Researchers at the University of Sussex in the UK have devised an algorithm that helps smartwatches track activity more effectively by learning new movements as they happen, rather than just having a limited number of pre-programmed activities they can recognize.
Wearable devices to track movement are currently very popular. At the moment, these devices can recognize when you are doing specific activities, such as yoga, but only if these types of movements have been pre-programmed into the device.
“Current activity-recognition systems usually fail because they are limited to recognizing a predefined set of activities, whereas of course human activities are not limited and change with time,” said Dr. Hristijan Gjoreski of the University of Sussex. “Here we present a new machine-learning approach that detects new human activities as they happen in real time, and which outperforms competing approaches.”
Traditionally, these types of wearable devices group bursts of activity together as a crude estimate of the activity a person has been doing, and for how long. So, for example, a series of steps could be considered a walk. However, existing devices do not account for pauses or interruptions in the detected activities, and so a typical device would consider a walk with two short stops as three separate walks, potentially providing misleading data. The new algorithm tracks activity more accurately, paying close attention to transitions between different types of activities, as well as the nature of the activities themselves. For example, it can recognize a short pause in a walk for what it is, and will hold the data while it waits for the walk to resume.
The algorithm can learn and account for a much greater range of activities, from brushing your teeth to chopping vegetables. This means that it can provide more representative data on movement. “Future smartwatches will be able to better analyze and understand our activities by automatically discovering when we engage in some new type of activity. This new method for activity discovery paints a far richer, more accurate, picture of daily human life,” said Daniel Roggen, another scientist involved in the research. “As well as for fitness and lifestyle trackers, this can be used in healthcare scenarios and in fields such as consumer behavior research.”
The scientists will present the system at the International Symposium on Wearable Computers in Hawaii, USA, in September.
