Last Modified : Tue, 12 May 15
Manipulation of Phonemes in Natural Speech
- ka 2 ta 2 pa m4v, wmv; Three different conditions (1-3) are played, with the AI gram showing for each. You might either hear /ka, ta, pa/ or /ka, pa, pa/. Listen for one, and then the other. Here is what we have done:
- You should always hear /ka/, as this is unmodified speech.
- You may hear /ta/ or /pa/. Here we have removed the mid-frequency /ka/ burst. See what you naturally hear first. Then try to listen for the other of /ta/, /pa/. This depends on your first language, and should only work for L1=English listeners.
- You should always hear /pa/. Here both the /ka/ and /ta/ acoustic cues have been removed.
* ka 2 ka 2 ta 2 pa wmv Δ
- ba 2 ga m4v, wmv @ 12 [dB-SNR]; Here there are two presentations:
- The natural /ba/ in 12 dB SNR of Noise. With no masking noise, it is 85% /ba/, else /va/. At 10 dB SNR it is /va/ 50% of the time. At 12 dB SNR it is frequently heard as /va/.
- The mid-frequency (i.e., /ga/) burst enhanced & the low-frequency onset removed. Many report it as /ga/.
- ga 2 da f103 wmv
- da 2 ga f103 m4v, wmv
- fa 2 ba f103 wmv
- Sa 2 cha 2 sa 2 za 2 Da f103 (m4v, wmv)
- ka 2 ta: f105 mpg Δ, wmv Δ
- ke 2 te: f105 mpg Δ, wmv Δ; m104: wmv Δ
- ka 2 ta 2 pa: m4v
- ma 2 na: f105 mpg Δ, wmv Δ
- S 2 s 2 z 2 d: m111 mpg Δ
- sa 2 za 2 Da: f105 mpg Δ, wmv Δ
- za 2 Da: f106 mpg Δ, wmv Δ
- She had your dark suit --> He has your dart shoot (m4v, wmv)