以功率頻譜分析研究發聲基頻與聽覺發聲反射的關係

Relationship between Fundamental Frequency System and Audio-Vocal Reflex: Illustration by Power Spectral Analysis of Vocal Fundamental Frequency

李國熙(Guo-She Lee) ； 蕭自佑(Tzu-Yu Hsiao) ； 郭博昭(Terry B. J. Kuo)

《台灣耳鼻喉頭頸外科雜誌》 39卷4期 (2004/10) Pp. 145-152

https://doi.org/10.6286/2004.39.4.145

摘要

The audio-vocal reflex is one of the most important examples of the close relationship between the auditory and the phonatory system. We attempted to elucidate the effects of noise masking upon the power spectral analysis of the vocal fundamental frequency (VF0) and to demonstrate the interaction between these two systems, and discussed the possible mechanisms for such interaction.
METHODS: Ten healthy subjects, (five males), who featured no history of any hearing loss, speech disorder or neurological disease, were enrolled into this study. None of these study participants had been trained as a singer. All subjects were instructed to vocalize the sustained vowel /a/ at an intensity of less than 75 dB for a period of five seconds in a sound-proofed booth. All peripheral environmental noise was controlled to the level of less than 40 dB. Voices were sampled at the rate of44 kHz and were stored electronically in digital format. Another voice recording of the same vowel was made by the same individual under the same conditions, except that 85-dB speech noise was applied to both participant ears using earphones at the time the participant was making the sound recording. Fundamental frequencies (F0) for each voice sample were derived using the algorithm for the auto-correlation function of sound-pressure signals, such frequencies being normalized to cents (1 octave=1200 cents). A vocal power spectrum for the VF0 was then derived using Fourier transformation of the normalized VF0s. The products of each frequency and its own power were summed in the frequency range of from 0.3 to 12 Hz, following which, this value was divided by the total power in order to provide a parameter to represent the mean powerfrequency (MPF) of the spectrum.
RESULTS: There were no significant changes of vocal fundamental frequencies, sound in tensities and percentage jitter found between the voices without noise masking and the voices with noise masking. However, the standard deviations of VF0 and the low frequency (0.3～3 Hz) power (LFP) revealed significant increase (p＜0.05, paired Student's t-test) with noise masking, and MPF was significantly reduced due to noise masking (p＜0.001, paired Student's t-test).
CONCLUSIONS: The increases in LFP under noise masking conditions revealed a possible negative feedback control of the human auditory system upon the phonatory system regarding the control of VF0. The auditory system helped reducing the 3-Hz oscillation of the VF0 in order to maintain a steady F0 during phonation. The interaction between these two systems was well demonstrated by the results of the power spectral analysis of VF0, which may become a promising measure for further researches and for certain clinical applications in the future.

關鍵字

聽覺發聲反射 ； 發聲基頻 ； 功率頻譜分析 ； 語音噪音

並列摘要

The audio-vocal reflex is one of the most important examples of the close relationship between the auditory and the phonatory system. We attempted to elucidate the effects of noise masking upon the power spectral analysis of the vocal fundamental frequency (VF0) and to demonstrate the interaction between these two systems, and discussed the possible mechanisms for such interaction.
METHODS: Ten healthy subjects, (five males), who featured no history of any hearing loss, speech disorder or neurological disease, were enrolled into this study. None of these study participants had been trained as a singer. All subjects were instructed to vocalize the sustained vowel /a/ at an intensity of less than 75 dB for a period of five seconds in a sound-proofed booth. All peripheral environmental noise was controlled to the level of less than 40 dB. Voices were sampled at the rate of44 kHz and were stored electronically in digital format. Another voice recording of the same vowel was made by the same individual under the same conditions, except that 85-dB speech noise was applied to both participant ears using earphones at the time the participant was making the sound recording. Fundamental frequencies (F0) for each voice sample were derived using the algorithm for the auto-correlation function of sound-pressure signals, such frequencies being normalized to cents (1 octave=1200 cents). A vocal power spectrum for the VF0 was then derived using Fourier transformation of the normalized VF0s. The products of each frequency and its own power were summed in the frequency range of from 0.3 to 12 Hz, following which, this value was divided by the total power in order to provide a parameter to represent the mean powerfrequency (MPF) of the spectrum.
RESULTS: There were no significant changes of vocal fundamental frequencies, sound in tensities and percentage jitter found between the voices without noise masking and the voices with noise masking. However, the standard deviations of VF0 and the low frequency (0.3～3 Hz) power (LFP) revealed significant increase (p＜0.05, paired Student's t-test) with noise masking, and MPF was significantly reduced due to noise masking (p＜0.001, paired Student's t-test).
CONCLUSIONS: The increases in LFP under noise masking conditions revealed a possible negative feedback control of the human auditory system upon the phonatory system regarding the control of VF0. The auditory system helped reducing the 3-Hz oscillation of the VF0 in order to maintain a steady F0 during phonation. The interaction between these two systems was well demonstrated by the results of the power spectral analysis of VF0, which may become a promising measure for further researches and for certain clinical applications in the future.

並列關鍵字

audio-vocal reflex ； vocal fundamental frequency ； power spectrum analysis ； speech noise