How to derive velocity from accelerometer data?

Force

I'm trying to measure a swimmer's speed by strapping a MetaTracker to the swimmer and integrating the accelerometer data along an axis. However, the resulting velocity is humanly impossible. What are some ways to derive velocity from acceleration data?

Thanks.

Eric

Post your data. What do you expect the values to be, what are the values you are seeing?

Force

Acceleration.

Velocity.

Expected velocity, or something similar.

Eric

• Where are you getting horizontal acceleration?
• Your accelerometer data is pretty noisy, you'll want to filter / smooth out the data before consuming it.
• The links I posted in a previous thread might be of some help

Force

I'm getting the horizontal acceleration directly from the y-axis of the accelerometer outputs.

What's a best practice for filtering/smoothing out the data?

Laura

Use a high pass filter and averager to smooth out the data.

Force

Just for clarification, both the high pass filter and averager are applied to the raw acceleration values?

Also, what's a best practice for dealing with drift when integrating?

Eric

@Force said:
Just for clarification, both the high pass filter and averager are applied to the raw acceleration values?

Yes

Also, what's a best practice for dealing with drift when integrating?

There will always be drift with just raw accelerometer data. You can try to offset this with sensor fusion and filtering.

Force

What kind of filters, other than a high pass filter, would be suitable for dealing with drift when integrating?

Eric

@Force said:
What kind of filters, other than a high pass filter, would be suitable for dealing with drift when integrating?

Dunno, that's what sensor fusion is for.

trabucco

@Force said:
What kind of filters, other than a high pass filter, would be suitable for dealing with drift when integrating?

Try reading about a Kalman Filter.