四路签名身份层与状态层的分层检验:什么参数能跨年认出一个人

Written by

in

Language

Evidence level: B

一个人的生理签名并非铁板一块:基因底色深厚的”身份层”参数(低阶脉搏谐波、IAF、QRS 形态)跨年稳定,而受状态调制的”状态层”参数(HRV、高阶谐波)随日波动——把两层拆开分层建模,是”望闻问切合参”科学化的关键一步。

正文

“认出一个人”靠什么? 脉诊、面相这类传统识别技艺,赌的是”同一个人的生理底版长期不变”。但生理底版不是铁板一块——同一段脉搏信号里,有些成分十年不变,有些成分一天三变。科学问题由此而来:能不能把”不变的部分”和”变的部分”拆成两层,分别验证?这正是”身份层 vs 状态层”参数分层假说的核心:低阶脉搏谐波、个体α频(IAF)这类基因底色深厚的参数属于身份层,跨年稳定、可作识别锚点;HRV、高阶谐波这类受年龄、情绪、疾病调制的参数属于状态层,随状态漂移、只适合状态判读。 若分层成立,则”合参”就不再是模糊的整体印象,而是可分别建模、可分别验证的两套特征。

身份层的证据链已经相当扎实,且直接指向”基因底色”。 脑电侧,IAF 在 20-31 岁与 65-80 岁两个年龄组中均通过 100 小时认知训练后保持不变,被视为稳定的神经特质标记(PMID 23551082);更大规模的双胞胎研究(237 对同卵 + 282 对异卵)显示 IAF 遗传度高达 0.81、alpha 功率遗传度 0.82,个体差异几乎全部由加性遗传解释(PMID 16338015)——这就是”身份层”的生物学底座。心电侧,460 对间隔一年以上的静息 12 导联 ECG 中,PQRST 形态模板匹配在 230 人参考库上识别率仍达 87.4-98.4%,跨年稳定(PMID 29382064)。脉搏侧,PPG 波形导数可以稳定描述个体特征(PMID 18003024),深度学习方法在 40 人实时识别中达 99% 准确率(PMID 39796830),1D 孪生网络 AUC 达 0.988(PMID 37430548)。

状态层同样有清晰的”漂移画像”。 HRV 在大规模人群队列(n=149,205)中,年龄与性别两个人口学变量解释了近 1/5 的个体差异,而生活方式、心理因素贡献不足 0.5%(PMID 29753022)——说明 HRV 与其说是”个体身份”,不如说是”人口学状态坐标”。更直接的证据来自 EEG 个体识别的跨会话衰减:同一批受试者同轮次内识别率 80-100%,换一天重新戴帽采集后显著下滑(PMID 31958785);静息脑电跨两周的识别虽仍达 80-100%,但已依赖特定频段(PMID 34588964)。2025 年一篇系统评估直言:PPG 生物识别在实验室数据集表现优异,换到真实世界、大样本、训练与测试时间分离后效果大幅下降(PMID 41471580)——这正是”状态层污染身份层”的典型表现。

分层的直接价值,是避免把”状态波动”误读成”身份改变”。 若不分层,某人的 HRV 因昨晚失眠而骤降,系统可能误判为”模板失效、此人变了”;分层之后,状态层参数的高波动被预期并单独建模,身份层参数仍稳定锚定身份——识别决策不再被一次偶然的生理波动动摇。反过来,若跨年某人的状态层参数集体位移且不可恢复,那恰恰是”此人有实质变化”(如疾病进展)的信号。这一”身份稳定 + 状态灵敏”的双层结构,也解释了为何单模态识别在真实世界总是高估自己:它把两层混在一根信号里,稳定性被状态噪声稀释。

于是检验设计变得清晰。 千人级跨年度队列中,把四路签名(ECG/PPG/EEG/IAF)的每个参数先按先验分成”身份层候选”(低阶谐波、IAF、QRS 形态)与”状态层候选”(HRV、高阶谐波),然后做三件事:其一,跨人一致性——计算同一参数在同一个人内部的跨年相关性 vs 不同人之间的差异,身份层应呈现”组内高、组间低”的聚类结构;其二,层间识别率对比——只用身份层参数建模 vs 只用状态层参数建模 vs 全参数建模,若分层成立,身份层在跨年识别任务中应显著优于状态层;其三,漂移关联——状态层参数与年龄、疾病、情绪评分的关联强度应显著高于身份层,从而验证”层”的语义划分不是任意的。三个检验一次队列全部可做,无需新设备。

“合参”的科学版本由此落地。 传统中医说”脉有根”、”五脏之气皆上注于面”,本质上是在描述一个人的”身份底版”与”状态投影”的关系。参数分层假说若被千人数据支持,就为这套直觉提供了一个可量化、可证伪的框架:身份层参数回答”你是谁”,状态层参数回答”你现在怎么样”——两者各司其职,合参才有意义。

证据与局限

已验证: IAF 是稳定神经特质、遗传度 0.81(PMID 23551082、16338015);ECG 形态跨年识别率 87-98%(PMID 29382064);PPG 可作生物特征、实验室识别率 97-99%(PMID 18003024、39796830、37430548);HRV 个体差异主要由年龄性别解释(PMID 29753022)。

部分验证: EEG 跨会话识别率明显低于会话内(PMID 31958785);静息 EEG 跨两周识别依赖特定频段(PMID 34588964);PPG 生物识别在真实世界条件下大幅衰减(PMID 41471580)——支持”状态层污染身份层”的存在。

待验证: 千人级跨年度队列中”身份层 vs 状态层”参数分层的跨人一致性;层间识别率对比是否稳定复现;身份层/状态层的语义划分是否与年龄、疾病漂移关联一致。

相关阅读

证据等级: B
参考文献: PMID 23551082, PMID 16338015, PMID 29382064, PMID 34588964, PMID 31958785, PMID 41471580, PMID 18003024, PMID 39796830, PMID 37430548, PMID 29753022


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。

Multi-Layered Verification of Signature Identity and State Layers: Which Parameters Enable Cross-Year Recognition of an Individual

Evidence level: B

An individual’s physiological signature is not monolithic: “Identity Layer” parameters with deep genetic grounding (low-order pulse harmonics, IAF, QRS morphology) remain stable across years, whereas “State Layer” parameters modulated by condition (HRV, higher-order harmonics) fluctuate daily. Separating these two layers for hierarchical modeling represents a critical step toward the scientific formalization of “inspection by observation, inquiry, and correlation.”

Main Text

What enables “recognizing someone”? Traditional diagnostic arts like pulse reading and physiognomy bet on the premise that “an individual’s physiological substrate remains fundamentally unchanged.” However, the physiological substrate is not static—within a single pulse signal, some components remain constant for decades while others shift three times daily. This raises a scientific question: Can we decompose the “constant” and “variable” components into two distinct layers for separate validation? This is precisely the core of the Identity Layer vs. State Layer parameter stratification hypothesis: Low-order pulse harmonics and individual alpha frequency (IAF)—parameters rooted in genetic makeup—belong to the Identity Layer, exhibiting cross-year stability and serving as recognition anchors; HRV and higher-order harmonics, which are modulated by age, emotion, and disease, belong to the State Layer, drifting with condition and suitable only for state assessment. If stratification holds, “correlation” ceases to be a vague holistic impression and becomes two distinct feature sets that can be modeled and validated separately.

The evidence chain for the Identity Layer is already robust and directly points to “genetic grounding.” On the EEG side, IAF remained unchanged after 100 hours of cognitive training in both the 20-31 and 65-80 age groups, regarded as stable neural trait markers (PMID 23551082). Larger-scale twin studies (237 monozygotic + 282 dizygotic pairs) showed IAF heritability reaching 0.81 and alpha power heritability at 0.82, with nearly all individual variation explained by additive genetics (PMID 16338015)—this constitutes the biological foundation of the “Identity Layer.” On the ECG side, among 460 pairs of resting 12-lead ECGs spaced by more than a year, template matching of PQRST morphology achieved recognition rates of 87.4-98.4% on a reference library of 230 individuals, demonstrating cross-year stability (PMID 29382064). On the pulse side, PPG waveform derivatives can stably describe individual characteristics (PMID 18003024); deep learning methods achieved 99% accuracy in real-time recognition of 40 individuals (PMID 39796830), and 1D twin networks reached an AUC of 0.988 (PMID 37430548).

The State Layer also possesses a clear “drift profile.” In large-scale population cohorts (n=149,205), age and gender explained nearly 1/5 of individual differences in HRV, while lifestyle and psychological factors contributed less than 0.5% (PMID 29753022)—suggesting HRV is less of an “individual identity” and more of a “demographic state coordinate.” More direct evidence comes from cross-session attenuation in EEG individual recognition: within the same session, recognition rates for the same subjects ranged from 80-100%, but significantly declined after re-capping on a new day (PMID 31958785); while cross-two-week resting EEG recognition still reached 80-100%, it relied on specific frequency bands (PMID 34588964). A 2025 systematic evaluation explicitly stated: PPG biometric recognition performs excellently on laboratory datasets but drops significantly when moved to real-world settings with large samples and separated training/testing times (PMID 41471580)—this is a typical manifestation of “State Layer contamination of the Identity Layer.”

The direct value of stratification is avoiding the misinterpretation of “state fluctuations” as “identity changes.” Without stratification, a sudden drop in an individual’s HRV due to insomnia might trigger a false judgment of “template failure, person changed.” With stratification, the high volatility of State Layer parameters is anticipated and modeled separately, while Identity Layer parameters remain stable anchors for identity—recognition decisions are no longer shaken by occasional physiological fluctuations. Conversely, if an individual’s State Layer parameters collectively shift irreversibly across years, that signals a “substantial change” (e.g., disease progression). This dual structure of “Identity Stability + State Sensitivity” also explains why single-modality recognition consistently overestimates its own performance in real-world settings: it mixes the two layers into a single signal, diluting stability with state noise.

Thus, the design for validation becomes clear. In a thousand-person multi-year cohort, each parameter of the four signatures (ECG/PPG/EEG/IAF) is first divided a priori into “Identity Layer Candidates” (low-order harmonics, IAF, QRS morphology) and “State Layer Candidates” (HRV, higher-order harmonics), then three actions are performed: First, Cross-Person Consistency—calculate the cross-year correlation of the same parameter within the same individual versus differences between different individuals; the Identity Layer should exhibit a clustering structure of “high within-group, low between-group”; Second, Inter-Layer Recognition Rate Comparison—compare modeling using only Identity Layer parameters vs. only State Layer parameters vs. all parameters; if stratification holds, the Identity Layer should significantly outperform the State Layer in cross-year recognition tasks; Third, Drift Correlation—the association strength between State Layer parameters and age, disease, and emotion scores should be significantly higher than for Identity Layer parameters, thereby verifying that the semantic division of “layers” is not arbitrary. All three validations can be performed in a single cohort without new equipment.

The scientific version of “correlation” thus takes ground. Traditional TCM describes “pulse has a root” and “Qi of the five organs all ascend to the face,” essentially describing the relationship between an individual’s “identity substrate” and “state projection.” If the parameter stratification hypothesis is supported by thousand-person data, it provides a quantifiable and falsifiable framework for this intuition: Identity Layer parameters answer “Who are you,” while State Layer parameters answer “How are you now”—both serve distinct functions, making correlation meaningful.

Evidence and Limitations

Verified: IAF is a stable neural trait with heritability of 0.81 (PMID 23551082, 16338015); ECG morphology cross-year recognition rates of 87-98% (PMID 29382064); PPG serves as a biometric feature with laboratory recognition rates of 97-99% (PMID 18003024, 39796830, 37430548); HRV individual differences primarily explained by age and gender (PMID 29753022).

Partially Verified: EEG cross-session recognition rates are significantly lower than within-session (PMID 31958785); cross-two-week resting EEG recognition relies on specific frequency bands (PMID 34588964); PPG biometric recognition attenuates significantly under real-world conditions (PMID 41471580)—supporting the existence of “State Layer contamination of Identity Layer.”

To Be Verified: Cross-person consistency of “Identity Layer vs. State Layer” parameter stratification in a thousand-person multi-year cohort; stable reproducibility of inter-layer recognition rate comparisons; consistency between semantic division of Identity/State layers and associations with age/disease drift.

Related Reading

Evidence Level: B
References: PMID 23551082, PMID 16338015, PMID 29382064, PMID 34588964, PMID 31958785, PMID 41471580, PMID 18003024, PMID 39796830, PMID 37430548, PMID 29753022


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.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *