I'm currently working on a project which involves using Accelerometer data. Having searched the internet for techniques in processing the data, I am unable to find simple techniques. Is there any suggestions, of what techniques to use in processing recordings? The data recorded is to do with body motion, walking up-stairs, down-stairs, falling etc.

The advantages and disadvantages of techniques to processing the data would also be welcome.

  • 1
    $\begingroup$ What is your metric or accuracy goal for the processed accelerometer readings? Without a goal, any process would work. Also note that it is possible that any "simple techniques" might be unlikely to get you close to your goal. $\endgroup$ – hotpaw2 Mar 8 '17 at 16:54
  • $\begingroup$ Here's an example: makerzone.mathworks.com/uncategorized/… $\endgroup$ – Atul Ingle Mar 8 '17 at 18:57

One suggestion (non-simple): Obtain thousands of recordings from a broad spectrum of thousands of people engaged in your chosen activities to measure (with the activity externally logged, witnessed log book, video, etc.). Feed all that data to several contemporary machine learning algorithms (DNN, etc.), and hope for a resulting inference method that meets your targets.

| improve this answer | |

Not the answer you're looking for? Browse other questions tagged or ask your own question.