NTT-Tohoku University Familiarity-Controlled Word Lists 2007 (FW07)
Tadahisa KONDO and Shigeaki AMANO, NTT Communication Science Laboratories
Yoichi SUZUKI and Shuichi SAKAMOTO, Research Institute of Electrical Communication, Tohoku University
The 1600 spoken words (20 words times 20 sets times 4 ranks) reconstructed from the "NTT - Tohoku University Familiarity-controlled Word Lists (FW03)" and edited under the following 5 conditions:
- Pseudo speech noise is superposed with four conditions (S/N ratio: 3, 0, -3, -6 dB).
- No noise superposition.
A part of this data set is contained in the 5 audio CD’s for the convenience of the examination, etc.*1
4 trained Japanese speakers including 2 males and 2 females; each word is spoken once.
Speech file format
WAV format (48 kHz, 16 bit, Mono)
1 DVD and 5 audio CDs
For research and development purposes only
11 429 yen (plus consumption tax for a domestic order)
The programs for sound pressure level transformation and noise superposition are included in the DVD.
*1 Noise data is superposed on the 800 words (20 words times 10 sets times 4 ranks) spoken by a female speaker under 5 conditions (no noise, +3 dB, 0 dB, -3 dB, -6 dB).
*2 The sound pressure level criteria are different for this FW07 and the formerly distributed FW03. FW07 is about 6 dB lower then FW03.