IIT Guwahati Develops Deep Learning Sensor to Convert Exhaled Air into Voice Commands

New Delhi: Researchers at the Indian Institute of Technology (IIT) Guwahati have developed an innovative underwater vibration sensor that enables contactless and automated voice recognition by detecting exhaled air—offering new possibilities for people with voice impairments.

Developed in collaboration with Ohio State University in the US, this sensor provides a groundbreaking alternative for individuals unable to use conventional speech-based communication systems.

The research hinges on the natural airflow that occurs when a person attempts to speak—even if they can’t produce sound. As this air passes over a water surface, it creates subtle waves. The underwater sensor detects these disturbances and interprets them as speech signals without needing audible voice input.

Published in the journal Advanced Functional Materials, the study outlines how this unique method could pave the way for new, non-verbal voice recognition systems.

“This is one of the rare material designs capable of recognizing speech by monitoring the water waves at the air-water interface caused by exhaled breath,” said Prof. Uttam Manna from the Department of Chemistry at IIT Guwahati. “It presents a promising communication method for individuals with partially or fully damaged vocal cords.”

The sensor is crafted from a conductive, chemically reactive porous sponge. Positioned just beneath the air-water interface, it picks up tiny vibrations from exhaled air and translates them into electrical signals.

To analyze these signals, the team employed Convolutional Neural Networks (CNNs)—a type of deep learning model adept at recognizing complex patterns.

This technology enables users to interact with devices without making any sound, potentially transforming assistive communication systems and smart device control.

According to the researchers, the prototype costs around ?3,000 at the lab scale. They are currently exploring industry partnerships to commercialize and reduce the cost of the device.

Key highlights of the sensor include:

• Contactless communication for individuals with speech disabilities

• AI-driven interpretation using CNN models

• Hands-free control of electronic and smart devices

• Durability, with stable performance even after extended underwater use

• Versatile applications, including fitness tracking, movement detection, and underwater sensing and communication

With inputs from IANS