Interaction between speech coding and semantic predictability in speech perception.
Nirmal Kumar Srinivasan, Thomas D. Carrell · The Journal of the Acoustical Society of America · 2008
The present experiments investigated how humans process speech presented in the form low-passed natural sentences (5000 Hz), cell phone coded sentences, and computer synthesized sentences. Spin sentences [Kalikow et al., (1977) and Bilger et al., (1984)] were presented in a background of multispeaker babble using these three coding representations. Word accuracy and visual-motor task performance were measured. Intelligibility was based on the participants' accuracy in repeating the final word of the sentence. Attention was based on the participants' performance on the simultaneous pursuit rotor task [Srinivasan and Carrell (2007)]. In the pursuit rotor task, a stylus was used to maneuver a cursor on a computer display with the goal of keeping the cursor aligned as closely as possible to a moving target. Word accuracy was higher for predictable sentences than for unpredictable sentences for all three encodings. However, pursuit rotor performance showed a complex set of interactions. For example, participants performed better on the pursuit rotor task for unpredictable sentences than predictable sentences, but only for the cell phone speech not for natural or synthetic speech. The findings suggest that the perception of cell phone speech does not degrade the same way that natural or synthetic speech does.