14 Nov

New: movement sensors and sensitivity – the ‘Mission Impossible’ test

New: movement sensors and sensitivity – the ‘Mission Impossible’ test

How easy is it to fool a movement sensor?

I can’t be the only person who has been working late at night in an office when suddenly I am left sitting in darkness, because the PIRs (passive infrared) movement sensor that controls the lighting has decided I’m too still.

In fact I know I’m not the only person, because an FAQ from customers who want to use our new movement sensors for measuring office space usage is how sensitive they are. And so we decided to do a little experiment in our office with two sensors, some software engineers and a very bouncy visitor.

We wanted to look at a range of activity levels and locations for sensors, including

  • distinguishing between one very active person (moving furniture around), several people talking and interacting and the classic ‘stationary deep thought’ pose beloved of software engineers
  • looking at the effect of location, in particular the difference between ceiling and wall mounted sensors. The ceiling mounted sensor was 2.7m off the ground in the middle of the workspace, and the wall mounted sensor was 1.5m off the ground on a column in a direction that covered half the office space.

Over a week, we found the ceiling sensor accurately reported all presence in the office. It is difficult to distinguish short, vigorous bursts of activity from one person – they look very like a few people interacting – but they did typically tend to be shorter in duration.

Meanwhile the wall mounted sensor did exactly what was expected and recorded all activity in the half of the office it could see. In the analysis below it is the light green trace:

Most usefully, our tests reveal it’s very unusual for anyone working at a desk to go for many minutes without some small adjustments of their body or working environment. Perhaps reaching for a pen, scratching your nose, or moving a mouse. It’s perfectly possible for someone who is focussed on their work (take a bow my software colleagues) to record no movement for a minute or more but if data is aggregated in 5 minute or larger buckets then we had a close fit between occupancy and movement.

To avoid over sensitivity, our movement sensors work with a target concept that is about the size of a human limb, so, if you are in Tom Cruise mode the best way not to trigger the sensor may be to practise moving one finger at a time at less than 1 m/minute. You might not get any work done, but that’s OK…no one will know you’re at work.

If you want a slightly more technical explanation of the sensitivity and range of our movement sensors you can find the data sheet here.