Recognise data patterns

Hey guys,

I’m using an mpu6050 sensor to get a total acceleration vector to determine the flight of a ball being thrown. I have to determine whether or not the ball was caught or dropped after being thrown. I’ve put the serial to processing for easier visualisation and am wondering how I can recognise the data patterns of these 2 situations.

In the attached image you can see the different graph of various situations. As you can see throwing the ball has a sharp acceleration followed by a drop to a steady reading. When the ball is caught it is a fairly clean reading where as when dropped it is quite noisy.

Thank for your help guys!

I see your only chance for the distinction of catch and miss by bouncing, assuming that the ball will bounce when it drops to the ground, but not when caught. A soft carpet may make it impossible to distinguish both cases.

I would try using the gyro-rate instead of acceleration. If the ball is caught, it probably won't rotate as much as while rolling on the ground.

lg, couka