Let's start with basic fitness monitoring for individuals who like to know how they are doing in daily activities. This is general-health monitoring as opposed to the type of monitoring performed during strenuous sporting activities.
We want to measure, track, and even alarm physiological factors such as temperature, heart rate, motion/activity level, and perhaps exertion and overheating. This information can be used, in part, to determine health-related characteristics such as calorie burn, sleep profile and number of steps taken over a defined time period.
In order to achieve a level of efficacy that keeps them out of the "gadget" classification, the measurement devices need to take on the role of lifestyle monitors that are personalized to our individual needs. They must be devices that fit in to our way of life non-intrusively and deliver on their promise of providing useful, accurate data.
To do this, they use strategically located sensors to monitor vital signs such as skin temperature, galvanic skin response, motion, heart rate, calories burned, activity level, and exertion. Whether the user is sitting, walking, driving, sleeping, working, or working out, these body-mounted monitors reveal calories burned and much more. When combined with information about food intake (calories consumed), a profile of lifestyle and fitness is developed.
Supporting the needs of the professional and amateur sports enthusiast provides a tougher challenge to system developers. Measurements taken during inactivity or carrying out moderate tasks is challenging enough, but achieving the same level of accuracy, consistency, and credibility while running, swimming or generally being pounded on, requires significant additional post-processing of the data using sophisticated algorithms, to address and eliminate the many artifacts which result from such motion.
It's important to note that it is not just these extreme user situations which need insight into the nature of the sensor output and significant signal analysis and processing. Consider a "fall detector" aimed primarily at the elderly living alone. What makes such detectors possible, among other factors, are ICs such as the ADXL362 from Analog Devices, an ultra-low power, ±2/±4/±8g, 3-axis MEMS accelerometer that consumes less than 3 µA across its full range of output data rates, and just 300 nA in motion-triggered wake-up mode.
This device forms the sensing core of a continuous motion-and-fall detector, an application where portability and long battery life are critical features of the product. Its internal 12-bit ADC provides 1 mg/LSB resolution on the 2g range, and interfaces to a microcontroller via an SPI port.
But it takes more than a sensor alone for a successful end product. OEM designers must implement algorithms which interpret and act on the waveforms provided by the device, and these waveforms are not trivial: they occur along multiple axes and with uncertain timings (Figure 1).
Figure 1: Complex waveforms captured by a three-axis accelerometer during a "fall."
Click on image to enlarge