The Effectiveness of Low-Pass Filters in English Pronunciation Training

抄録

The purpose of this study is to investigate the effectiveness of applying low-pass filters to computer-assisted pronunciation training of English. Electric low-pass digital filtering of speech has been used in a considerable number of experiments to highlight the prosodic features of speech. Some attempts to apply low-pass filters to language learning have been made but the effectiveness has not yet been empirically proved. If the application is found to be effective to acquire the prosody of the target language, we might develop the training software using low-pass filtered sound. The experiment, using a pretest-posttest design, provided 13 native Japanese-speaking learners of English with 10 time training sessions focused on prosody using a real-time computerized pitch display. Multiple exemplars produced by native speakers of English provided training feedback. A group of seven learners were trained with low-pass filtered models while another six students were given non-filtered examples. Learners' recorded pre- and posttest productions were analyzed by computer software. Acoustic analyses were used to determine how accurately the suprasegmentals were produced. In comparison of the low-pass group with the non-filtered group, a Regression ANOVA was used and it revealed a significant difference between the two. Questionnaire responses indicated a greater awareness of the various aspects of speech and increased confidence in producing the target language.