A new wearable device has been developed to enable stroke patients with dysarthria—a motor-speech disorder—to regain natural, fluent communication.
The “intelligent throat” system, created by an international team of researchers, combines advanced sensors and artificial intelligence (AI) to process silent speech and emotional cues in real time.
The system integrates textile strain sensors, which detect vibrations from throat muscles, and carotid pulse signal monitors, with large language models for speech processing.
Unlike existing technologies, the device translates silent speech into coherent, delay-free sentences while incorporating emotional and contextual nuances.
Tested on five dysarthria patients, the system achieved a 4.2% word error rate and a 2.9% sentence error rate, significantly improving over existing silent speech systems.
Additionally, user satisfaction increased by 55%, highlighting its ability to deliver personalized, expressive communication.
“The system generates personalized, contextually appropriate sentences that accurately reflect patients’ intended meaning,” a research paper submitted Wednesday reads.
The wearable’s design features a choker embedded with graphene-based strain sensors, providing high sensitivity and comfort for daily use.
A built-in wireless module ensures continuous data transmission with minimal energy consumption, enabling all-day functionality.
LLM agents embedded in the system analyze speech tokens and emotional signals, refining and expanding sentences to match the user’s intended meaning.
The personalized approach allows for dynamic, real-time expression, bridging the gap between patient communication needs and technological capabilities, the researchers wrote.
The researchers also envision broader applications, including support for other neurological conditions like ALS and Parkinson’s and the potential for multilingual adaptations.
The team is now focusing on miniaturizing the device and integrating it into edge-computing frameworks for improved usability, they wrote.
Generally Intelligent Newsletter
A weekly AI journey narrated by Gen, a generative AI model.
Source: https://decrypt.co/293950/ai-powered-wearable-device-restores-natural-speech-for-stroke-patients
