Adachi S, Yamada M
ATR Human Information Processing Research Laboratories, Kyoto, Japan.
J Acoust Soc Am. 1999 May;105(5):2920-32. doi: 10.1121/1.426905.
A theory that the high melody pitch of biphonic singing, Xöömij, is produced by the pipe resonance of the rear cavity in the vocal tract is proposed. The front cavity resonance is not critical to the production of the melody pitch. This theory is derived from acoustic investigations on several three-dimensional shapes of a Xöömij singer's vocal tract measured by magnetic resonance imaging. Four different shapes of the vocal tract are examined, with which the melody pitches of F6, G6, A6, and C7 are sung, along with the F3 drone of a specific pressed voice. The second formant frequency calculated from each tract shape is close to the melody pitch within an error of 36 cents. Sounds are synthesized by convolving a glottal source waveform provided by the Rosenberg model with transfer functions calculated from the vocal tract shapes. Two pitches are found to be successfully perceived when the synthesized sounds are listened to. In a frequency range below 2 kHz, their spectra have a strong resemblance to those of the sounds actually sung. The synthesized sounds, however, fail to replicate the harmonic clustering at 4-5 kHz observed in the actual sounds. This is speculated to originate from the glottal source specific to the "pressed" timbre of the drone.
提出了一种理论，即双声唱法“呼麦”的高音调是由声道后腔的管道共振产生的。前腔共振对旋律音高的产生并不关键。该理论源于对通过磁共振成像测量的呼麦演唱者声道的几种三维形状进行的声学研究。研究了四种不同形状的声道，演唱者用这些声道唱出了F6、G6、A6和C7的旋律音高，以及特定紧喉音的F3持续低音。根据每个声道形状计算出的第二共振峰频率与旋律音高接近，误差在36音分以内。通过将罗森伯格模型提供的声门源波形与根据声道形状计算出的传递函数进行卷积来合成声音。当聆听合成声音时，可以成功地感知到两个音高。在2kHz以下的频率范围内，它们的频谱与实际演唱声音的频谱非常相似。然而，合成声音未能复制在实际声音中观察到的4-5kHz的谐波聚集现象。据推测，这源于持续低音“紧喉”音色特有的声门源。
