Development of the Mandarin speech banana

Abstract

Purpose: For Indo-European languages, “speech banana” is widely used to verify the benefits of hearing aids and cochlear implants. As a standardised “Mandarin speech banana” is not available, clinicians in China typically use a non-Mandarin speech banana. However, as Chinese is logographic and tonal, using a non-Mandarin speech banana is inappropriate. This paper was designed to develop the Mandarin speech banana according to the Mandarin phonetic properties.

Method: In the first experiment, 14 participants read aloud the standard Mandarin initials and finals. For each pronounced sound, its formants were measured. The boundary of all formants formed the formant graph (intensity versus frequency). In the second experiment, 20 participants listened to a list of pre-recorded initials and finals that had been filtered with different bandwidths. The minimum bandwidth to recognise a target sound defined its location on the formant graph.

Result: The Mandarin speech banana was generated with recognisable initials and finals on the formant graph. Tone affected the shape of the formant graph, especially at low frequencies.

Conclusion: Clinicians can use the new M andarin speech banana to counsel patients about what sounds are inaudible to them. Speech training can be implemented based on the unheard sounds in the speech banana.