Presentation is loading. Please wait.

Presentation is loading. Please wait.

Janine Wotton, Kristin Welsh, Crystal Smith, Rachel Elvebak, Samantha Haseltine (Gustavus Adolphus College) and Barbara Shinn-Cunningham (Boston University).

Similar presentations


Presentation on theme: "Janine Wotton, Kristin Welsh, Crystal Smith, Rachel Elvebak, Samantha Haseltine (Gustavus Adolphus College) and Barbara Shinn-Cunningham (Boston University)."— Presentation transcript:

1 Janine Wotton, Kristin Welsh, Crystal Smith, Rachel Elvebak, Samantha Haseltine (Gustavus Adolphus College) and Barbara Shinn-Cunningham (Boston University). Sentence context influences vowel perception in reverberant conditions

2 Pseudo-anechoic Corner 1m Frequency (Hz) dB Time (ms) 0° H 0° V

3 gn a tn e t

4 Hillenbrand, Getty, Clark and Wheeler JASA 1995 97(5) 3099-3111. æε

5 F1 Hz F2 Hz gn a t n e t Sampled every 6 milliseconds 20 samples 24 samples æ ε

6 F2 Hz F1 Hz gn a t n e t + noise + noise + reverb Mean and standard deviation of formant values for single vowel utterance

7 Sentences convolved with impulse responses in either pseudo-anechoic or reverberant conditions Sentence conditions Target word consistent with context e.g. The cattle grazed in the meadow. Target word inconsistent with context e.g. The cattle boiled the water. Target word neutral with context e.g. The cattle went shopping. Stimuli presented dichotically (TDT). Response - subjects type what they hear.

8 No added noise open ended response N = 23 Significant comparisons 1) Reverberant Vs. pseudo- anechoic 2) Inconsistent condition Vs. others ** 3) Reverberant condition in inconsistent and neutral conditions * * * ** Substitution errors

9 Sum of errors Reverberant Pseudo-anechoic Number of trials each condition = 23*8 = 184 Number of trials each condition = 23*3*8 = 552

10 Added noise (S/N -8dB) open ended response N = 6 Significant comparisons 1) More errors in noise than no noise in all conditions 2) More errors in inconsistent than others. ** 3) More errors in inconsistent in reverberation than pseudo-anechoic * * ** Substitution errors

11 Conclusions 1)Increase the ambiguity of the vowel by adding noise or reverberation and the errors in the inconsistent condition increase. 2) In the consistent condition, noise increases the errors but the addition of reverberation does not. 3)The sentence context becomes more influential if the uncertainty of the vowel choice is increased. 4) The error pattern reveals a shift in bias rather than sensitivity for correctly responding with a consistent sentence.

12 Added noise 2AFC N = 6


Download ppt "Janine Wotton, Kristin Welsh, Crystal Smith, Rachel Elvebak, Samantha Haseltine (Gustavus Adolphus College) and Barbara Shinn-Cunningham (Boston University)."

Similar presentations


Ads by Google