Grant Packet
Luka Rekhviashvili, Giorgi Rurua


Funding Request Focus: Phase 2

This grant request is focused on Phase 2, where conceptual and single-channel groundwork is transformed into a multi-channel wearable system ready for structured user validation. Funding enables custom electronics fabrication, signal-grade materials procurement, participant testing logistics, and concentrated model benchmarking. The immediate objective is to produce a reliable eight-channel prototype and a rigorously documented dataset that supports reproducible decoding performance analysis.


Financial Justification and Budget Breakdown

As an independent, student-led project, the software architecture, high-VRAM computing infrastructure, and workspace are already secured and self-funded. The requested micro-grant will strictly cover the lean material costs required to build the signal-acquisition hardware. This includes the analog front-end integrated circuits, custom printed circuit board (PCB) fabrication, and the electrodes necessary to capture real-world SVR data.

Category Specific Item / Resource Estimated Cost (USD) Justification
Integrated Circuits Biopotential AFE (ADS1299) $55 Core analog front-end for high-precision, low-noise measurement of sub-vocal biopotentials.
Microcontroller Wireless MCU (ESP32 / nRF52) $25 Low-latency data transmission from the sensor board to the local compute workstation.
Fabrication Custom PCBs & Assembly $60 Fabrication and shipping for low-volume, custom flexible or rigid-flex circuit boards.
Sensors Surface Electrodes & Paste $45 Ag/AgCl electrodes and skin-prep materials for reliable, low-impedance signal acquisition.
Hardware Passives, Cabling & Connectors $50 Resistors, capacitors for hardware filtering, and shielded cables to minimize environmental noise.
Prototyping 3D Printing Filament / Casing $30 Materials for iterating on an ergonomic, wearable enclosure for the neck/jaw placement.
Logistics Shipping & Contingency $85 Buffer for international shipping costs, import duties, and minor component iterations.
Total Requested $350

Ethics, Privacy, and Safety

Amtavla treats biosignal data as sensitive human data. Participation is voluntary, consent-driven, and revocable at any time. All collected sessions are assigned randomized participant identifiers, and personally identifying metadata is stored separately from signal files. Access to raw recordings is restricted to authorized project members and controlled through local encrypted storage practices where available.

The system is designed as non-invasive and low-risk: dry-contact wearable electrodes with no penetration, low-voltage electronics, and continuous monitoring for skin irritation or discomfort during test sessions. Participants can pause or terminate sessions immediately without penalty.

Data governance emphasizes minimal retention and purpose limitation. Only data required for model training, validation, and methodological reporting is retained. Any publication output is aggregated and de-identified, with no release of private personal records.


Demo and Validation Artifacts

The grant period produces concrete artifacts for reviewers and technical auditors:

1) Multi-channel wearable prototype (hardware v1)

2) End-to-end capture-to-decoding software pipeline

3) Dataset protocol and anonymized benchmark splits

4) Quantitative report including baseline word error rate, latency, and session stability metrics

5) Demonstration package with images, architecture diagram, and short end-to-end test video

These outputs are structured to support independent review, replication, and phase-gate decisions for post-grant deployment work.

For reference, our Phase 1 demo, AI system, and software architecture are documented on the following pages: Methodology (Hardware) and Methodology (Software).


Expected Phase 2 Outcomes

By the end of Phase 2, Amtavla will have a tested eight-channel silent speech capture platform, a validated experimental protocol across multiple users, and a technical baseline that de-risks miniaturization and mobile deployment in Phase 3. This closes the gap between concept-stage R&D and practical, repeatable system performance.