From Pen Strokes to Sleep States: Detecting Low-Recovery Days Using Sigma-Lognormal Handwriting Features

Chisa Tanaka; Andrew Vargo; Anna Scius-Bertrand; Andreas Fischer; Koichi Kise

cs.HCMar 12, 2026

From Pen Strokes to Sleep States: Detecting Low-Recovery Days Using Sigma-Lognormal Handwriting Features

Authors: Chisa Tanaka, Andrew Vargo, Anna Scius-Bertrand, Andreas Fischer, Koichi Kise

Organizations: Osaka Metropolitan University, Osaka, Japan · University of Applied Sciences and Arts Western Switzerland, Fribourg, Switzerland · University of Fribourg, Fribourg, Switzerland · German Research Center for Artificial Intelligence, Japan Laboratory, Osaka, Japan

Abstract

While handwriting has traditionally been studied for character recognition and disease classification, its potential to reflect day-to-day physiological fluctuations in healthy individuals remains unexplored. This study examines whether daily variations in sleep-related recovery states can be inferred from online handwriting dynamics. % We propose a personalized binary classification framework that detects low-recovery days using features derived from the Sigma-Lognormal model, which captures the neuromotor generation process of pen strokes. In a 28-day in-the-wild study involving 13 university students, handwriting was recorded three times daily, and nocturnal cardiac indicators were measured using a wearable ring. For each participant, the lowest (or highest) quartile of four sleep-related metrics -- HRV, lowest heart rate, average heart rate, and total sleep duration -- defined the positive class. Leave-One-Day-Out cross-validation showed that PR-AUC significantly exceeded the baseline (0.25) for all four variables after FDR correction, with the strongest performance observed for cardiac-related variables. Importantly, classification performance did not differ significantly across task types or recording timings, indicating that recovery-related signals are embedded in general movement dynamics. These results demonstrate that subtle within-person autonomic recovery fluctuations can be detected from everyday handwriting, opening a new direction for non-invasive, device-independent health monitoring.

Figures & tables

Feature	Description	Type
$D$	Amplitude of motor command	Model parameter
$t_{0}$	Onset time of motor command	Model parameter
nblog	Number of lognormal distributions	Reconstruction index
SNR	Signal-to-noise ratio	Reconstruction index
SNR/nblog	Reconstruction accuracy per distribution	Reconstruction index

Table 1: Five basic features extracted from the Sigma-Lognormal model. The mean and standard deviation of each feature are computed, yielding ten features in total.

Figure 1: Illustration of the Sigma-Lognormal model. The velocity profile (black line) is described as a sum of lognormal distributions (yellow lines).

Panel of Figure 3 — Figure 3: Devices used for data collection

Panel of Figure 4 — Figure 4: Visualization of writing speed for three participants (P1, P7, P13). Each row represents one participant, and each column shows data from days 1, 14, and 28 of the experiment. Color indicates stroke speed (0–300 mm/s).

Panel of Figure 5 — Figure 5: Stroke order instructions for handwriting tasks

Figure 6: Distribution of target variables for each participant. Box plots show the median (red line), interquartile range (box), and minimum to maximum (whiskers). Blue dots represent individual daily data. Participants are ordered by ascending median of Total Sleep (P1–P13).

Panel of Figure 7 — Figure 7: Classification performance by task (each target variable). Each point represents one user’s result. The red dashed line indicates the baseline (0.25).

Panel of Figure 8 — Figure 8: Classification performance by timing (each target variable). Each point represents one user’s result. The red dashed line indicates the baseline (0.25).

Target variable	mean $\pm$ SD	$p$	$q$	sig.	mean $\pm$ SD	$p$	$q$	sig.
	PR-AUC				Recall@25%
Avg HRV	$0.391\pm 0.131$	0.0023	0.0046	**	$0.416\pm 0.139$	0.0023	0.0046	**
Lowest HR	$0.438\pm 0.164$	0.0001	0.0008		$0.350\pm 0.174$	0.0327	0.0374
Avg HR	$0.365\pm 0.146$	0.0031	0.0050	**	$0.344\pm 0.185$	0.0471	0.0471
Total Sleep	$0.352\pm 0.125$	0.0023	0.0046	**	$0.322\pm 0.108$	0.0199	0.0265

Table 2: PR-AUC and Recall@25% for each target variable (user mean

\pm

SD) with Wilcoxon test results.

p

: raw p-value;

q

: FDR-corrected p-value (Benjamini-Hochberg, 8 comparisons). The baseline (random classifier) is 0.25 for both metrics.

	Task effect						Timing effect
	PR-AUC			R@25%			PR-AUC			R@25%
Target	$\chi^{2}$	$p$	$q$	$\chi^{2}$	$p$	$q$	$\chi^{2}$	$p$	$q$	$\chi^{2}$	$p$	$q$
Avg HRV	2.40	0.663	0.847	2.26	0.688	0.847	0.46	0.794	0.907	6.84	0.033	0.344
Lowest HR	6.09	0.192	0.680	3.31	0.507	0.847	4.31	0.116	0.619	1.10	0.576	0.847
Avg HR	9.85	0.043	0.344	0.40	0.982	0.982	2.92	0.232	0.680	0.91	0.633	0.847
Total Sleep	2.65	0.619	0.847	0.82	0.935	0.982	1.08	0.584	0.847	2.73	0.255	0.680

Table 3: Results of Friedman tests. Tests were conducted for task effect (5 levels) and timing effect (3 levels) on PR-AUC and Recall@25% respectively (16 tests in total).

q

: FDR-corrected p-value (Benjamini-Hochberg, 16 comparisons). None remained significant after correction.

Prediction of Grade, Gender, and Academic Performance of Children and Teenagers from Handwriting Using the Sigma-Lognormal Model
Mar 12, 2026Adrian Iste, Kazuki Nishizawa, Chisa Tanaka +4Educational Assessment Educational Data Mining
Osaka Metropolitan University, Osaka, Japan · University of Applied Sciences and Arts Western Switzerland, Fribourg, Switzerland · University of Fribourg, Fribourg, Switzerland +1
Handwriting Trajectory Recovery with Diffusion Models
Jul 3, 2026Hiroki Nagamatsu, Shoji Toyota, Seiichi Uchida
Kyushu University, Fukuoka, Japan
Beyond AHI: An Interpretable Causal-Discovery-Guided Framework for Sleep Recovery in Connected Health
Jun 16, 2026Saba A. Farahani, Elahe Khatibi, Manoj Vishwanath +2Clinical Outcome Prediction
University of California, Irvine, CA, USA

Mar 12, 2026J/K move · Enter open · S save

Prediction of Grade, Gender, and Academic Performance of Children and Teenagers from Handwriting Using the Sigma-Lognormal Model

Adrian Iste, Kazuki Nishizawa, Chisa Tanaka, Andrew Vargo+3

Osaka Metropolitan University, Osaka, Japan · University of Applied Sciences and Arts Western Switzerland, Fribourg, Switzerland · University of Fribourg, Fribourg, Switzerland+1

Educational Assessment Educational Data Mining Classification

Open paper AskPDF arXiv

Digital handwriting acquisition enables the capture of detailed temporal and kinematic signals reflecting the motor processes underlying writing behavior. While handwriting analysis has been extensively explored in clinical or adult populations, its potential for studying developmental and educational characteristics in children remains less investigated. In this work, we examine whether handwriting dynamics encode information related to student characteristics using a large-scale online dataset collected from Japanese students from elementary school to junior high school. We systematically compare three families of handwriting-derived features: basic statistical descriptors of kinematic signals, entropy-based measures of variability, and parameters obtained from the sigma-lognormal model. Although the dataset contains dense stroke-level recordings, features are aggregated at the student level to enable a controlled comparison between representations. These features are evaluated across three prediction tasks: grade prediction, gender classification, and academic performance classification, using Linear or Logistic Regression and Random Forest models under consistent experimental settings. The results show that handwriting dynamics contain measurable signals related to developmental stage and individual differences, especially for the grade prediction task. These findings highlight the potential of kinematic handwriting analysis and confirm that through their development, children's handwriting evolves toward a lognormal motor organization.

Abstract

Explore similar work

Prediction of Grade, Gender, and Academic Performance of Children and Teenagers from Handwriting Using the Sigma-Lognormal Model

Handwriting Trajectory Recovery with Diffusion Models

Beyond AHI: An Interpretable Causal-Discovery-Guided Framework for Sleep Recovery in Connected Health