Non-Invasive Device for Voice Restoration After Laryngectomy

Background & Unmet Need

• Laryngectomy is a procedure in which part or all of the larynx is removed from a patient, affecting speech
• The main voice restoration options following a laryngectomy are esophageal speech, the electrolarynx, or tracheoesophageal puncture (TEP)
• These options have drawbacks as the electrolarynx is often noted to sound robotic, and TEP is associated with complications such as leakage
• Even with these voice restoration options, patients with laryngectomy experience limited vocal capacity and decreased vocal control
• Unmet Need: a voice restoration device with better vocal control, intensity, and intelligibility for laryngectomy patients

Technology Overview

• The Technology: a novel, personalized device for voice restoration using machine learning applied to surface EMG (sEMG) signal
• The inventors have created a tailored device to conform to a patient's unique anatomy with sensors on the articulatory muscles of the face and neck
• The device detects the sEMG signals and applies a predictive machine-learning model to translate silent speech into words
• PoC Data: The inventors collected data using this device from a laryngectomy patient silently articulating ‘Tedd’ and ‘Ed’
• The team trained a predictive model for automatic speech recognition of these words, which had an 86.4% word recognition accuracy

Technology Applications

• Voice restoration for laryngectomy patients
• Silent speech recognition for noisy or difficult environments

Technology Advantages

• Portable, all-in-one device
• Inconspicuous profile to be held in place only during speech, like a phone
• Personalized to patient based on 3D scan of head and neck geometry