5D Library › 歌手共振峰簇、声门闭合商与声学权威度心理物理学 · English
第六章 半阻塞声道练习:吸管物理与一条十四天的闭合商曲线
一根吸管产生多少背压
半阻塞声道练习(对着吸管发声、抿唇发声或把管子插入水中发声)会提高声门上方的压力。Titze 的解释是,增加的惯性与阻性负载有助于声带以较小的碰撞力高效振动 (Titze 2006)。p11 计算不同直径的吸管在中等气流下的背压。
推导
# Chapter 6. p11: Back pressure of a straw in semi-occluded phonation.
# Air flow 0.15 L/s through a straw 0.15 m long. Laminar (Poiseuille)
# loss 8 mu L U / (pi r^4) plus the exit loss of the jet, rho v^2 / 2;
# air viscosity mu = 1.8e-5 Pa s, density rho = 1.2 kg/m^3. Reynolds
# number Re = rho v d / mu: above about 2,300 the flow is not laminar
# and the Poiseuille term is only a lower bound. Pressure is compared
# with 0.6 kPa, a typical subglottal pressure of conversational speech
# (1 cm H2O = 98.07 Pa).
from math import pi
MU, RHO, L, U = 1.8e-5, 1.2, 0.15, 0.15e-3
if __name__ == "__main__":
print(" diameter (mm) speed (m/s) Re back pressure (Pa)"
" cm H2O / 0.6 kPa")
for d_mm in (2, 3, 4, 5, 6, 8):
r = d_mm / 2000
v = U / (pi * r * r)
lam = 8 * MU * L * U / (pi * r ** 4)
jet = 0.5 * RHO * v * v
p = lam + jet
print(" {:13d} {:11.1f} {:5,.0f} {:18,.0f} {:6.1f} {:9.2f}"
.format(d_mm, v, RHO * v * d_mm / 1000 / MU, p, p / 98.07,
p / 600))
# diameter (mm) speed (m/s) Re back pressure (Pa) cm H2O / 0.6 kPa
# 2 47.7 6,366 2,399 24.5 4.00
# 3 21.2 4,244 474 4.8 0.79
# 4 11.9 3,183 150 1.5 0.25
# 5 7.6 2,546 61 0.6 0.10
# 6 5.3 2,122 30 0.3 0.05
# 8 3.0 1,592 9 0.1 0.02在 0.15 L/s 的气流下,3 毫米的吸管产生 474 Pa(4.8 cm H2O),约为对话时 0.6 kPa 声门下压的 0.79;4 毫米的产生 150 Pa(为其 0.25);6 毫米的只有 30 Pa。2 毫米的吸管在这一气流下需要 2,399 Pa,是声门下压的四倍,所以实际上会是气流下降。直径 5 毫米及以下的吸管,雷诺数都超过 2,300,所以层流项只是下限,真实背压会更高一些。决定负载的是直径,而不是练习的名称:直径减半,层流损失变为十六倍。
一条十四天的曲线
训练的说法常以曲线表达:「闭合商在十四天里上升」。p12 假设这样一条曲线:闭合商以四天的时间常数指数地趋近 0.06 的增幅,并问一个人与一个群体能测到什么。
推导
# Chapter 6. p12: A 14-day straw-phonation programme and the closed
# quotient (CQ, the share of each cycle the folds are closed, from
# electroglottography). Illustrative model: CQ(day) = 0.40 + g (1 -
# exp(-day / 4)), measured once a day with noise sd 0.03. Persons
# differ in g (sd 0.03 around the mean gain). (a) One person, mean gain
# 0.06: the change estimated from day 0 to day 14, from single sessions
# or from three-day averages. (b) Persons needed for a paired test
# (two-sided 5%, power 80%, normal approximation, rounded up) with
# three-day averages, by mean gain.
# 4,000 persons simulated, seed 99.
import random
from math import ceil, exp, sqrt
def series(rng, g):
return [0.40 + g * (1 - exp(-d / 4)) + rng.gauss(0, 0.03)
for d in range(15)]
if __name__ == "__main__":
rng = random.Random(99)
one, avg = [], []
for _ in range(4000):
s = series(rng, 0.06)
one.append(s[14] - s[0])
avg.append(sum(s[12:15]) / 3 - sum(s[0:3]) / 3)
for name, v in (("single sessions", one), ("three-day means", avg)):
m = sum(v) / len(v)
sd = sqrt(sum((x - m) ** 2 for x in v) / (len(v) - 1))
print(" (a) {:<16} mean change {:.4f}, sd {:.4f}, share < 0: {:.3f}"
.format(name, m, sd, sum(x < 0 for x in v) / len(v)))
true0 = 0.40 + 0.06 * (sum(1 - exp(-d / 4) for d in range(3)) / 3)
true14 = 0.40 + 0.06 * (sum(1 - exp(-d / 4) for d in (12, 13, 14)) / 3)
frac = (true14 - true0) / 0.06
sd_change = sqrt((0.03 * frac) ** 2 + 2 * 0.03 ** 2 / 3)
print(" (b) mean gain expected change persons needed")
for g in (0.06, 0.03, 0.015):
n = ((1.959964 + 0.841621) * sd_change / (g * frac)) ** 2
print(" {:15.3f} {:15.4f} {:14d}".format(g, g * frac, ceil(n)))
# (a) single sessions mean change 0.0583, sd 0.0422, share < 0: 0.083
# (a) three-day means mean change 0.0453, sd 0.0242, share < 0: 0.028
# (b) mean gain expected change persons needed
# 0.060 0.0453 5
# 0.030 0.0227 18
# 0.015 0.0113 69对一个人来说,用单次测量得到的第 0 天到第 14 天的变化,平均 0.0583、标准差 0.0422,虽然真实变化为正,仍有 0.083 的情形测出负值;三天平均把离散降到 0.0242(负值 0.028),但测得的变化也降到 0.0453,因为头三天已经包含部分进步。对一个群体来说,平均增益为 0.06 时五人就够,0.03 时需要 18 人,0.015 时需要 69 人。由于练习效应以及对电声门图仪的熟悉也会改变闭合商,在把任何曲线归功于练习之前,都需要一个不做吸管发声的对照组。
假说 · 尚未验证
作者的诠释(尚未验证): 未经训练的成年说话者每天用 3 至 4 毫米的吸管发声、连续十四天,其电声门图闭合商的提高比同样时长的朗读多至少 0.03。什么会证明它错: 在一项预先登记、每组至少 18 人的随机试验中,两组闭合商变化之差低于 0.03,或其置信区间包含零。
小结
吸管发声的背压由吸管直径决定:3 毫米时数百帕,6 毫米时几十帕。一条十四天的闭合商曲线,在一个人身上只有用平均后的测量才能看出;若真实增益不大,要把它归功于训练,需要约二十人或更多的对照组。
参考文献
Fan, Y. (2026). Singer Formant Clustering, Glottal Closure Quotients and Acoustic Authority Psychophysics: Subglottal Pressure Dynamics, Pitch-Floor Stability and Pause-to-Phoneme Persuasion Ratios [歌手共振峰簇、声门闭合商与声学权威度心理物理学] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090258 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).