Effects of high-pass filtering on perception of dialect and talker sex
MetadataShow full item record
Publisher:The Ohio State University
Series/Report no.:The Ohio State University. Department of Speech and Hearing Science Undergraduate Research Theses; 2018
Abstract: Linguistic (message related) and indexical (related to individual talker characteristics) information are conveyed in spoken language. It has been demonstrated that listeners are sensitive to indexical cues such as regional dialect spoken in their speech community (Clopper et al., 2006; Jacewicz & Fox, 2012). However, we have relatively little information in terms of how listeners form a perceptual representation of speaker identity and how the indexical information is conveyed by the vocal source (related to voice) and filter (related to the changing shape of the vocal tract during speech production). We do not know the necessary and appropriate acoustic cues that listeners use to determine speaker identity. This project analyzes how listeners process information about talker sex (male, female) and dialect (Ohio, North Carolina) with high-pass filtered speech. Research shows that intelligibility remains relatively high even when large portions of the speech spectrum are eliminated by filtering (Stickney & Assmann, 2001), indicating that speech cues are widely distributed. This study explores the contribution of cues at different frequency ranges to the identification of speaker dialect and gender by systematically removing the segmental and semantic content from speech using high-pass filtering. Two types of tests will be given: Identification and Intelligibility. In the Identification (ID) task, listeners will be required to identify each sentence as having been produced by an Ohio speaker or a North Carolina speaker; male or female. In the Intelligibility task participants will be asked to write down (using a Matlab program) the messages (i.e., linguistic content). Intelligibility data will be analyzed using Signal Detection Theory to separate sensitivity to indexical cues from the response bias. Based on previous findings for low-pass filtered speech, I expect to find greater amount of dialect cues than gender cues in high-pass filtered speech. Intelligibility is expected to increase with each decreasing cutoff frequency of the filter.
Academic Major: Speech and Hearing Science