Applying AI to detect gestures using Accelerometer

Hi. I was wondering how to apply accelerometer data to detect motions like clapping, waving, and things like that. Will it be possible in that regards?

How should I go about this, should I just get the accelerometer data and used it to train a model? will I need to do features extraction?

Yes, that’s possible. It’s not simple though.

On the input, you would have a time-stamped sequence of accelerometer measurements.

On the output, you would need time-stamped labels for when each action you’re trying to predict is occurring (clapping, waving, etc).

Then you could train a Sequence Model on this data set, so that you could make predictions based on new accelerometer measurements.


I do not know exactly how to do that but I remember that some banks try to use accelerometer to identify/secure users of their mobile apps.

An I found this article which may be interesting for you:

Smartphone User Identification/Authentication Using Accelerometer and Gyroscope Data

1 Like