Day 4

  • by

Every morning my day starts with going out to outdoor toilette in the dark. Swelovenian days are short. This is a picture from 9am, sometime after sunrise.. and it stays like this half gray sky till 3pm. I am starting to loose the sense of the time. It is hard to get up when it is completely dark, and when you notice it is already dark outside.

Today, it is finally ready for MAG to test the sensor interface. It has 8 bend/pressure sensor attached, and we tried out on her wrists, elbows, arm joints (both on shoulders and arm pits) and knees. The sensors are directly taped on her body with kinesiology tape. These tapes are for sports taping and DM (German drag store) sells them with not much cost. It is good that it is cheap for our budget… but not so good that it makes a lot of waste… I should think of another way to do this for the long-run.

The connections looks good so far. I can monitor the sensor move on PD patch on my computer. This patch is to monitor the sensors and to send out some commands for development phase. This will all be integrated on Organelle later on.

We make the first test with the Analog Style synth patch on Organelle. I made a modification on it so it subscribe to zmq proxy to get data from the Bela and use this data as knob inputs. After trying out, we decided to use only 2 knobs: oscillator ratio and cut-offs from the sensors. The Bela reads out 8 sensors on Mag, and we train 3 poses she makes with 3 different parameter settings of the knobs using Machine Learning object (ml lib) on PD. Now “theoretically” when Mag makes these poses, the parameter shifts to what we set as example.

It is hard to get full reading range from these types of self-made textile sensors. In the first tryout, it was used as how it was read, then I added calibration function to detect min-max reading range and stretch the input range. This improved the Machine Learning results but still it is doing something else than we “taught” and seems too sensitive to switch from one output example to another. We will need to work on it tomorrow.