The system requires only a few minutes of training data, such as reading a series of numbers, to learn a user’s speech patterns. Once trained, it uses sound waves sent and received across the user’s face to sense mouth movements and analyzes echo profiles in real time using a deep learning algorithm, achieving an accuracy of about 95%.
One of the key advantages of the system is its privacy features. The data processing is offloaded wirelessly to the user’s smartphone, ensuring that no data leaves the phone and eliminating privacy concerns. The current version of the system offers around 10 hours of battery life for acoustic sensing, making it practical for everyday use.
