Zerocrossing-based fine structure representation to convey Mandarin tonal information: a study on the noise effect

Since the fine structure cue of the speech signal has been believed to be important for the pitch perception of cochlear implant (CI) users, studies are actively ongoing with attempts to propose novel CI speech processing strategies incorporating the fine structure cue. A speech synthesis model with the zerocrossing-based fine structure representation has been recently developed. This paper further examines its ability to convey Mandarin tonal information in the noisy conditions. Acoustic simulation experiment of Mandarin tone identification was conducted. The experimental results supported that the zerocrossing-based speech processing strategy could more efficiently convey Mandarin tonal information than the traditional continuous-interleaved-sampling (CIS) method, even when the Mandarin voices were contaminated by the speech-spectrum shaped noise at low signal-to-noise-ratio levels. It is believed that the zerocrossing technique would facilitate the development of novel CI speech processor to enhance the pitch perception of cochlear implantees speaking tonal language, such as Mandarin.