The interface between
human and machine.
We develop intelligent neuroprosthetic systems that interpret biological signals, learn individual movement patterns, and translate human intent into precise physical action.
- 01HUMANIntent
- 02BIOLOGICAL SIGNALRaw EMG
- 03SENSORAcquisition
- 04THALAMOInterface
- 05PROCESSINGEnvelope
- 06INTENTClassification
- 07PROSTHESISCommand
- 08MOVEMENTActuation
01/ The problem
Today’s prostheses replace a limb.
We want to restore an interface.
INTENT
Understand what the user wants to do.
Active research
CONTROL
Translate biological signals into precise movement.
Prototype
FEEDBACK
Ultimately return meaningful information back to the user.
Long-term vision
Feedback is a direction we are working toward, not a capability we have today.
02/ The system
From signal to movement.
Six stages sit between a muscle contraction and a moving hand. Each one is its own engineering problem.
Biological signal
Electrical activity generated by the human body contains information about intended movement.
03/ The interface
The interface.
Three technological areas define the system. Each sits at a different stage of maturity — and we are explicit about which is which.
Signal acquisition
Current prototype work with surface EMG and biological signal acquisition. Electrode placement, amplification and envelope extraction that stay stable outside a lab bench.
Advanced sensing
A research direction toward higher-density and eventually implantable sensing technologies, capable of resolving finer motor detail than surface electrodes allow.
Intelligent decoding
Machine learning models that map biological signals to intended movement, and adapt to the individual wearing the system rather than the average of a dataset.
THALAMO is a research and development project. Nothing described here is a medical device, none of it is clinically approved, and no implantable system has been used in a human.
04/ Trajectory
What exists, and what does not.
The further right you read, the less certain it becomes. We would rather be exact about that than impressive about it.
NOW
In the lab today
- Surface EMG
- Prototype development
- Signal classification
NEXT
Actively working toward
- Advanced sensing
- Improved signal resolution
- Personalised control
FUTURE
Research direction
- Implantable interfaces
- Long-term biological signal acquisition
VISION
Where this is going
- Bidirectional neuroprosthetics
- Movement + sensory feedback
05/ Architecture
A prosthesis should adapt to its user.
Every person differs in anatomy, muscle structure, proportions and movement patterns. So we treat the prosthesis as a modular system rather than one fixed product.
SENSOR MODULE
Signal acquisition placed against the residual limb. Electrode count and placement change per person.
ELECTRONICS
Amplification, filtering and the compute that turns raw biopotential into a decoded intent.
ACTUATION
Motors and transmission. Grip force and speed budgeted against battery and thermal limits.
STRUCTURAL CORE
The load path. Carries actuation forces into the socket without transferring them to tissue.
CUSTOM EXTERIOR
Fitted to the individual’s anatomy and proportions — the only layer most people ever see.
06/ Decoding
The system learns the person.
Instead of forcing every user to adapt to a fixed control scheme, we are building toward systems that adapt to the individual.
Surface EMG, sampled continuously.
Amplitude, timing and frequency descriptors per channel.
- CLOSE HAND97.8%
- OPEN HAND12.1%
- INDEX8.4%
HAND → CLOSE
Command issued to the actuation layer.
CONCEPTUAL DEMONSTRATIONThe values shown above are illustrative and animate on a fixed loop. They represent the structure of the decoding pipeline, not measured THALAMO performance. We have not published benchmark results.
Restoring the connection between human and machine.
We are a small team building this in Prague. If you work in prosthetics, neural interfaces, signal processing or robotics — or you would use a system like this — we want to hear from you.