Intelligent Throat Wearable Syaytem (Adaptaded from paper)
What is the news?
The researchers have developed a new wearable that could help people with neurological diseases regain the ability to communicate naturally and fluently without the need for invasive brain implants.
What is the problem?
Neurological diseases such as stroke, amyotrophic lateral sclerosis (ALS), and Parkinson’s disease often cause dysarthria, a motor speech disorder that interferes with normal speaking abilities.
Patients may know what they want to say but cannot vocalize it due to weakened neuromuscular control, resulting in lower quality of life.
What are the existing solution?
Augmetive and alternative communication (AAC) technology, which utilise head and eye tracking to translate the speech letter-by-letter. While this technology is easy to implement, but suffer from slow communication.
Brain computer interfaces devices, relied on invasive surgeries and complex signal processing.
What solution researcher develop?
The team created an AI-driven wearable Intelligent Throat (IT) system, a soft wearable around neck.
The IT system captures laryngeal throat muscle vibrations (20-200 Hz) to detect silent speech and cartoid pulse signals (0.5-5 Hz) to detect emotional state analysis.
The signals are captured by ultrasensitive textile strain sensors, fabricated using advanced printing techniques.
Later, signals are processed by a large language model (LLM) and machine learning networks to decode and expand the patient’s intended speech into fluent, context-aware spoken sentences in real time.
How the IT system work (Adapted from paper)
How well the device work?
Using machine learning trained on datasets from 10 healthy participants, the system achieved a word error rate of 4.2%, meaning ~96% of words were decoded correctly.
After fine-tuning on data from five stroke patients, the system reached a sentence error rate of just 2.9%, indicating that about 97% of sentences were decoded accurately.
Additionally, incorporating emotional statement analysis led to an ~55% increase in user satisfaction with the IT system.
How does the device compare to existing solutions?
Faster communication: It captures signals continuously rather than letter-by-letter or slow word-by-word systems.
Non-invasive and portable: Unlike brain-computer interface implants, the IT is lightweight and wearable like a choker.
Context-aware output: Thanks to LLMs, it doesn’t just transcribe silent mouthing, it expands phrases into meaningful, emotionally expressive sentences, closer to natural speech.
What are the limitation of the device?
The current study only involved a small sample (n=5), larger clinical trials are needed to validate robustness across diverse users.
Broader language support and improved AI-based emotional/context modeling will be needed before widespread adoption.
Long-term usability and comfort in day-to-day life remain to be demonstrated.