This document just summarises my work and links to the detailed description of works is marked via underlined text.
Information theory
- The 12 Balls problem with unknown defect — I explored this problem from perspective of information gained after each weighing and came up with a new solution.
- Receiver Synchronization in absence of phase locked loop- When I was creating Automatic speaker volume control for Samsung I ran into the synchronisation problem at receiver and it took quite some time via cross correlation to detect inaudible signals generated by phone on receiving phone calls. I solved this problem by using De Bruijn Sequences.
Laser Microphones
3. While at Samsung I once had come across a problem to detect very low frequency changes and no of the shelf commercial microphone could detect it due to frequency response characteristics. I remembered once I had seen a lecture from Royal Institution where they used a laser for covertly picking up sound from a room. I made the same setup and was able to record fizz from a drink via laser microphone
Work in improving Accessibility
4. Blind Navigation — At Samsung Research Institute Delhi, I designed and implemented a prototype for blind navigation by using the concept of spatial audio and rendering the surroundings and navigational cues in 2D sound setting allowing the person to see the map by audio rather than plain old turn by turn navigation. I developed it on MacOS using panda3d. The user perceived sound via microphones. This I developed at Samsung no public code is available.
5. improvements to low cost bidirectional Deaf-Blind Communication
I have developed 2 improvements to current method of deaf blind communication using morse code. Morse code has been used from time of Helen Keller for communicating with deaf blind but it has a serious flaw which does not allow it to be readily adapted to modern day devices like phones and laptops. I propose a change that will make morse code uniquely decodable and how to interface with mobile phones that will give tactile input capability.
github- https://github.com/afopriest/Deaf-Blind-Communication-Interface
Computational imaging and its applications
6. Video Deep Fake detection based on extracting Heart Rate from changes in face color.
This was developed to try and detect heart rates from facial video by trying to extract PPG signals in video’s RGB channels.
8.
Failed or currently unfeasible projects on a Back Burner for now
1. Application of laser Microphones to counter muffled speech in Covid-19 —
We all have experienced how our voice gets muffled after wearing face mask.
I tried to pick up audio by shining laser on to larynx and capturing it via phototransistor but I was not able to make it work correctly. The incidence angle was always changing as surfaces kept deforming and photransistor’s frequency Response was not suited for laser and limit in fabrication methods resulted in a failed project.
