Testing controlling methods

We had a meeting on Tuesday where we prepared the content and slides for the Friday’s presentation. In addition, we discussed how we could improve our control glove. One of the current difficulties is how we could use sensors such as gyroscopes and accelerometers to detect the movement of the glove.

On technical side:

We experimented with gyroscope and accelerometer for controlling computer cursor by creating a mobile app and a server for the computer being controlled. The mobile app sends sensor data to the computer and the server program parses the data and moves the mouse accordingly. We would like to use the accelerometers to deduce hand position for controlling the cursor, but that kind of data is prone to errors due to double integration. Some ideas to reduce error in accelerometer measurements:

  • Multiple sensors
  • Resetting data integration when the movement stops
  • Filtering data for example with Savitzky-Golay filter
  • Using machine learning methods, such as neural networks

Another option is to use the gyroscopes and control the cursor by tilting the hand. This is demonstrated with a mobile phone in the video below.

The controls here are a bit clumsy, but with a wearable controller it would be better.

Next in our plans is to perform some interviews with potential customers. We will ask them, for example, about features they would like to see in such a product and how much they could see themselves paying for a working device.

Leave a Reply