®æéåæãåãã¢ã«ã´ãªãºã ã«DTWã¨ããã®ãããã®ã§ä½¿ã£ã¦ã¿ããDTWã¯ãäºã¤ã®æç³»åãã¼ã¿ããªãã¹ãéãªãããããã«ä¼¸ã°ãããã
çµ±è¨ç声質å¤æ (2) ãã¤ã¹ãã§ã³ã¸ã£ã¼ãä½ããï¼2015/2/25ï¼ã®ç¶ãã ååã¯é³å£°ããæ½åºããã¡ã«ã±ãã¹ãã©ã ããããã®ãã©ã¡ã¼ã¿ãç´æ¥ããããã¨ã§ç°¡åãªãã¤ã¹ãã§ã³ã¸ã£ã¼ãä½ã£ããä»åããAããã®é³å£°ãBããã®é³å£°ã«å¤æããè¦åãæ©æ¢°å¦ç¿ã®ææ³ã§å¦ç¿ãã声質å¤æã®å®é¨ããã¦ãããããä»åã¯å¦ç¿ãã¼ã¿ã¨ãªãé³å£°ãã¼ã¿ããã¦ã³ãã¼ãããã¨ããã¨ã¡ã«ã±ãã¹ãã©ã ãæ½åºããã¨ããã¾ã§ã é³å£°ãã¼ã¿ã®ãã¦ã³ãã¼ã ãã¥ã¼ããªã¢ã«ã¨åãããã«CMU ARCTIC Databasesã¨ããå ¬éãã¼ã¿ã使ãããã®ãã¼ã¿ã¯7åã®è±èªè©±è ãåãæç« ãæèªããé³å£°ãã¤ãã¦ãããä»åä½ãçµ±è¨ç声質å¤æã®ææ³ã¯ãå¤æå ã®Aããã¨å¤æå ã®Bãããåãæç« ãèªãã é³å£°ãå¿ è¦ã«ãªãã®ã§ã¡ããã©ãããã¼ã¿ãè±èªãªã®ãæ®å¿µãæ¥æ¬èªã®é³å£°ã¯æ¢ãããã©ãªãã£ãããã¨ã§èªåã®å£°ã§ãã£ã¦ã¿ããã ä¸æ¬ãã¦ã³ãã¼ãããã¹ã¯
çµ±è¨ç声質å¤æ (1) ãã¼ããããï¼2015/2/11ï¼ã®ç¶ãã çµ±è¨ç声質å¤æã®ç¬¬äºåã¨ãããã¨ã§ã¾ãã¯çµ±è¨çãããªã声質å¤æã®æ çµã¿ã§ç°¡åãªãã¤ã¹ãã§ã³ã¸ã£ã¼ãä½ã£ã¦ã¿ããããããªãæ¬é¡ã¨ããã¦ããããã©ããããã£ããç解ã§ãã¦ããªãã¨çµ±è¨çãªæ¹ã¯ã¾ã£ããæ¯ãç«ããªãããæ´çãã¦ããããã ã½ã¼ã¹ã»ãã£ã«ã¿ã¢ã㫠人éã®é³å£°ã¯ãããã®å£°å¸¯ãæ¯åããããã¶ã¼é³ã声éãå£ãåãééãããã¨ã§åºã¦ããä»çµã¿ã«ãªã£ã¦ããããããæ°å¦çã«ã¢ãã«åããã®ãã½ã¼ã¹ã»ãã£ã«ã¿ã¢ãã«ã http://www.kumikomi.net/archives/2010/08/ep30gose.php ããå¼ç¨ ãã®ã¢ãã«ã§ã¯ãé³æºã«ããããã¶ã¼é³ãä½ãåºãããã¶ã¼é³ããã£ã¸ã¿ã«ãã£ã«ã¿ã«éããã¨ã§é³å£°ãä½ããé³æºã®ãã©ã¡ã¼ã¿ã¨ãã¦å£°ã®é«ãã表ããããã声éã®ãã©ã¡ã¼ã¿ã¨ãã¦ã¡ã«ã±ãã¹ãã©ã ã¨ããã®ããã使ãã
ã©ã³ãã³ã°
ã©ã³ãã³ã°
ã©ã³ãã³ã°
ãªãªã¼ã¹ãé害æ å ±ãªã©ã®ãµã¼ãã¹ã®ãç¥ãã
ææ°ã®äººæ°ã¨ã³ããªã¼ã®é ä¿¡
å¦çãå®è¡ä¸ã§ã
j次ã®ããã¯ãã¼ã¯
kåã®ããã¯ãã¼ã¯
lãã¨ã§èªã
eã³ã¡ã³ãä¸è¦§ãéã
oãã¼ã¸ãéã
{{#tags}}- {{label}}
{{/tags}}