Evidence level: B
中医说脉有”常”——那份能代表”你是谁”的底子到底能放多久?生物识别领域已经把”特征随间隔时长衰减”做成了严格的纵向科学,脉搏谐波这份身份签名同样面临”保质期”问题:它衰减多快、何时失效、能否回卷。
正文
上一篇文章建立了”低阶谐波承载身份、高阶谐波反映状态”的谱段分工:一至三阶谐波被左心室射血波形与主动脉弹性锚定,跨时辰跨季节可复现,甚至能在 5 个心动周期内达到 AUC 0.988 的个体识别(PMID 37430548)。但这引出一个更”要命”的问题:这份身份签名到底能放多久? 今天录下的脉象,三个月、三年后还能认出是同一个人吗?中医”脉有常”的”常”,有没有明确的保质期?
生物识别领域已经替我们回答了问题的框架。 指纹识别的奠基性纵向研究用 15,597 名受试者、每人至少五份十年印痕、跨 12 年的运营数据库证明:随着两次采集间隔拉长,真实匹配分数显著下降(身份信号确实在衰减),但在运营阈值下识别准确率在 12 年跨度内保持稳定——直到图像质量变差时,时间稳定性的不确定性才急剧放大(PMID 26124106)。自动人脸识别的纵向研究得出更精确的量化:尽管真实分数随间隔衰减,0.01% 误接受率下约 99% 的受试者在 6 年内仍可被识别;而年龄、性别、种族对衰减趋势只产生边际影响(PMID 28092523)。结论很反直觉:特征”在漂”,识别力”没崩”——衰减的是相似度分数,而不是任务准确率本身。
脉搏谐波的衰减速度不是零,而是有梯度的。 把这两个纵向模板放回血流动力学:谐波谱的个体底版由心-主动脉耦合决定,而这个耦合会随血管老化缓慢重排。无创动脉搏波的经典横断面研究显示,1,005 名 2-91 岁正常人的桡动脉脉搏幅度从第一个十年到第八个十年增大约 67.5%,反射波随年龄提前返回(PMID 2598428)。径向增强指数随年龄显著上升、脉搏锐度指数随年龄下降(r=-0.55),韩国医学研究院的 528 例桡动脉分析直接证明波形形态随年龄重排(PMID 34615945)。更精细地,舒张期振荡幅度每十年下降约 2%(PMID 37038502)——这些”年龄效应”就是身份签名被缓慢改写的物理机制。 换句话说,脉搏身份签名的衰减曲线主要由”血管老化速率”驱动,其时间常数是”年”量级,而非”天/周”量级。
那么天/周/月的短期衰减呢?这才是”保质期”实验的真正空白。 目前已知的重测信度证据都集中在分钟级到周级:谐波前 11 阶在标准协议下的操作者内 ICC 超 0.9、操作者间 0.83-0.96(PMID 27170904);动脉僵硬度指标(PWV 与 AIx)同日 ICC 0.90-0.96、间隔一周复测降至 0.73-0.77(PMID 39474348);高血压人群间隔四周的 PWV/AIx 仍显示良好可重复性(PMID 15824467)。但没有任何一项研究把”个体识别 AUC”作为终点,按天、周、月、季四个间隔系统性地画出衰减曲线。 这正是本主题要填补的空白:短期衰减如果可忽略、长期衰减遵循血管老化斜率,那么”保质期”就是一条平滑下降的曲线;如果短期就有陡降,说明身份签名被状态噪声污染的程度比”低阶身份层”假设所允许的更严重。
“可逆边界”是这条曲线最有用的产物。 生物识别领域已确立”模板更新”(template update)作为对抗特征漂移的标准操作:当相似度分数跌破阈值,就用新样本刷新底版。Mayo Clinic 的 ECG 纵向研究展示了这个思路——约 35 年数字记录中,间隔 5-7 年相似度分数分布显著改变(P<0.001),据此提出”进化基线”:每次认证成功且分数达标时,把新记录渐进并入个人模板(PMID 36713596)。映射到脉搏:每个受试者需要一份”底版刷新时间表”——确定 C1-C4 低阶谐波的相似度分数随间隔的衰减斜率后,就能算出”隔多久需要重新录入一次身份底版”,把中医”常脉”变成可运维的工程参数。
可检验的设计。 同队列(数百人)按标准协议采集桡动脉谐波全谱,设置天(1/7 天)、周(30 天)、月(90 天)、季(180/365 天)多个重测间隔,用低阶谐波(C1-C4)做个体识别,画出”识别 AUC-间隔时长”衰减曲线;同时以血管老化指标(PWV、增强指数、血压)为协变量,检验衰减斜率是否被血管老化速率解释;最后做”模板更新”增益实验——间隔后是否刷新底版,决定识别力恢复多少。三种结局都是知识:衰减曲线平缓→”脉有常”有明确保质期,脉诊跨年可比;短期即陡降→身份层边界比预想更窄,脉诊必须标准化间隔;模板更新能大幅恢复→”常脉参变脉”获得工程化操作规则,体质研究可用动态底版。
为什么这对科学易学重要。 “脉有常”不再是一句经验判断,而是一条可测量的衰减曲线与一个可运维的底版刷新周期。“常”不是”永恒”,而是”有保质期的稳定”——把中医脉诊从”这一刻像不像”推进到”这份底子能信多久、何时该刷新”,正是把身份层从玄学层剥离出来的决定性一步。
证据与局限
已验证: 指纹跨 12 年匹配分数衰减但识别准确率稳定(PMID 26124106);人脸 6 年内约 99% 受试者仍可识别、年龄/性别/种族影响边际(PMID 28092523);脉搏谐波重测 ICC 0.83-0.96(PMID 27170904);动脉僵硬度间隔一周复测 ICC 0.73-0.77(PMID 39474348)。
部分验证: 脉搏波形随年龄显著重排(幅度、增强指数、锐度指数、舒张振荡幅度的年龄梯度,PMID 2598428/34615945/37038502);ECG 相似度分数间隔 5-7 年显著改变、模板更新可行(PMID 36713596)——但”脉搏身份识别 AUC 随天/周/月/季间隔的系统衰减曲线”尚无直接研究。
待验证: 低阶谐波个体识别力的多间隔衰减曲线形状与斜率;衰减是否由血管老化速率解释(PWV/血压/增强指数中介);模板更新能否恢复识别力及最优刷新周期。
相关阅读
证据等级: B(指纹/人脸/ECG 的模板衰减纵向证据与脉搏谐波老化机制有 PubMed 直接支撑;”脉搏身份识别 AUC 的多间隔衰减曲线与可逆边界”本身无先例,属研究设计提案)
参考文献: PMID 26124106, PMID 28092523, PMID 2598428, PMID 34615945, PMID 37038502, PMID 27170904, PMID 39474348, PMID 15824467, PMID 36713596, PMID 37430548
免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。
Expiry of Pulse-Based Identity Signature: Decay Curve and Reversible Boundaries of Individual Recognition with Time Interval
Evidence level: B
Traditional Chinese medicine posits that the pulse possesses a “constant” (常)—but how long does the foundational signature representing “who you are” remain valid? The biometric field has rigorously established longitudinal science regarding “feature decay over time intervals.” The pulse harmonic signature faces a similar “shelf life” question: at what rate does it decay, when does it fail, and can it be restored?
Main Text
The preceding article established a spectral division of labor: “low-order harmonics carry identity, while high-order harmonics reflect state.” Harmonics 1–3 are anchored by left ventricular ejection waveforms and aortic elasticity, reproducible across time scales and seasons, even achieving an individual identification AUC of 0.988 within five cardiac cycles (PMID 37430548). However, this raises a more critical question: How long does this identity signature remain valid? Can a pulse recorded today still identify the same individual after three months or three years? Does the “constant” (常) in traditional medicine have a defined shelf life?
The biometric field has already provided a framework for answering this question. Foundational longitudinal studies in fingerprint recognition, utilizing 15,597 subjects with at least five prints each and spanning a 12-year operational database, demonstrated that while real-match scores significantly decline as the interval between acquisitions increases (confirming signal decay), recognition accuracy remains stable within a 12-year operational threshold—until image quality degradation amplifies uncertainty regarding temporal stability (PMID 26124106). Longitudinal studies in automatic face recognition yielded more precise quantification: despite real-score decay, approximately 99% of subjects remain identifiable within a 6-year window under a 0.01% false acceptance rate; meanwhile, age, gender, and race exert only marginal effects on the decay trend (PMID 28092523). The conclusion is counterintuitive: features are “drifting,” but recognition capability has not “collapsed”—the decaying metric is the similarity score, not the task accuracy itself.
The decay rate of pulse harmonics is non-zero but exhibits a gradient. Returning these longitudinal templates to hemodynamics: the individual baseline of the harmonic spectrum is determined by heart-aortic coupling, which slowly rearranges with vascular aging. Classic cross-sectional studies of non-invasive arterial pulse waves in 1,005 normal adults aged 2–91 showed that radial pulse amplitude increased by approximately 67.5% from the first to the eighth decade, with reflected waves returning earlier with age (PMID 2598428). The radial enhancement index significantly rises with age, while the pulse sharpness index declines (r=-0.55); a direct analysis of 528 radial arteries by the Korea Advanced Institute of Science and Technology confirmed waveform morphological rearrangement with age (PMID 34615945). More finely, the amplitude of diastolic oscillations decreases by approximately 2% per decade (PMID 37038502)—these “age effects” constitute the physical mechanism slowly rewriting the identity signature. In other words, the decay curve of the pulse identity signature is primarily driven by “vascular aging rate,” with a time constant on the order of “years” rather than “days/weeks.”
So, what about short-term decay over days, weeks, or months? This is the true gap in “shelf life” research. Existing retest reliability evidence focuses on minute-to-week intervals: the first 11 harmonics show an intraclass correlation coefficient (ICC) exceeding 0.9 under standard protocols for intra-operator and 0.83–0.96 for inter-operator (PMID 27170904); arterial stiffness indicators (PWV and AIx) show ICCs of 0.90–0.96 on the same day but drop to 0.73–0.77 after one-week retesting (PMID 39474348); PWV and AIx in hypertensive populations still demonstrate good reproducibility at four-week intervals (PMID 15824467). However, no study has used “individual identification AUC” as the endpoint to systematically plot decay curves across daily, weekly, monthly, and quarterly intervals. This is precisely the gap this topic aims to fill: if short-term decay is negligible and long-term decay follows vascular aging slopes, the “shelf life” is a smooth downward curve; if there is a steep drop even in the short term, it indicates that identity signatures are contaminated by state noise to a degree more severe than the “low-order identity layer” hypothesis allows.
The “reversible boundary” is the most useful product of this curve. The biometric field has established “template updating” as a standard operation against feature drift: when similarity scores fall below a threshold, the baseline is refreshed with new samples. A Mayo Clinic longitudinal ECG study illustrated this approach—among approximately 35 years of digital records, the distribution of similarity scores significantly changed over intervals of 5–7 years (P<0.001), leading to the proposal of an “evolutionary baseline”: progressively integrating new records into the personal template upon successful authentication and score sufficiency (PMID 36713596). Mapping this to pulse: each subject requires a “baseline refresh schedule.” Once the decay slope of similarity scores for low-order harmonics (C1–C4) over time intervals is determined, one can calculate “how often the identity baseline needs re-entry,” transforming the traditional concept of a “constant pulse” into an operational engineering parameter.
Testable Design. Collect the full harmonic spectrum of radial arteries from a cohort (hundreds of individuals) using standard protocols, setting multiple retest intervals at days (1/7 days), weeks (30 days), months (90 days), and quarters (180/365 days). Use low-order harmonics (C1–C4) for individual identification to plot the “Identification AUC vs. Time Interval” decay curve. Simultaneously, use vascular aging indicators (PWV, enhancement index, blood pressure) as covariates to test whether the decay slope is explained by vascular aging rates. Finally, conduct a “template update” gain experiment—determine how much recognition capability recovers after refreshing the baseline post-interval. All three outcomes yield knowledge: a flat decay curve implies a clear shelf life for “constant pulses,” allowing cross-year comparison; a steep short-term drop suggests narrower identity layer boundaries than anticipated, necessitating standardized intervals for pulse diagnosis; significant recovery via template updating provides engineering rules for “constant pulses with variable parameters,” enabling dynamic baselines in constitution studies.
Why this matters for scientific Yijing Studies. “The pulse has a constant” is no longer just an empirical judgment but a measurable decay curve and a maintainable refresh cycle. “Constant” does not mean “eternal,” but rather “stable within a shelf life.” Moving pulse diagnosis from “does it look like this right now?” to “how long is this baseline trustworthy and when should it be refreshed” is the decisive step in stripping the identity layer from the realm of metaphysics.
Evidence and Limitations
Verified: Fingerprint match scores decay over 12 years but recognition accuracy remains stable (PMID 26124106); ~99% of subjects identifiable within 6 years for faces, with marginal effects from age/gender/race (PMID 28092523); Pulse harmonic retest ICCs range from 0.83–0.96 (PMID 27170904); Arterial stiffness indicators show ICCs of 0.73–0.77 after one-week retesting (PMID 39474348).
Partially Verified: Pulse waveforms significantly rearrange with age (gradients in amplitude, enhancement index, sharpness index, diastolic oscillation amplitude; PMID 2598428/34615945/37038502); ECG similarity scores significantly change over 5–7 year intervals with feasible template updating (PMID 36713596)—but no direct study exists for “systematic decay curves of pulse identity AUC across daily/weekly/monthly/quarterly intervals.”
To Be Verified: Shape and slope of multi-interval decay curves for low-order harmonic individual identification; whether decay is explained by vascular aging rates (mediated by PWV/blood pressure/enhancement index); whether template updating restores recognition ability and the optimal refresh cycle.
Related Reading
- Spectral Division of Pulse Harmonics: Low-order Carry “Who You Are,” High-order Reflect “How You Are Now”
- Retest Consistency: Rhythm-Corrected Spectrum vs. Raw Spectrum (Stability Boundaries of Individual Signatures)
- Circadian and Seasonal Stability of Pulse Spectra
- Research Pool: Biological Frequency Codes
Evidence Level: B (Longitudinal evidence of template decay in fingerprints/faces/ECGs and pulse harmonic aging mechanisms have direct PubMed support; “multi-interval decay curves of pulse identity AUC and reversible boundaries” itself has no precedents, representing a research design proposal)
References: PMID 26124106, PMID 28092523, PMID 2598428, PMID 34615945, PMID 37038502, PMID 27170904, PMID 39474348, PMID 15824467, PMID 36713596, PMID 37430548
Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.
Leave a Reply