Perception of speech in cartilage conduction

Abstract Objective By attaching a transducer to the aural cartilage a relatively loud sound is audible even with a negligibly small fixation pressure applied to the transducer. This form of conduction is referred to as cartilage conduction (CC). Utilizing CC, novel audio devices can be developed, and one possible application is a CC hearing aid. However, there are no studies on speech perception in CC. In this study, CC speech recognition performance was measured and compared with that for air and bone conduction (AC and BC, respectively). Methods Nine volunteers with normal hearing participated in the study. The performance-intensity functions were measured for AC, BC and CC. These measurements were performed in the conditions with and without an earplug. Results Without the earplug, no differences in speech recognition scores were observed among AC, BC, and CC. With the earplug, the level at which the maximum speech recognition score was obtained did not increase in CC, which agreed with the result of BC but not AC. The maximum speech recognition CC score decreased with the earplug. The performance-intensity functions for AC and BC shifted in parallel with the earplug. These shifts approximated the average threshold shifts. In contrast, for CC, the performance-intensity function did not shift in parallel with the earplug. As for the CC threshold shifts with the earplug, although the threshold at 500 Hz decreased by 15.4 dB, those at 2000 and 4000 Hz increased by 13.8 and 31.1 dB, respectively. Compared with AC and BC, CC excessively emphasized low over high frequency sounds when the earplug was inserted. Confusion matrices analysis demonstrated that 4%, 22%, and 74% of the errors occurred at low, intermediate, and high frequency speech sounds, respectively. Thus, this excessive low frequency sound emphasis probably prevented the recognition of high frequency speech sounds. Conclusion The decrease in the maximum speech recognition score for CC with the earplug was derived from the biased frequency composition. It can be improved by frequency composition adjustment.