Everyone knows that data scientists love data and the more of it, the greater the love. As a result, the surging interest in wearables is just what the doctor ordered because these electronic devices collect enormous treasure troves of data. In turn, it is the job of data scientists to make sense of it all, unlock secrets, and assign economic value. As a data scientist, it is a dream come true!
Today’s wearable apps are embracing the 3 Vs of big data to translate growing data stores containing telemetry data originating from wearable devices into significant value for users. Healthcare is one industry seeing immeasurable benefit from wearables. One group of researchers at UCLA, for example, is using sensor data from an iPhone for characterizing Parkinson’s disease tremor through a wireless accelerometer application. We’ll see later that preventative healthcare (in the form of fitness and training) is also receiving attention from the wearables industry.
Deluge from Sensor Data Sources
Wearables are included in a growing list of special devices with variety of embedded sensors. Depending on the specs of the wearable hardware to which the app interfaces, you potentially have access to a myriad of sensors – accelerometer, gyroscope, geolocation, magnetometer, pressure, altimeter, temperature, electrodermal response as well as security and health biometrics. The wearable app takes these disparate data points from sensors to capture a view of data describing users in a multifaceted manner including their behavioral, cognitive, social, and health profiles. These models will then drive a predictive, classification engine using statistical learning methods. The key here is that more data always trumps clever algorithms, but with wearables that doesn’t appear to be a problem – sensor data collection is becoming pervasive.
Developers are an important term in the wearables equation. For example, the Samsung Architecture Multimodal Interactions (SAMI) is an open sensor data platform designed to aggregate health data from Samsung (Simband) and non-Samsung sensor laden devices in order to give developers API access to granular sensor data to run analytics. Additionally, Apple plans to launch its HealthBook health-monitoring software for its iOS8 mobile operating system. The software would monitor and collect data like heart rate, blood pressure and sleep through the iPhone’s sensors.
The core value of your wearable app will be the actionable intelligence derived from sensor data sources. Fortunately, there are powerful, scalable distributed computing platforms and open source tools like Hadoop and Spark to control, configure and run deep learning on both streaming and historical data to yield insights into patterns and outliers.
Big data analytics with machine learning requires building predictive models through intensive model training (supervised learning). Using statistical pattern recognition, predictive analytics is able to discover patterns in historical data that can predict what would happen with out-of-sample data. Techniques such as random forests, linear regression, logistic regression, decision tree classifiers, neural networks, support vector machines, k-means clustering, K-nearest neighbors, and naive Bayes classifiers can be used to build predictive models.
As the algorithms learn over time, the modeling takes an iterative approach. New data is used to go back and train the algorithm further to improve its predictive power. But wearable users will never see this complexity. What they will experience is the benefit from the sophistication in terms of unparalleled intelligence.
A Use Case from a Wearables Start-up: Focus
One early stage start-up firmly positioned to find success in this space is Focus. The company is developing the next generation of motion tracking using current wearable devices . The initial Focus app, TRAINR, is targeted for the fitness and athletic market – aspirational gym-goers, athletes, and fitness enthusiasts. Their software automatically identifies and records repetitions, sets, rest periods, and exercises/motions. Focus combines wearable technology with data processing and exercise regimens designed by professional trainers and coaches to provide real-time feedback to help users reach their athletic, aesthetic, fitness, or health goals more quickly and more effectively.
Focus uses digital signal processing algorithms as well as supervised and unsupervised statistical learning models to utilize sensor data coming from the accelerometer and gyroscope in the wearable device. The company had distinct challenges to overcome including: multi-rate and adaptive filter techniques, the need to define filter specifications from signal analysis and noise characteristics, power and resource constrained systems, plus all the difficulties involved with embedded programming in iOS and Android environment.
FocusMotion is an innovative solution that makes it easy to track and analyze movements with wearables. The firm’s hardware and OS agnostic SDK unlocks new consumer insights and behaviors. The product does all of the complex signal processing, statistical learning, and device integration work so developers can focus on making engaging applications.
[Source: Fast and Furious: Why Millisecond Wearable Interactions Will Have You Wanting More]
Sign up for the free insideBIGDATA newsletter.