3 ms·

That's most probably not lying, just statistical noise not detected properly. I made a sleep tracking app a couple of years ago, using the accelerometer built

That's most probably not lying, just statistical noise not detected properly. I made a sleep tracking app a couple of years ago, using the accelerometer built into smartphones. To normalize data between different environments (beds, matresses, phone models...), I tried to estimate activity/sleep phases compared to a low quantile over all data points, after trying to remove, or at least mitigate noise. However, if there is only noise, the algorithm will interpret that as if recorded movements were just really fine and still try to derive useful information from it. There were lots of complaints that my app would "make up" data, but this was never the case. It was just not using an absolute lower threshold of non-movements.

moftz 10y ago
But you would think that the device would be able to detect zero motion for long periods of time and stop recording data and ignore any data recorded since ceasing to move. It's not that complicated to have a rule where if there is no motion (relative to someone moving their arm) for a few hours, then the device isn't being worn or the person is dead. Naively recording what accounts to noise for long periods of time is dumb; it wastes battery and reduces trust in the product.
- blauditore 10y ago
  The problem is that while one is sleeping, body motion switches between phases of intensive, slight and no movement. Slight movements are quite close to noise, so drawing the line there is not trivial. And during certain phases, the person lays almost completely still for up to more than one hour.
  - moftz 10y ago
    Of course there are going to be periods of slight movements that appear as noise while sleeping. But the device should be smart enough to realize when there has been a very long (see 8 hours) period of inactivity. No one is staying still for that long while they sleep so the device should just assume any data collected during those 8 hours of inactivity was just the device sitting on a bedside table and that the data should be discarded. Additionally, ask the user if they want to record sleep data and let them choose the hours they typically sleep. Using that can greatly help save battery life during the night by just having the device be more liberal in its use of a low power mode when long periods of noise-level activity is detected. Use the general sleep window to put the device into a mode where it listens for the slight, noise-level movements. During the day, those movements aren't going to matter as much compared to the night so the window is important in clearly defining when they should count more.

moftz 10y ago

But you would think that the device would be able to detect zero motion for long periods of time and stop recording data and ignore any data recorded since ceasing to move. It's not that complicated to have a rule where if there is no motion (relative to someone moving their arm) for a few hours, then the device isn't being worn or the person is dead. Naively recording what accounts to noise for long periods of time is dumb; it wastes battery and reduces trust in the product.

blauditore 10y ago
The problem is that while one is sleeping, body motion switches between phases of intensive, slight and no movement. Slight movements are quite close to noise, so drawing the line there is not trivial. And during certain phases, the person lays almost completely still for up to more than one hour.
- moftz 10y ago
  Of course there are going to be periods of slight movements that appear as noise while sleeping. But the device should be smart enough to realize when there has been a very long (see 8 hours) period of inactivity. No one is staying still for that long while they sleep so the device should just assume any data collected during those 8 hours of inactivity was just the device sitting on a bedside table and that the data should be discarded. Additionally, ask the user if they want to record sleep data and let them choose the hours they typically sleep. Using that can greatly help save battery life during the night by just having the device be more liberal in its use of a low power mode when long periods of noise-level activity is detected. Use the general sleep window to put the device into a mode where it listens for the slight, noise-level movements. During the day, those movements aren't going to matter as much compared to the night so the window is important in clearly defining when they should count more.