The Moody Foodie
Wearable sensing meets everyday cooking
CS467 Final Project

Investigating stress-reduction effects of culinary activities through wearable monitoring

We studied whether everyday cooking behaviors (snacking, assembly, cooking, baking) produce measurable changes in physiological stress (ECG, GSR) and perceived stress (modified PSS-4).

Study windowFeb 23, 2026 to Mar 6, 2026
DesignWithin-subjects
SignalsECG, GSR, Survey

Overview

Motivation

Cooking illustration

Stress is pervasive and impacts long-term health. Wearables can detect stress responses, but less work models the daily behaviors that might help people regulate stress. Cooking is a promising coping behavior: it can be mindful, sensory, and routine.

Research Question

Do different culinary activities produce different acute changes in stress? We compared four categories: snacking (≤30s), assembly (≤10m), cooking (stovetop), and baking (oven).

Hypothesis Passive prep decreases stress more than heat-based cooking.
Null No differences across culinary activities.

Methods

Participants

College juniors/seniors who cook at least once per day. Convenience sampling due to time constraints. We recorded multiple activities per participant for intra-subject comparison.

Procedure

  1. 2 minutes seated baseline (ECG + GSR)
  2. Modified PSS-4 ("right now")
  3. Complete one culinary activity
  4. 2 minutes seated post (ECG + GSR)
  5. Modified PSS-4 again

Analysis

We tested normality (Shapiro-Wilk, α = 0.05). Depending on normality, we used paired t-tests or Wilcoxon signed-rank tests with Bonferroni correction (p < 0.008 across 6 pairwise comparisons). We also reported effect sizes and ran OLS regression per metric.

Signals and Processing

ECG

Heart rate and HRV were derived using a modified Pan-Tompkins pipeline (band-pass filter, differentiate, square, moving window integration) and cleaned to physiologically realistic RR intervals (0.5s to 1.5s).

Filter5 to 20 Hz
Integration150 ms window
Min distance0.4 s

GSR

We decomposed electrodermal activity into tonic (SCL) and phasic (SCR) components. From phasic GSR, we also counted peaks and computed a zero-crossing rate.

Tonic1 Hz low-pass
Phasic0.05 Hz high-pass
Peak prominence0.0003 V

Results

Key finding
No significant predictors
Activity type did not significantly predict changes across metrics.
OLS example
HR: R² = 0.076
p = 0.73
Perceived stress
PSS-4: R² = 0.140
p = 0.480

Effect Size Shortlist (Moderate to Large)

With low power, effect sizes help surface potential trends even when p-values are not significant.

Activity Metric Effect Size Magnitude Stress
Assembly SDNN -0.710 Moderate
Assembly RMSSD -0.586 Moderate
Assembly PSS-4 0.572 Moderate
Cooking SDNN 1.025 Large ↓↓
Cooking RMSSD 1.145 Large ↓↓
Cooking Tonic GSR -0.531 Moderate
Cooking Phasic Peaks -0.695 Moderate
Cooking PSS-4 0.500 Moderate
Baking Phasic GSR 0.600 Large ↑↑
Snacking Phasic GSR 0.333 Moderate
Snacking Phasic Peaks -0.404 Moderate

Correlation Snapshot

Spearman correlation analysis suggested a strong inverse relationship between HR and GSR in several conditions, while perceived stress (PSS-4) did not consistently align with physiological markers.

Snacking HR vs GSR: ρ = -0.655 (p = 0.021)
Assembly PSS-4 vs HR: ρ = 0.708 (p = 0.050)

Takeaway

Across a small sample, we fail to reject the null hypothesis: No culinary activity affects stress levels.

The most important next step is increasing sample size and reducing sensor noise to improve power and measurement quality.

Discussion

Interpretation

Stress responses varied by person, and the same activity could look “stressful” in one metric but not another. This is consistent with psychophysiology: perceived stress does not always track physiological arousal.

Limitations

  • Small sample size (low statistical power).
  • Sensor noise and movement artifacts (ECG + GSR are sensitive to setup and motion).
  • Convenience sampling may limit generalizability.
  • Contextual stressors (exams, deadlines) not controlled.
  • Non-randomized activity order may introduce carryover effects.

Future Work

Weekday vs weekend
Test whether context changes stress responses to cooking.
Longer studies
Track chronic effects, not only acute pre/post changes.
Recipe familiarity
Compare comfort foods vs unfamiliar recipes within cooking and baking.

Team

Alison Bai
Northwestern University
Bhavi Barnwal
Northwestern University
Montse Villasenor
Northwestern University
Ludi Yu
Northwestern University