Detecting human activity on mobile and wearable devices

7th April 2016

Joe Bush

0 0

To aid the development of motion sensing applications, STMicroelectronics has introduced three additions to its Open.MEMS portfolio of free and software libraries.

The new libraries allow designers to combine ST’s motion sensing technology with the range of price/power/performance options offered by the STM32 ARM Cortex-M 32-bit microcontroller family. This provides a suitable route to implementing contextual awareness in mobile, wearable and IoT (Internet of Things) applications.

The new software allows the detection of human activities from data acquired by inertial sensors embedded in the end user equipment. Optimised to minimise power consumption, they are particularly suited for fitness and healthcare applications in portable or wearable platforms that monitor human physical activities in real time over long periods.

The three new software packages are:

The osxMotionAR Activity Recognition package is an algorithm that identifies the user activity from a range of movements and transportation scenarios such as stationary, walking, fast walking, jogging, cycling and driving. Exploiting the high precision of ST’s LSM6DS3, LSM6DS3H and LSM6DSL inertial modules, the Activity Recognition algorithm manages the data acquired from the sensors at a low sampling frequency and returns the identified activity in real time with a low power consumption.
The osxMotionCP Carry Position package detects how the device containing the motion sensors is being carried. For example, the algorithm can detect whether a portable device such as a mobile phone is placed on a desk, held in hand to view the display or in a swinging arm, near the user’s head, or put in a shirt or trouser pocket. To minimise power consumption, sensor data is acquired at a low sampling frequency (50Hz).
The osxMotionGR Gesture Recognition package recognises the actions carried out on a mobile or handheld device, including pick-up, glance or wake-up, which allows designers to develop controls for different functions on the device. This algorithm acquires data from inertial modules with a sampling frequency of 100Hz and recognises the gestures carried out by the user platform in real time.