Sukhraj Singh

Research

Two papers.
A closer look.

Questions explored through computer vision and voice AI. Open a paper, turn the pages, or read the full PDF.

Computer vision · Independent research

Temporal-Consensus OCR

Can neighbouring video frames help read a headline more accurately? This paper explores combining evidence over time to improve text recognition.

Evaluated on 50 synthetic clips; real-world broadcast performance remains unverified.

Voice AI · Co-authored course research

SCAMPA / ScamShield

A proposal for a local voice agent that connects speech recognition, language models and speech synthesis for controlled, defensive scam research.

Phase 1 architecture proposal in an IEEE-formatted course report; not a deployed voice system.