Scent-Synced Movie Scenes and Viewer Memory

ISEF Category: Technology Enhances the Arts

Subcategory: Other  ·  Difficulty: Advanced  ·  Setup: School Lab  ·  Time: Full Year

The Hook

Movies already use sound and color to shape how you feel. Smell can do the same, but most screens ignore it. Your project asks a bigger question, can scent make a scene feel more real, and help you remember it better?

What Is It?

This project links a film scene to a matching smell. Think of it like subtitles for your nose. When your system sees a forest, ocean, fire, or food scene, it sends a signal to a diffuser that releases a related scent.

The computer vision part often uses a CNN, or convolutional neural network, which is a type of AI model that learns to sort images. Here, the model classifies frames from the video, then decides which scent to trigger. You are not just building a gadget. You are testing whether a sensory cue changes how people experience media.

The human side matters just as much as the tech side. You can compare scented and unscented viewings, then measure immersion, memory recall, or scene recognition. That gives you a project that mixes machine learning, interface design, and behavioral testing.

Why This Is a Good Topic

This works well for a science fair because you can test a clear cause and effect. You change one thing, the scent cue, and measure outcomes like immersion or recall. That makes the project easy to explain and easy to judge. It also connects to real problems in media design, accessibility, museum exhibits, and interactive art.

Research Questions

• How does scene-matched scent affect viewers' reported immersion compared with no scent?
• What is the effect of scent timing on memory recall for key scene details?
• Does matching the scent category to the visual scene improve recognition more than a random scent does?
• To what extent does scent intensity change comfort, immersion, and distraction ratings?
• Which scene type, forest, ocean, fire, or food, produces the strongest response when paired with smell?
• How does a CNN-based trigger system compare with manual scent triggering for consistency?

Basic Materials

• Laptop or desktop computer with webcam or video playback capability.
• Simple microcontroller such as Arduino Uno or Raspberry Pi.
• Low-cost piezo-mist diffuser.
• Food-safe or cosmetic-grade essential oils.
• Small fans or tubing for scent delivery.
• Timer or stopwatch.
• Participant survey forms or digital survey tool.
• Consent forms and debrief sheet.
• Video clips with clearly labeled scene types.
• Notebook or spreadsheet for data logging.

Advanced Materials

• Laptop or workstation with GPU access for training or fine-tuning a CNN.
• Python environment with computer vision libraries.
• Pretrained image classification model.
• Microcontroller for real-time control.
• Multiple piezo-mist diffusers with relay or driver circuits.
• Airflow sensor or odor sensor for system checks.
• Standardized scent cartridges or dilution setup.
• Calibrated rating scales for human subjects testing.
• Screen recording or logging software.
• Statistical analysis software or scripts.

Software & Tools

• Python: Handles video analysis, model inference, and experiment data processing.
• OpenCV: Extracts frames and supports scene detection from video clips.
• TensorFlow or PyTorch: Trains or runs a CNN that classifies visual scenes.
• Google Forms: Collects viewer ratings and recall responses in a clean format.
• ImageJ: Helps measure image features if you compare frame properties across scenes.

Experiment Steps

1. Define the exact scene classes you want the system to recognize and trigger.
2. Choose one delivery design for scent output and decide how you will keep it consistent across trials.
3. Build a scene-classification pipeline and verify that it labels test clips with reasonable accuracy.
4. Plan a control version that shows the same film clips with no scent or with mismatched scent.
5. Design your viewer study so immersion and recall have clear scoring rules before data collection starts.
6. Decide how you will analyze whether the scent cue changes ratings, recall, or both.

Common Pitfalls

• Using scents that linger too long, which causes later scenes to inherit smell from earlier ones.
• Letting room airflow vary between sessions, which changes how strong each scent feels.
• Choosing scene classes that look too similar, which makes the CNN trigger the wrong aroma.
• Asking viewers vague questions, which makes immersion and memory data hard to compare.
• Ignoring scent comfort and sensitivity, which can skew ratings or make participants stop early.

What Makes This Competitive

A stronger project goes beyond a fun demo. You can compare multiple trigger methods, test whether scene matching beats random scent cues, and analyze both user ratings and recall scores. A careful design with balanced controls, repeated trials, and clear statistics will stand out. The best version also checks whether the system works differently for different scene types, not just overall.

Project Variations

• Test the system with nature documentaries instead of fiction films and compare scent effects on recall.
• Replace essential oils with neutral air pulses or no scent to isolate whether smell, not extra hardware attention, drives the response.
• Compare human-coded scene triggers with CNN-triggered scene triggers to measure accuracy and viewer experience.

Learn More

• PubMed: Search for review articles on olfaction, memory, and multisensory perception to ground your viewer study in existing research.
• NIH National Library of Medicine: Use PubMed and related health resources to find studies on smell, attention, and recall.
• NASA: Search for articles on sensory perception and environment monitoring if you want background on smell detection and human factors.
• MIT OpenCourseWare: Look for machine learning and computer vision course materials to understand CNN basics.
• IEEE Xplore Abstracts: Search for papers on scent interfaces, multimedia systems, and human-computer interaction.
• NOAA: Find educational resources on ocean, forest, and fire environments if you want to match scents to scene types more thoughtfully.