Effects of training format on perceptual learning of spectrally degraded voices.
Sangsook Choi, Karen Iler Kirk, Tom M Talavage, Vidya Krull, Chris Smalt, Steve Baker · The Journal of the Acoustical Society of America · 2009
There is evidence that familiarity with the talker facilitates speech recognition (see Nygaard, 2005 for reviews). Identifying talker voices from an auditory speech signal is difficult for cochlear implant users due to the lack of spectral detail conveyed by their processors. Recently, Sheffert and Olson (2004) demonstrated facilitation in voice learning when listeners with normal hearing were trained using audiovisual speech. In the present study, normal hearing listeners were presented with speech processed through a 12-channel noise band cochlear implant simulator and trained to recognize voices in an auditory-only or audiovisual format. Listeners in the audiovisual format required fewer sessions to reach criterion (80% accuracy) than those in the auditory-only format. Voice learning varied as a function of the talkers included in the training set. For the talkers that were more difficult to identify, audiovisual training yielded greater improvements in voice identification than did auditory-only training. After training, participants completed a word recognition task using processed sentences produced by training talkers and novel talkers. Talker familiarity did not influence word recognition performance for either training group. This may be due to ceiling effects with the 12-channel noise band processor. Additional testing in noise is underway.