Turn Noisy Sensors into Clean Data

Often data coming from sensors is unstable or noisy. For example, the pressure sensors under the stomp pads were jittering around a lot depending on where a foot was placed. So we need to get clean triggers from the input devices, looking for real gestures and trying to eliminate unwanted triggers. At the IDC, some of the data cleaning was done in the Opcode MAX music software, and some in Houdini.

In Houdini CHOPs there are several tools to clean data. It can make an envelope around noisy data, it can apply various filters to data, it can lag the data, and it can look for thresholds. Here we look for thresholds.

The first channel is sample data from an input device. Then second channel is the result if we just use .2 as a threshold for ON. There are too many spikes. If we raise the threshold, we miss events.

The Trigger CHOP is a lot smarter about detecting events. The third channel triggers and holds a peak for a period, then it looks for another trigger. But there is a case when the players moves was not smooth, so the next trigger was unintended.

This is fixed in the fourth channel where we set the minimun time of .5 seconds before the next possible trigger.

Finally in the last channel, the trigger is shaped to allow an effect to ease on, peak out, hold, and decay slowly. This is used in Trick 9.