关于「停顿的结构:顿笔、迟疑与三分之二幂律」,《书法运笔平滑度与分形维数》给出的定量结论是什么?

结论:第三章区分有意的顿笔与迟疑:顿笔分段后得 100.0 分,四段迟疑的笔只得 25.0 分。

数据口径:书中轨迹、纸墨与评分参数均为示意;全部计算由书中程序生成;本书是非个人化的分析,不构成经营、设计或医疗建议。 第二章证明盒计数维数看不见运笔:同一条弧线以三种时序写出,维数都是 1.2251,平滑度却从 -7.653 落到 -11.140;示意的两组笔迹中,维数的效应量只有 -0.36(95 % 区间 -0.98 到 +0.27),平滑度的效应量是 3.74;随机游走的维数 1.704 已接近滴画研究报告的 1.72。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 100.0, 25.0

reference DJ0 of a quintic cut at 5 % of peak: 225.9
logwhole N_eqpausesshareseg N_eqscore
one sweep1.02100.0001.02198.0
press-pause-sweep3.40510.2440.991100.0
hesitant sweep3.99600.0003.99625.0
tremor sweep1.49000.0001.49067.1
查看完整验算程序
# Chapter 3. p31: a deliberate pause is not a hesitation
# Model: a pen-tip log (t, x, y) sampled at 200 Hz is cut at pauses,
# runs where speed stays below 5 % of the peak for at least 40 ms.
# Each moving segment gets DJ = J T^5 / A^2 from finite differences,
# divided by DJ0, the value for an ideal quintic cut at the same 5 %
# threshold, so N_eq = (DJ / DJ0)^(1/4) and a perfect segment scores
# 100 / N_eq = 100. The whole-stroke N_eq treats the log as one piece.
# Strokes: one sweep, a press-pause-sweep (a dun, a planned stop),
# a hesitant sweep of 4 weakly overlapping pieces, and a sweep with a
# 9 Hz tremor of 0.4 px. All four are synthetic and illustrative.
import math

RATE = 200.0
THR, MIN_PAUSE = 0.05, 0.04


def q(u):
    if u <= 0.0:
        return 0.0
    if u >= 1.0:
        return 1.0
    return 10 * u ** 3 - 15 * u ** 4 + 6 * u ** 5


def sample(pieces, total, trem=0.0):
    n = int(total * RATE) + 1
    ts, xs, ys = [], [], []
    for i in range(n):
        t = i / RATE
        xs.append(math.fsum(dx * q((t - t0) / d)
                            for t0, d, dx, dy in pieces))
        yv = [dy * q((t - t0) / d) for t0, d, dx, dy in pieces]
        yv.append(trem * math.sin(2 * math.pi * 9.0 * t))
        ys.append(math.fsum(yv))
        ts.append(t)
    return ts, xs, ys


def dj0():
    k = 0
    while 30 * (k * 1e-5) ** 2 * (1 - k * 1e-5) ** 2 < THR * 1.875:
        k += 1
    lo = k * 1e-5
    n = 20000
    h = (1 - 2 * lo) / n
    js = [(60 - 360 * u + 360 * u ** 2) ** 2
          for u in (lo + (k + 0.5) * h for k in range(n))]
    a = q(1 - lo) - q(lo)
    return math.fsum(js) * h * (1 - 2 * lo) ** 5 / a ** 2


def segments(xs, ys, h):
    n = len(xs)
    sp = [0.0] * n
    for i in range(1, n - 1):
        sp[i] = math.hypot(xs[i + 1] - xs[i - 1],
                           ys[i + 1] - ys[i - 1]) / (2 * h)
    vmax = max(sp)
    mov = [s >= THR * vmax for s in sp]
    first = mov.index(True)
    last = n - 1 - mov[::-1].index(True)
    segs, pauses, a, i = [], [], first, first
    while i <= last:
        if not mov[i]:
            j = i
            while j <= last and not mov[j]:
                j += 1
            if (j - i) * h >= MIN_PAUSE:
                segs.append((a, i - 1))
                pauses.append((j - i) * h)
                a = j
            i = j
        else:
            i += 1
    segs.append((a, last))
    return segs, pauses, sp, (last - first + 1) * h


def seg_dj(xs, ys, sp, h, a, b):
    jj = []
    for i in range(max(a, 2), min(b, len(xs) - 3) + 1):
        jx = (xs[i + 2] - 2 * xs[i + 1] + 2 * xs[i - 1] - xs[i - 2])
        jy = (ys[i + 2] - 2 * ys[i + 1] + 2 * ys[i - 1] - ys[i - 2])
        jj.append((jx * jx + jy * jy) / (4 * h ** 6))
    dur = (b - a + 1) * h
    amp = math.fsum(sp[a:b + 1]) * h
    return math.fsum(jj) * h * dur ** 5 / amp ** 2, amp


def analyse(ts, xs, ys):
    h = 1.0 / RATE
    ref = dj0()
    segs, pauses, sp, active = segments(xs, ys, h)
    whole, _ = seg_dj(xs, ys, sp, h, segs[0][0], segs[-1][1])
    parts = [seg_dj(xs, ys, sp, h, a, b) for a, b in segs]
    tot = math.fsum(amp for _, amp in parts)
    neq = math.fsum((d / ref) ** 0.25 * amp / tot for d, amp in parts)
    return {"whole": (whole / ref) ** 0.25, "pauses": len(pauses),
            "share": math.fsum(pauses) / active, "neq": neq,
            "score": min(100.0, 100.0 / neq)}


if __name__ == "__main__":
    logs = {
        "one sweep": sample([(0.1, 0.5, 150.0, 0.0)], 0.7),
        "press-pause-sweep": sample([(0.1, 0.15, 30.0, 0.0),
                                     (0.37, 0.4, 120.0, 0.0)], 0.87),
        "hesitant sweep": sample([(0.1 + 0.14 * k, 0.18, 37.5, 0.0)
                                  for k in range(4)], 0.8),
        "tremor sweep": sample([(0.1, 0.5, 150.0, 0.0)], 0.7, 0.4),
    }
    print("reference DJ0 of a quintic cut at 5 % of peak: {:.1f}".format(
        dj0()))
    print("log                 whole N_eq  pauses  share  seg N_eq  score")
    for name, (ts, xs, ys) in logs.items():
        r = analyse(ts, xs, ys)
        print("{:<18s}  {:>9.3f}  {:>6d}  {:.3f}  {:>8.3f}  {:>5.1f}".format(
            name, r["whole"], r["pauses"], r["share"], r["neq"],
            r["score"]))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《书法运笔最小加加速度轨迹、飞白墨量毛细流与笔势美感分形维数母典》第 3 章「停顿的结构:顿笔、迟疑与三分之二幂律」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-229 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026229,
  author    = {Fan, Yafu},
  title     = {Minimum-Jerk Brush Trajectories, Capillary Ink Flow in Flying-White Strokes and the Fractal Dimension of Stroke Beauty},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论