Magnitude estimation scaling of computerized (digitized) speech by young adults under different listening conditions

The purpose of this study was to evaluate the intelligibility of natural, tape recorded speech sentences and digitized speech sentences digitized at low, moderate, and high sampling rates by young adults under different listening conditions (quiet vs. noise) using magnitude estimation and transcription measures. This study also explored the relationship between the magnitude estimation responses and transcription measures.;A single group of twenty-four young adults participated as subjects in this study. Forty sentences randomly selected from "Sentences for Phonetic Inventory" (Fairbanks, 1960) with slight modification served as stimulus material for this study. CSL Model 4300B was used to record and play back the digitized versions of these sentences. These sentences were digitized using the CSL Model 4300B at low, moderate and high sampling rates of 4000 Hz, 8000 Hz, and 16000 Hz, respectively. Thus two audio CDs were produced with digitized sentences which were used to deliver this stimulus material to subjects.;Natural, tape recorded speech was used as a control for this study. The natural, tape recorded speech sentences and digitized speech sentences were presented to each subject in quiet and in the presence of background noise at a signal-to-noise ratio of 0 dB. The subjects were instructed to utilize magnitude-estimation scaling to evaluate these sentences by assigning a number that corresponded to the intelligibility of each sentence. The subjects were also instructed to immediately write down each sentence soon after they assigned a number to it.;Separate 4 x 2 repeated measures analyses of variance were performed to assess the effects of the two independent variables (mode of speech and listening conditions) on each of the dependent variables (magnitude estimation responses and total number of words correctly transcribed). Results showed that the main effects for mode of speech and listening condition were statistically significant. The interaction effects of mode of speech by listening condition were also statistically significant. Pairwise comparisons showed that the magnitude estimation responses were higher in the quiet condition than in the corresponding noise condition for each level of mode of speech. Regarding the total number of words correctly transcribed, significantly more words were transcribed in the quiet condition for the sampling rates of 4,000 Hz, 8,000 Hz, and natural, tape recorded speech than in the corresponding noise condition. The range of correlations between the magnitude estimation responses and total number of words correctly transcribed was from noninterpretable (ns) to moderate. Based on the results of the present study, several avenues for future research are suggested.