Plant Bioelectrical Signals for Emotion Detection

Machine learning integration for real-time plant bioelectrical signal visualization

Overview

This ongoing pilot investigation explores the use of bioelectrical signals from a Purple Heart plant (Tradescantia pallida) for dual-purpose classification: environmental state detection and human emotion recognition. The completed project will be presented at the Phaenomena Conference in Zurich on March 14th.

Problem

Translating plant bioelectrical responses into meaningful, real-time visual representations presents a unique challenge at the intersection of biology, signal processing, and human-computer interaction. Traditional analysis methods fail to capture the dynamic nature of these biological signals in ways that are engaging and interpretable.

Solution

Using an AD8232 ECG sensor at 400 Hz sampling rate, bioelectrical signals are recorded from a single Purple Heart plant and converted to mel-spectrograms for ResNet18 CNN classification. My role focuses on developing machine learning integration for real-time visualization, creating movable live animations in TouchDesigner that respond dynamically to human emotional states detected through the plant's bioelectrical signals.

Impact

• •85.4% accuracy in environmental state (lamp on/off) detection across 2767 samples
• •73% accuracy in human emotion classification (happy vs. sad) with 1-second lag across 1619 samples
• •Real-time animation system translating bioelectrical signals into interactive visual experiences
• •Project to be presented at Phaenomena Conference, Zurich (March 14th)

Key Skills Applied