Devices equipped with LSM6DSOX iNEMO sensor can provide a convenient and responsive ‘always-on’ user experience without impacting the battery life
STMicroelectronics, a leading global company in semiconductor and electronics manufacturing, announced that it has integrated machine learning (ML) technology into its advanced inertial sensor in order to enhance performance of activity-tracking and battery life in wearables and mobiles.
The LSM6DSOX iNEMO sensor consists of an ML core to classify motion data, based on the known patterns. Eliminating this first level of activity tracking from the primary processor helps in saving energy and also accelerates motion-based apps like fitness logging, personal navigation, wellness monitoring and fall detection.
Commenting on the development, Andrea Onetti, Analog, MEMS and Sensors Group Vice President, STMicroelectronics, said, “Machine learning (ML) is already used for fast and efficient pattern recognition in social media, financial modelling, or autonomous driving. The LSM6DSOX motion sensor integrates ML capabilities to enhance activity tracking in smartphones and wearables.”
Features and specifications
Devices equipped with this sensor can provide a convenient and responsive ‘always-on’ user experience without impacting the battery life. The sensor has more internal memory as compared to the conventional sensors. In addition, it is incorporated with a state-of-the-art high-speed I3C digital interface that allows longer times between interactions with the primary controller and shorter connection periods for additional saving of energy.
The sensor is simple to integrate with all the popular mobile platforms such as iOS and Android, thus simplifying use in smart devices for medical, consumer and industrial markets.
The sensor contains a 3D MEMS gyroscope and 3D MEMS accelerometer. In addition, it tracks complex movements by leveraging the ML core at low typical current consumption of only 0.55mA to reduce load on the battery.
The ML core works in combination with the integrated finite-state machine logic of the sensor to manage motion pattern recognition or vibration detection.