A new frequency lowering technique for Mandarin-speaking hearing aid users

Frequency lowering technologies have been developed to improve the hearing experience of people with high-frequency hearing loss by preserving high-frequency information of the speech in a lower-frequency band. The technique was shown effective on the English speech for English-speaking people with severe to profound high-frequency hearing loss. The unique phonology of the Mandarin makes it worthwhile to examine the effect of this technique on the Mandarin speech, which has not been well investigated. In this paper, we propose a new frequency transposition framework that incorporates the language-specific characteristics of the Mandarin for efficient signal processing, a deep neural network (DNN) technique for Mandarin consonant/vowel classification, and the equal loudness model in psychoacoustics for time-domain compensation. Our results show that the proposed algorithm achieves statistically significant improvements in Mandarin speech recognition performance in a simulated severe high-frequency hearing loss condition.