Hardware

Custom 3D-printed PETG chassis housing an 8-channel OpenBCI Cyton board sampling at 250Hz. Eight gold cup electrodes maintain constant pressure across facial muscles, preventing signal loss in noisy environments. Final weight: ~420g.

Signal Processing

0.5-45Hz bandpass filtering with Independent Component Analysis (ICA) removes artifacts like eye blinks and jaw clenches. PCA-LDA dimensionality reduction compresses 256 features into 45 discriminative dimensions.

Machine Learning

CNN-LSTM architecture trained on 440 phoneme utterances, generating 3,220 training samples. Achieves 82% single-subject phoneme classification accuracy and 68% continuous speech accuracy.

Performance

Under 300ms end-to-end latency. Maintains signal quality across 2-hour sessions. LLM integration provides context-aware error correction for natural output.

Explore the Device

Interact with our 3D model to see VOCL's ergonomic design and electrode placement.

Loading 3D model...

This may take a moment for large models (3.2MB)

💡 Tip: Drag to rotate, scroll to zoom. The model may take a few seconds to load.

Ergonomic Design

Custom 3D-printed PETG chassis optimized for comfort during extended wear sessions.

8-Channel Electrodes

Gold cup electrodes positioned at key facial muscle locations for optimal signal capture.

Lightweight

Final weight of ~420g ensures comfortable long-term use without fatigue.

Validation Results

1,847 Words Successfully Decoded
82% Phoneme Classification Accuracy
<300ms System Latency
7 Prototype Iterations

How VOCL Compares

AAC Tablets Brain Implants VOCL
Cost $10,500 $50,000-$100,000 ~$2,500 target
Speed 10-15 WPM 62 WPM 30-50 WPM target
Invasive No Yes (surgery) No
Abandonment 30-50% N/A TBD