Problem and Solutions
Discovery of the Problem
We like playing with integrating real world electronics into games and we wanted to experiment with EEG headsets, from knowing only the basic fundamentals. After some research, we came up with the ideal specs of: 8-16 channels, dry electrodes, and modifiable.
We started looking into what made the headsets so expensive, and we came up with two key inflated parts of the cost: The probes themselves (~$40 per probe), which we believe is primarily because of the very
Derived Problem Statement
The current limiting factor to current EEG technology is the inefficient use of expensive analog to digital converters, making large probe counts virtually impossible due to the cost.
Existing Solutions
Industry Standard
The industry has thus far been restricted to enterprise/research use cases, which means that they just bite the bullet and spend whatever it costs to get the headset they need. This means that they have a direct line from the probe to an A->D converter on the controller board.
Positives
- Simple to engineer
- Least possible chance for noise
- Usually comes with software support
- Almost sane pricing in small probe counts (<$1000)
Negatives
- Very wasteful in parts, meaning as probe count increases, so does the cost, linearly
DIY Microcontroller Method
This is the alternative method. It uses amplifiers connected to general purpose microcontrollers and their embedded analog inputs, for one or two channels. The most common microcontroller of choice is the arduino uno or similar. These aren’t designed to be scientific devices, but rather as an introduction to EEG and can do basic detection.
Positives
- Very cheap due to mass production of microcontrollers.
- Great as an introduction to the technology
Negatives
- Not very accurate (Noisy and low resolution)
- Low channel counts mean reduced precision
Opportunity
We saw an opportunity to improve in this space by reducing the number of expensive A->D converters. After a bit of deliberation, we decided to bring the concept that cameras use to EEG. By having a switching system that can quickly iterate through all of the probes in use, faster than the frequency of brainwaves, 300Hz, thereby allowing us to use only one really expensive A -> D converter that runs at a high speed to service multiple probes. We came up with some different ways of accomplishing this.
Many Probes to Many Multiplexer channels to One A->D
This is the simplest way to increase the number of probes. Each probe is mapped directly to its own channel on the multiplexer. This does not add much complexity to the design, but it increases the ratio of probes to A->D converters to a maximum of 16:1.
Positives
- Not much extra complexity
- Dramatically reduces cost
Negatives
- Multiplexers tend have a max of 16 inputs, reducing the theoretical limit of the approach
- Multiplexers could introduce significant noise into the super-low voltage data lines.
This is the more complicated solution, but with a vastly increased theoretical limit. Instead of using the multiplexer to switch between the inputs, it is possible to use it in reverse as a “demultiplexer”, and use it to switch amplifiers located directly on the probes. This allows a grid layout to select one probe out of many, which means that instead of the other method’s 1:1 multiplexer channel to probe ratio, the multiplexer channel count can be calculated with f(x)→⌈2x⌉, and x is the probe count.
Positives
- Multiplexer channels is a exponential falloff as more probes are added
- Dramatically reduces costs
Negatives
- Significantly more complex
Why we chose to do this
Implement EEG into our games and tools
We have played around before with implementing external mechanics into games, this kind of like the reverse of when we made a horror game with spiders that bite with a tens machine
To learn, explore and play
We always love to learn and try new technologies, and we thought of some fun stuff that we could do with an EEG headset, on the AI side of things.