You are viewing the SNL 2018 Archive Website. For the latest information, see the Current Website.

Poster Slam Session D
Friday, August 17, 4:30 – 4:45 pm, Room 2000C, Chair: Angela Grant

Sensorimotor adaptation in speech is sensitive to vowel targets of altered feedback

Hardik Kothare1,2, Inez Raharjo1,2, Kamalini Ranasinghe2, Vikram Ramanarayanan2,3, Benjamin Parrell4, John Houde2, Srikantan Nagarajan1,2;1UC Berkeley-UCSF Graduate Program in Bioengineering, 2University of California, San Francisco, 3Educational Testing Service R&D, 4University of Wisconsin- Madison

Auditory feedback plays an important role in speech production. Any alteration in auditory feedback usually engenders a change in speech output. The speech motor control system also learns to anticipate and compensate for consistent feedback alterations. This counteractive response or sensorimotor adaptation persists temporarily even after feedback returns to normal. Sensorimotor adaptation in speech is a form of learning that helps the motor system maintain desired speech output in spite of changes in the properties of the vocal tract. An important type of sensorimotor adaptation in speech is that seen in response to formant-shifted auditory feedback. From one study, we know that the extent of such formant adaptation is heterogeneous across different target vowel productions (Mitsuya et al, 2015) but the formant shifts investigated were only upshifts or downshifts in the single formant F1. What if we shift multiple formants (F1 and F2) at once, shifting feedback in different directions in F1-F2 space? Does the accuracy and consistency of sensorimotor adaptation depend on the size and direction of the feedback alteration in F1-F2 space? To investigate these questions, we employed real-time auditory feedback alteration to shift the frequency values of the first and second formants (F1 and F2) of participants’ speech. The experiment comprised six cases; the shift was different in each case (from /ɛ/ to /I/, /i/, /e/, /æ/, /ɑ/ and /u/). In each case, participants produced 90 repetitions of the nonsense word ‘bep’ (vowel /ɛ/). A case started with a non-altered block of 10 trials, followed by a block of 50 trials with a constant alteration and then by a non-altered washout block of 30 trials. Shifts were designed on a subject-by-subject basis using pre-collected baseline formant frequencies of vowels. We find that adaptive control of vowel formant frequency depends on the magnitude and direction of the applied shift in the two-dimensional F1-F2 vowel space. We also observe that all shifts, except the one from /ɛ/ to /u/, elicit a response of a compensatory nature. A two-dimensional vector resolution analysis of the response vectors reveals that they have a component orthogonal to the axis of the shift and a component parallel to the shift axis. Statistical analyses reveal that the orthogonal component appears to be more resistant than the parallel component to changes in shift magnitude, suggesting that the orthogonal response may be more related to the detection of errors rather than nature of the errors. Additionally, the response seems to depend on whether the applied shift is upwards or downwards along the F2 axis in the F1-F2 plane. This suggests that the altered feedback’s degree of vowel backness may play a role in determining the extent of formant adaptation. These results suggest that sensorimotor adaptation in speech is sensitive not only to the vowel sound to which altered feedback is applied but also on the spatial directionality and magnitude of applied shift in formant space.

Topic Area: Speech Motor Control and Sensorimotor Integration

Poster D7

Back to Poster Slam Sessions