Graphene-Based Wearable Intelligent Throat Enables Stroke Patients to Speak

Why It Matters

Revoice offers a scalable, non‑invasive solution to restore natural communication for stroke survivors, addressing a critical unmet need in neuro‑rehabilitation and expanding the market for AI‑enhanced wearables.

Key Takeaways

• Graphene textile sensors detect sub‑0.1% throat strain.
• Wearable choker translates silent speech into real‑time text.
• Integrated LLM expands fragmented words into fluent sentences.
• Device remains washable and motion‑artifact free.
• Trial with five stroke patients showed low error rates.

Pulse Analysis

Speech loss after stroke remains a major barrier to rehabilitation, affecting roughly half of survivors. Traditional assistive technologies rely on invasive brain‑computer interfaces or cumbersome mouth‑piece amplifiers, limiting adoption. The Revoice system sidesteps these constraints by mounting a graphene‑based textile sensor on a soft choker that captures the minute vibrations of the larynx and carotid pulse. By converting these biomechanical cues into digital signals, the device creates a non‑invasive bridge between the patient’s intent and audible language, opening a new pathway for real‑time communication.

The core of the sensor is a screen‑printed graphene film whose controlled micro‑cracks act as a resistive strain gauge. When stretched as little as 0.1 % along the throat’s expansion axis, the gauge exhibits a gauge factor near 100, delivering a linear response that outperforms conventional textile sensors. A rigid isolation frame surrounds each sensing region, confining macroscopic deformation and suppressing crosstalk, which preserves signal fidelity during everyday movements. Because the graphene ink is compatible with roll‑to‑roll printing, the choker can be mass‑produced, washed repeatedly, and retained without performance loss.

Signal acquisition is paired with a two‑stage large language model pipeline that first decodes phonetic tokens and then enriches them with contextual and emotional cues drawn from heart‑rate and time‑of‑day data. In a pilot with five dysarthric stroke patients, the system delivered word error rates below 10 % and markedly higher satisfaction scores than existing silent‑speech prototypes. This performance suggests a viable commercial route for hospitals and home‑care providers seeking scalable, non‑invasive communication aids. As AI‑driven wearables mature, Revoice could catalyze a broader market for neuro‑rehabilitation devices that blend advanced materials with real‑time language generation.