Chapter 10: Exercises

  1. (**) ­pºâ¤¤¤å¥À­µÀWÃФ§¤@: ½Ð¼g¤@­Ó MATLAB µ{¦¡ vowelSpectrumPlot01.m¡A§¹¦¨¤U¦C¥\¯à¡G
    1. ¶i¦æ¤T¬í¿ý­µ¡A¿ý­µ³W®æ¬O 16KHz/16Bits/Mono¡A¿ý­µ¤º®e¬O¤¤¤å¥À­µ¡u£¸¡v¡C¡]½ÐºÉ¶qºû«ùí©wªº­µ°ª¤Î­µ¶q¡A¥i¥H¸ÕÅ¥¦¹½d¨ÒÀɮסC¡^
    2. ¨Ï¥Î buffer2.m ¤Î frame2volume.m ¨Ó¤Á¥X­µ®Ø¨Ã­pºâ­µ¶q¡A¬ÛÃö³W®æ¬O frameSize=32ms, overlap=0ms¡C¡]´£¥Ü¡G§A¥²¶·¥ý±N frameSize ¤Î overlap ÂনÂI¼Æ¡C¡^½Ð§ì¥X­µ¶q³Ì¤jªº­µ®Ø¡A¦b¨ä«e«á¦U§ì¥X 2 ­Ó­µ®Ø¡AÁ`¦@¦³ 5 ­Ó­µ®Ø¡C
    3. ¨Ï¥Î fft ¨Ó­pºâ³o¤­­Ó­µ®ØªºÀWÃСA§t¤j¤p©M¬Û¦ì¡A¨Ãµe¥X©Ò¦³¬ÛÃö¹Ï§Î¡G
      1. subplot(3,1,1)¡G¤­­Ó­µ®Øªº°T¸¹¹Ï¡C
      2. subplot(3,1,2)¡G¤­­Ó­µ®Øªº³æÃäÀWÃбj«×¹Ï¡C
      3. subplot(3,1,3)¡G¤­­Ó­µ®Øªº³æÃäÀWÃЬۦì¹Ï¡C
      µe¥Xªº¹Ï§ÎÀ³¸ÓÃþ¦ü¤U¹Ï¡G
      ¡]´£¥Ü¡G(a)¹ï­µ®Ø¶i¦æ fft ®É¡A¥i¥H¥ý­¼¤W Hamming window¡A¥i¥HÅýºâ¥X¨ÓªºÀWÃФñ¸û°®²b¡Cµ{¦¡½X¬O¡Gframe=frame.*hamming(length(frame))¡C(b)¦bµe¥X³æÃäÀWÃбj«×¹Ï¡A¥i¥H¨Ï¥Î log ¨è«×¡A¤ñ¸û®e©ö¬Ý¥X¿Óªiµ²ºc¡A¤èªk¬O¦bµe²Ä¤G­Ó¹Ï«á¡A¤U¹F¤U¦C©R¥O¡Gset(gca, 'yscale', 'log')¡C(c) ¬°¨Ï¬Û¦ì²£¥Í³sÄòªº¦±½u¡A¥i¥H¨Ï¥Î unwarp «ü¥O¡C¡^
    4. ½ÐÆ[¹î¬Ý¬Ý©Òµe¥Xªº¤T­Ó¹Ï¡A¨º¤@­Ó¹Ïªº¤­±ø¦±½uªº¬Û¦ü«×³Ì°ª¡H¡]´£¥Ü¡G¥Ñ©ó§Ú­Ìªºµo­µ³£¬O¡u£¸¡v¡A¦]¦¹¬Û¦ü«×³Ì°ªªº¹Ï©Ò¥Îªº¯S¼x¡A´N¬Oí©w¤£Åܪº¯S¼x¡A¥i¥Î©ó»y­µ¿ëÃÑ¡C¡^
  2. (**) ­pºâ¤¤¤å¥À­µÀWÃФ§¤G: ½Ð¼g¤@­Ó MATLAB µ{¦¡ vowelSpectrumPlot02.m¡A¥\¯à¦p¦P¤W¤@ÃD¡A¦ý±N¿ý­µ¤º®e§ï¦¨¤¤¤å¥À­µ¡u£¸£½¡v¡C¦¹ÃD¥i¥Î¨Ó±´°Q¦b­µ°ªÅܤƪº±¡ªp¤U¡AÀWÃЪºÅܤƱ¡ªp¡C
  3. (***) ­pºâ¤¤¤å¥À­µÀWÃШöi¦æ¤ÀÃþ: ½Ð¼g¤@­Ó MATLAB µ{¦¡ vowelSpectrumRecogChinese01.m¡A§¹¦¨¤U¦C¥\¯à¡G
    1. ¶i¦æ¤­¬í¿ý­µ¡A¿ý­µ³W®æ¬O 16KHz/16Bits/Mono¡A¿ý­µ¤º®e¬O¤¤¤å¥À­µ¡u£«¡B£¸¡B£¹¡B£®¡B£¬¡v¡C¡]µo­µ½ÐºÉ¶q¥­Ã­¡A¨Ã¦b¥À­µ¤§¶¡µy³\°±¹y¡A¥H«K«áÄò¶i¦æ¦Û°Ê¤Á­µ¡A¥i¥H¸ÕÅ¥¦¹½d¨ÒÀɮסC¡^
    2. ¨Ï¥Î endPointDetect.m (in SAP Toolbox) ¨Ó§ä¥X³o¤­­Ó­µªº¶}©l©Mµ²§ô¦ì¸m¡C½Ð½Õ¾ã¬ÛÃöºÝÂI°»´ú°Ñ¼Æ¡A¨Ï±o§Aªºµ{¦¡¯à°÷¦Û°Ê¦a¥¿½T¤Á¥X³o¤­­Ó­µ¡C§A¥i¥H³]©w plotOpt=1¡A¥H«K¨Ï¥Î endPointDetect.m ¨Óµe¥XºÝÂI°»´úµ²ªG¡A¥¿½T¹Ï§ÎÃþ¦ü¤U¹Ï¡G
    3. ¨Ï¥Î buffer2.m ¨Ó¹ï³o¤­­Ó­µ¤Á¥X­µ®Ø¡]frameSize=32ms, overlap=0ms¡A¦ý§A¥²¶·¥ý±N frameSize ¤Î overlap ÂনÂI¼Æ¡^¨Ã­pºâÀWÃбj«×¡A½Ð±N³o¤­­Ó­µªºÀWÃбj«×µe¥X¨Ó¡A¨C¤@­Ó¥À­µ¹ïÀ³¨ì m ±ø¦±½u¡Am ¬O¦¹¥À­µªº­µ®Ø­Ó¼Æ¡A½Ð¤ñ¸û¦¹ m ±ø¦±½uªº¬Û¦ü«×¡Cµe¥X¹Ï§ÎÀ³¸ÓÃþ¦ü¤U¹Ï¡G
    4. ½Ð¥Î knncLoo.m (in Machine Learning Toolbox) ¨Óºâ¥X¨Ï¥Î KNNC ±N¸ê®Æ¤À¦¨¤­Ãþ¥À­µªº leave-one-out ¿ëÃѲv¡C¡]¥i¥H³]©w k = 1¡C¡^
    5. ½Ð±N¯S¼x¼Æ¥Ø¥Ñ 1 ¼W¥[¨ì 128¡Aµe¥X KNNC ªº leave-one-out ¿ëÃѲv¹ï¯S¼x¼Æ¥Øªº¦±½u¡Cµe¥X¹Ï§ÎÀ³¸ÓÃþ¦ü¤U¹Ï¡G
    6. ½Ð¥Î PCA ©Î LDA ±N³o¨ÇÀWÃбj«×¸ê®Æ§ë¼v¨ì¤G«×ªÅ¶¡¡A¨Ã±N¹Ï§Îµe¥X¨Ó¡AÆ[¹î¬Ý¬Ý¬O§_¦P¤@Ãþ¥À­µ¡A¨ä¸ê®ÆÂI¦³»E¶°¦b¤@°_ªº¶É¦V¡C
  4. (**) ­pºâ­^¤å¥À­µÀWÃШöi¦æ¤ÀÃþ: ½Ð¼g¤@­Ó MATLAB µ{¦¡ vowelSpectrumRecogEnglish01.m.m¡A¥\¯à¦p¦P¤W¤@ÃD¡A¦ý±N¿ý­µ¤º®e§ï¦¨­^¤å¥À­µ¡ui¡Be¡B Ë¡Bo¡Bu...¡v¡C

Audio Signal Processing and Recognition (­µ°T³B²z»P¿ëÃÑ)