An Evaluation of Bone-conducted Ultrasonic Hearing-aid regarding Transmission of Japanese Prosodic Phonemes

Human listeners can perceive speech signals from a voice-modulated ultrasonic carrier which is presented through a bone-conduction stimulator (bone-conducted ultrasound, BCU), even if they are sensorineural hearing loss patients. As an application of this phenomenon, we have been developing a bone-conducted ultrasonic hearing aid (BCUHA). The performance of BCUHA has been evaluated considering syllable articulation and word intelligibility. These studies showed that the syllable articulation scores when using BCU were over than 60%, and the word intelligibility scores for words with high familiarity were over than 85%. The patterns of confusion in speech perception in the case of BCU have many points of similarities with those for air-conduction (AC). Although performance of BCUHA regarding perception of segmental units were evaluated, performance of perception of suprasegmental or prosodic units have not been evaluated. Japanese is a pitch accent language. Many words are contrasted by its tonal features, for example, “a’ka” (red) vs. “aka” (dirt). Also Japanese is a mora-timed language. All vowels and some consonants are contrasted by its lengths, like “obasan” (aunt) vs. “oba:san” (grandma), “kita” (came) vs. “kit:a” (cut). Since prosodic units of Japanese function as phonemes as described above, thus the perception of prosodic elements plays important role for Japanese. The purpose of this study is to evaluate the performance of perception of prosodic phoneme through BCUHA. A series of experiments which consist of minimal pairs judgement task was conducted. The minimal pairs were differentiated by prosodic elements. A pair of logatomes “etete” and “ete:te” was selected for long/short vowel discrimination task, “etete” and “etette” were for single/geminate consonant, and a pair of real words “a’ka” and “aka” was for pitch accent. Stimuli were speech sounds continua which manipulated its prosodic element. Experiments were designed as singlestimulus two-alternative forced-choice identification tasks. Ten native Japanese listeners participated in the experiments. To examine whether difference between normal air-conducted hearing (AC) and BCU hearing regarding perception pattern of prosodic elements was observed, the same tasks were conducted in AC and BCU condition. From results of the experiments, the position and the sharpness of categorical boundaries were computed by using logistic regression analyses, and then examined whether significant differences were observed by using t-tests. From analysis of t-tests, significant difference between AC and BCU conditions was not observed in all tasks. This result indicates that the BCUHA can effectively transmit prosodic phoneme as well as segmental element. INTRODUCTION For patients with acute sensorineural hearing loss who are not able to hear using a normal hearing aid, we have been developing a bone-conducted ultrasonic hearing aid (BCUHA) (Nakagawa et al. 2006). Ultrasound is defined as sound waves which travel at such a high frequency that they cannot be heard by humans. However, if the ultrasound is presented through a bone-conducted stimulator (bone-conducted ultrasound, BCU) the ultrasound is perceived by human listeners (Lenhardt et al. 1991). In addition, if BCU signals are amplitude-modulated by speech signals, listeners can perceive the original speech signals Lenhardt et al. (1991). These voice-modulated BCU signals enable patients with acute sensorineural hearing loss perceive speech signals Nakagawa et al. (2006). The BCUHA being developed is based on these observations. Performance of the BCUHA have been evaluated by using syllable articulation and word intelligibility. Syllable articulation scores were over 60% (Okamoto et al. 2005) and word intelligibility scores for words with high familiarity were over 85% (Nakagawa et al. 2006). The patterns of confusion in speech perception in the case of BCU have many points of similarities with those for air conduction (AC) (Okamoto et al. 2005). Although the usability of the BCUHA has been evaluated as mentioned above, the evaluations have been restricted to the transmission of segmental information, or textural messages. In other words, little attention has been paid to the transmission of prosody. For Japanese speaker, perception of suprasegmental, especially lengths of segments and pitch patterns, are very important because those features can be distinctive for Japanese language. For example, the lengths of vowels /a/ contrast in the words /obasan/ (“aunt”) and /oba:san/ (“grandmother”), the lengths of consonants /t/ contrast in /kita/ (“come”) and /kit:a/ (“cut”). Likewise, /ka’ki/ (“oyster”) and /ka‘ki’/ (“persimmon”) are discriminated by their pitch patterns. The purpose of this study is to evaluate the performance of perceptions of those suprasegmental phoneme via BCUHA. It