The interaction of acoustic and linguistic aids to sentence intelligibility.

摘要

One of the earliest explanations for good speech intelligibility in poor listening situations was context (Miller, Heise & Lichten, 1951). Context was thought to allow listeners to group and predict speech appropriately and is known as a “top-down” listening process. Amplitude comodulation is another mechanism that has been shown to improve sentence intelligibility. More recently, amplitude comodulation was shown to provide acoustic grouping information without changing the linguistic content of the desired signal (Carrell & Opie, 1992). The use of amplitude comodulation in this way is considered a bottom-up process. The present experiment investigated how amplitude comodulation and semantic information combined to improve speech intelligibility. The sentence material used in these experiments accounted for any a priori probability of correct recognition (Boothroyd & Nittrouer, 1988). Experiment I examined the influence of semantic information and amplitude comodulation on amplitude degraded speech, and frequency degraded speech. Semantic information did not benefit listeners for either type of sentence. This result contrasted with results of an examination of semantic information on sentence presented in noise. Amplitude comodulation was found to aid listeners' ability to extract speech from amplitude degraded sentences but hindered their ability with frequency degraded sentences. The results obtained in Experiment I led to the questions: Were listeners not able to use semantic information due to the extremely low intelligibility of signals in Experiment I? Did the type of acoustic degrading processes used interfere with listeners' abilities to access semantic information? Experiment II examined these questions and revealed that semantic information aided listeners when speech was degraded with noise and intelligibility decreased as signal-to-noise ratios (SNR) decreased. Furthermore, −9 dB SNR produced comparable intelligibility levels for the low-predictability sentences as were found in low-predictability degraded amplitude and degraded frequency sentences; however, unlike the results from the degraded sentences, semantic information was found to improve listener intelligibility for the −9 dB SNR sentences. Although further research must be conducted to support this claim, the results of the present work support the notion that auditory object formation is likely to be a process that is separate from and occurs prior to speech perception.