EEG topographic brain activity map showing green-teal motor cortex activation zones on a dark background

BCI signal decoding SDK

Reliable motor-imagery decoding.
Every electrode, every session.

Synaptiq converts surface EEG into accurate device commands for rehabilitation exoskeletons — handling electrode drift and inter-session non-stationarity that makes raw BCI data unusable in clinical workflows.

Request SDK Access View Signal Pipeline
<100ms
End-to-end decode latency
~94%
Motor-imagery classification accuracy
Adaptive
Drift correction across sessions
IEC 62304
Software lifecycle development process

The core problem

Raw EEG data is unusable.
We fix that.

Electrode impedance rises across a two-hour therapy session. Gel dries, cap shifts, scalp conductance changes. The spatial filters your decoder trained on Monday morning are stale by Friday afternoon — and a static classifier has no way to compensate.

Inter-session non-stationarity is the fundamental obstacle to clinical BCI deployment. EEG covariance structure shifts between sessions by more than any within-session calibration can cover. Device manufacturers know it. Clinicians feel it. No one has solved it cleanly for a production SDK.

Synaptiq solves it with Euclidean alignment and online Riemannian covariance updating — re-centering each session's feature distribution before classification, so session 5 performs like session 1.

Session accuracy — with vs. without alignment

Signal processing pipeline

How the pipeline works

01
Raw EEG Input
Multi-channel surface EEG from 8–32 electrode cap. 250–1000Hz sampling. LSL, BrainVision, BDF/GDF compatible.
02
Preprocessing & Artifact Rejection
Band-pass 8–30Hz (mu/beta), EOG/EMG artifact rejection, CSP spatial filter bank computation.
03
Riemannian Geometry Classification
Covariance matrices on SPD manifold. Minimum distance to Riemannian mean. Euclidean alignment for session normalization.
04
Device Command Output
Discrete class labels mapped to device commands. Digital I/O, serial UART, UDP. <100ms total pipeline.
Explore the full pipeline

Hardware compatibility

Designed for rehabilitation devices

Rehabilitation Exoskeletons

Lower-limb gait rehab, upper-limb reach-grasp, and full-body robotic orthoses. SDK outputs discrete command events over digital I/O, serial, or UDP.

See compatibility

Clinical EEG Headsets

8-channel to 32-channel dry and gel electrode caps at 250–1000Hz. BrainVision, Lab Streaming Layer (LSL), and BDF/GDF format support.

See compatibility

Research BCI Systems

Offline batch decoding for research protocols. Python SDK integrates with MNE-Python pipelines. REST API for cloud batch analysis.

View SDK

Clinical validation

Clinical validation summary

93.7%
Session-to-session accuracy retention with Euclidean alignment (vs. 72.4% without)
±18μV
Electrode drift compensation range — full therapy-session drift window
68ms
Typical real-time command latency floor at 250Hz acquisition

Development follows the IEC 62304 software lifecycle process. Designed for FDA 510(k) regulatory pathway as a software accessory to Class II rehabilitation devices.

View methodology and validation details
View full evidence

Start integrating

Start integrating Synaptiq today.

Python and C++ SDKs with a REST API for cloud batch decoding. Three lines to decode your first motor-imagery event.

Request SDK Access
decode_example.py 
import synaptiq

# Initialize and load session calibration
session = synaptiq.Session(device="lsl", channels=16)
session.calibrate(duration_s=120)

# Start real-time motor-imagery decode
def on_command(cmd):
    print(f"Command: {cmd.label} [{cmd.confidence:.2f}]")

session.decode_stream(callback=on_command)