Page 40

EETE JUNE 2013

Sensors & Data Loggers Data Fusion: the next frontier of software integration By Pierre Jallon In the last few years, the functionality of our mobile devices has come so far that consumers now treat them more like personal assistants, using them for everything from remembering to complete a task to planning a night out to finding their way around to creating, sharing and enjoying content. In fact, consumers have come to expect their mobile devices to be intelligent enough to sense the world around them and even learn to anticipate their needs. This is the future, but while improvements in hardware in terms of better cameras, sound, battery life, and additional sensors are an indication of great progress, these will only take us so far. Software will make the difference between a device laden with many great but disparate features and a truly intelligent device able to synthesize its many features and offer users the most intuitive and seamless experience. In order to achieve this, the hardware must be raised to an intelligent level via the integration of a top grade software solution. Although the industry reached an advanced level of software integration with compelling functionalities, it remains application specific, with the ability to agglomerate data for smarter services yet to come. Advanced services and applications require more integrated data from several sources, which individual designers and developers cannot master on their own due to the breadth of data available and limited individual expertise. As such, there is a need for data fusion, which has the ability to meld data from different sources and platforms, whether the data resides on devices, on accessories, or in the cloud. Software: the key differentiator A currently significant trend is that sensors manufacturers, such as Invensense, ST Microelectronics believe that adding a software layer will increase value of their hardware, which is becoming Fig. 1: Data fusion implementation architecture (source: Movea) increasingly less expensive. In addition, microprocessor providers wish to increase the value of their package by adding intelligence to their systems using data from sensor fusion to generate data results from multiple sensors that measure many physical parameters in the intended devices’ environment. To illustrate the added value of software on a mobile device, let’s look at the Galaxy S4, that includes unprecedented software capabilities, enabling Samsung to create new gamechanging applications such as “Group Play,” “WatchON,” or even the “S Band” accessory and its activity monitoring services, enhancing user experience with their device. Given the plethora of disbursed information from all of these data sources and platforms, one now needs to meld everything for next-generation services and apps. Advanced data fusion models and application layer services can merge and process data to analyze specific situations and adapt or propose options without the user being aware. To do this, services need to combine the Web, email, GPS, calendar, etc. with embedded sensors data and information from a user’s personal-area-network data. So how is this accomplished? The first step is to merge the data, which is not easy to do because data sets come from a variety of sources that are not meant to work with each other. Higher intelligence - a collation and a central piece - is needed. For example, Movea can provide the sensor expertise at the device level as well as at the operating system level, merging different data, processes and packages making it easy for applications developers to use the data. Data fusion becomes a necessity to enable the integration of multiple sources of information and processes to generate smarter content. The data fusion platform relies on three pillars as shown in figure 1. Hardware as a data source The sensor hub is a dedicated entity that processes sensor data and activity and provides high-quality calibrated sensor data output that can be leveraged in the applications space such as continuous activity monitoring, indoor navigation, and context aware apps, requiring the sensors to be active even when the rest of the phone is idle. Sensor hub embedded models are optimized for low power consumption and use input from accelerometer, gyroscope, magnetometer and pressure sensors and deliver qualitative output data, such as auto-continuous calibration, 3D orientation, step count and posture recognition. A sensor hub implementation will not only deliver better quality data as input for advanced models, but also optimize device power consumption, externalizing the sensor data processing to a smaller, less power-hungry processor. Another data source to be considered is mobile accessories, which have begun to proliferate the market thanks to the increasing popularity of activity monitoring devices, such as Nike Fuelband, Fitbit, Ondaily, etc. Pierre Jallon is Data Fusion product manager at Movea – www.movea.com 32 Electronic Engineering Times Europe June 2013 www.electronics-eetimes.com


EETE JUNE 2013
To see the actual publication please follow the link above