Demonstration sounds for ICASSP-2003

• Reference:
  Nakatani, T., and Miyoshi, M, ``Blind dereverberation of single channel speech signal based on harmonic structure,'' submitted to ICASSP-2003.
• Task: dereverberation of male/female word utterances
• Training data: 5240 Japanese word utterances included in ATR database convolved with impulse responses. The reverberation times (rtime) are 0.1, 0.2, 0.5, 1.0 second, respectively.
• Presentataion document [PDF]

1. Female speech

   Rtime (second)	Original sound	Reverberant sound	Dereverberated sound
   0.1
   0.2
   0.5
   1.0

   
   
2. Male speech

   Rtime (second)	Source sound	Reverberant sound	Dereverberated sound
   0.1
   0.2
   0.5
   1.0