Handimajig

Inspiration

Prosthetic hands that respond to a user's actual neural intent are usually locked behind high-channel EEG systems and complex mechanics that make them expensive and inaccessible. We wanted to see how far we could push a "less is more" approach — could we build something that still works reliably, but at a fraction of the cost, by rethinking how much hardware a functional BCI prosthesis actually needs?

What it does

Handimajig is a cost-optimized EEG neuroprosthesis. It reads a user's brain signals through a low-channel EEG headset, decodes their movement intention using a BCI pipeline, and translates that intention into physical motion in a 3D-printed prosthetic hand. By minimizing the number of EEG channels and simplifying the hand's mechanical design, we bring the total cost of the system down while keeping it reliable enough for real-world use.

How we built it

Our pipeline runs in five stages: EEG signal capture → preprocessing → intention classification → command mapping → movement.

Signal acquisition: an EEG reader with a reduced channel count, paired with strategically placed electrodes to maximize signal quality despite fewer channels.
Intention decoding: an EEG-based BCI classifier trained to extract usable intent from a low-channel signal, squeezing more utility out of each channel.
Actuation: a 3D-printed prosthetic hand driven by servo motors and microcontrollers, with a simplified command set and mechanical design to match the decoder's output.
Evaluation: we compared different EEG channel counts, command sets, and prosthetic mechanisms against accuracy, task success, latency, reliability, and total cost, to find the best cost-to-functionality configuration.

Challenges we ran into

Low-channel EEG comes with real trade-offs — signals were often noisy, inconsistent, and harder to classify reliably with fewer channels to work with. Command recognition in particular struggled when intent signals overlapped or were ambiguous. We addressed this by refining electrode placement strategically and simplifying our command codes, which in turn let us simplify the hand's movement system to match — trading a bit of granularity for a lot more reliability.

Accomplishments that we're proud of

We're proud of designing an end-to-end system — from signal to motion — that specifically targets cost as a first-class constraint rather than an afterthought, without abandoning reliability. Grounding our design decisions in existing BCI and prosthetics research (rather than building in a vacuum) let us make informed trade-offs at every stage.

What we learned

We learned just how much can be recovered from a constrained EEG setup with the right preprocessing and classification choices, and how tightly the "cost" and "complexity" of a prosthesis are linked — simplifying the command set didn't just cut cost, it directly improved feasibility by reducing points of failure.

What's next for Handimajig

Next, we want to expand testing across more channel/command/mechanism configurations to further refine the cost-functionality frontier, and explore adapting the same low-channel BCI framework to other prosthetic systems (e.g., lower-limb devices) to test its scalability beyond the hand.