The Contribution of Nasal Murmur to the Perception of Nasal Consonant

ABSTRACT

Identification of perceptual cues can be very helpful in almost all areas of speech signal processing. Recently, a new methodology called the 3-Dimensional-Deep Search and a visualized intelligible time-frequency computer-based model AI-gram have been introduced for research on the perceptual cues. Based on the technique, the acoustic cues for stop consonants [1], fricative consonants [2] and nasal consonants [3] are successfully found. However, these have limitations for studying the contribution of nasal murmur to the recognition of nasal consonants due to the following reasons: Firstly, they only allow the investigation of individual recognition effects along the time, frequency and amplitude axes. The effects of frequency and amplitude in a combinatorial way cannot be studied. Secondly, the initial value for the highpass filter in the filter experiment HL07 [4] is set to 697 Hz, but the nasal murmur region lies around 250 Hz. The perceptual contribution of nasal murmur to the nasal consonants cannot be assessed. To solve these problems, a new experiment has been designed by analyzing the experiment data and comparing them with the stimuli under different SNRs via AIgram. It is revealed that when the primary cue of nasal consonant is clear, which is usually under high SNRs, filtering out nasal murmur does not affect its correct perception. However, when the primary cue is weak usually under low SNRs, nasal murmur has strong complementary effects on the primary cue, and can greatly suppress confusions. This conclusion can be used for noise-robust speech recognition.

Keywords: