Multisensor Machine Learning for Gestures

Offentlig prosjektsammendrag

As a world leader in ultrasound technology for embedded systems, Elliptic Labs has developed and launched the first proximity sensor using ultrasound for smartphones in 2016. Because the sensor does not require additional hardware, it can replace the existing infrared sensor. Such functionality is an essential step for having bezel-less smartphone display which has become a key design trend. In addition to proximity sensing, Elliptic has also been pushing for wider adoption of ultrasound based gesture recognition and presence sensing in both smartphones and IoT (Internet of Things) markets. To meet the increasing market demand and at the same time stay technically and cost competitive, Elliptic aims to build a new technology platform based on machine learning through the Skattefunn project. The platform will support the elements that are essential to the design and deployment of machine learning based software for airborne ultrasound on embedded platforms. They include for example, infrastructure for recording and collecting ultrasound and other types of sensor data, signal processing tools for pre and post processing, sensor fusion, optimization of machine learning classifier using artificial neural networks, and efficient implementation on power and complexity constrained computation platforms.