Perception of noise-vocoded Mandarin Chinese lexical tones by native and L2 listeners

Abstract

The production, acoustics, and perception of Mandarin lexical tones have been studied extensively. Mandarin Chinese listeners rely heavily on the perception of lexical tones to distinguish word meanings and decode linguistic information. Mandarin Chinese has four lexical tones: high-level tone 1, high-rising tone 2, low-fall-rise tone 3, and high-falling tone 4. Fundamental frequency (F0) is the acoustic correlate of pitch in tones, but other acoustic cues can also contribute to tonal information. Noise-vocoded speech is thought to simulate the speech perceived by hearing-impaired listeners through a cochlea implant. When the speech passed through a noise vocoder with a specified channel number, the spectral information was replaced by random noise in the amplitude envelope across several frequency bands. F0 contour is absent in such degraded speech signals of noise-vocoded speech. The study aims to investigate normal-hearing listeners' ability to identify noise-vocoded tones with a limited number of acoustic cues. The results show that native listeners are able to distinguish the four tones, but the level tone is the most difficult to identify. L2 listeners show a different picture in tone identification. An effect was observed in the tone identification across vocoded bands and tone categories. Besides, there is a certain order of perceptual difficulty and a pattern of mislabeled tendency.