Hisham Bedri

December 1, 2014

Projects + Passions

Filed under: — hbedri @ 5:32 am

Seeing Around Corners with a Mobile Phone? Synthetic Aperture Audio Imaging:

audio_teaser Seeing around corners, in the dark, and through smoke is difficult without specialized sensors[Velten et al. 2012], and so far impossible with a mobile phone. We use an active audio system to sense objects around occluders. Current techniques perform passive localization of sound sources with a microphone array, however, we demonstrate that with one microphone and one speaker pair, such as the ones found in mobile phones, it is possible to sense the specular
reflection of silent objects such as mannequins around occluding objects. We demonstrate this technique by sensing a mannequin occluded by a wall.
Siggraph 2014 Poster, Student Research Competition, PDF: Hisham Bedri, Micha Feigin, Michael Everett, Ivan Filho, Gregory L. Charvat, and Ramesh Raskar. 2014. Seeing around corners with a mobile phone?: synthetic aperture audio imaging. In ACM SIGGRAPH 2014 Posters (SIGGRAPH ’14). ACM, New York, NY, USA, , Article 84 , 1 pages. DOI=10.1145/2614217.2614282 http://doi.acm.org/10.1145/2614217.2614282

Grid-Free time domain super-resolution for audio imaging:

superresolution The resolution of audio ranging and imaging is heavily limited by bandwidth. Here is an attempt to increase the resolution through continuous basis pursuit, a grid-less sparse algorithm. Comparison of backprojection, basis-pursuit denoising, and continuous basis pursuit shown.

PDF: Grid-Free time domain super-resolution for audio imaging: , Matlab Code: Code (readme coming soon)

Wrist-Wave: Context-Aware Clothing that listens to its environment:

wrist When you hear music, your foot starts tapping. What if your clothes interacted with the music around you! The wrist-wave is a battery operated wearable LED display which changes patterns according to the music around you! Great for dance clubs! The next version will have motion control, any motion correlation in this video is a coincidence.

Youtube: Wrist-Wave:

Orthographic Tensor Display Implementation:

wrist This is an orthographic implementation of the tensor-display paper. Lanman et al show that layering two LCD screens with content adaptive patterns can create multiview displays. Wetzstein et al extend this work to N-screens and directional backlights. This is an orthographic implementation of their paper, using iterative non-negative matrix factorization to solve for the appropriate patterns for 3 layered screens.

Matlab Code: Orthographic NMF Tensor .m Original Paper (Wetzstein et al): Tensor Displays Other awesome links (Matt Hirsch + Doug Lanman): Build Your Own 3D

Smoke Tomography for a floating 3d display:

floating By spinning a DLP projector and exploiting the >4000Hz refresh rate, it’s possible to create images that float in mid-air by projecting onto a scattering medium from many directions. Here a simulation and concept for a prototype is shown. There’s still much work to be done in optical alignment and synchronization before prototype works.

Youtube link: Work in Progress: Smoke Tomography for a floating 3d display:

Hearing the shape of a room: GUI/Simulator/Forward Model

roomGUI Can you hear the shape of a room? Here is a matlab based simulator for exploring the problem, capable of simulating convex rooms with arbitrary numbers of microphones / speakers. The number of bounces is controllable, and occlusion and multiple reflections are automatically handled.
Matlab code: Room echo simulator

Grow! Ultra-fast Multiview Stereo:

floating floating Goesselle et al show in “Multiview Stereo Revisited” show an accurate algorithm for recovering 3d shape from a set of photos, however their implementation is exhaustive and slow. By assuming the structure of the object being scanned is continuous, it’s possible to seed a single point and grow the object. This method is efficient and yields comparable results. This work was done with Aziz Alghunaim for 6.869: Advances in Computer Vision (Fall 2014).

Code coming soon!

Theoretical paper for transparent display using Coupled Total Internal Reflection / LCD panel :

CTIR A design for a high-resolution solid-state volumetric display using liquid crystals in their nematic phase is explored. The display is created by stacking transparent 2D displays. Each display is illuminated from the side and operates by coupling out a totally internally reflected wave. Analysis shows that this display is possible, however highly inefficient with a coupling efficiency of 5% in some cases and suffers from slow refresh rates due to time-multiplexing of colors. The exploration shows that nematic crystals can be used for other electro-optic applications such as multi-view displays and dynamic lenses.

PDF: Coupled Total Internal Reflection

Immersion Project, Make your laptop into a virtual-window through face-tracking + google streetview :

immersion Imagine diving into another world without ever having to leave your desk. The immersion project turns your computer into a window to another world using head-coupled-perspective. You can now interact with google streetview panoramas by moving your head. This website uses the newest features of HTML5 and javascript head-tracking libraries to provide you an amazing experience using just your webcam for input.

Demo (requires Web-RTC): Demo Youtube Video: Youtube

Designs for Directional Backlight:

backlight By projecting onto a partial diffuser, it’s possible to light-up a semi-transparent slab and make a semi-transparent display. An interesting effect that one sees is that areas with concentrated light perceptually seem less transparent (background washed out). This can be used to make simple transparent displays, or magical window effects without projecting directly onto the surface (and having a bright projector in the background).

Acoustic Landmine Detector (Senior Design Project, Tufts University):

landmine Buried landmines can be hard to detect, especially if they have been buried for a long time. When vibrated. however, landmines interact dynamically with the soil around them. The detector works by transmitting two sound tones (two frequencies) and records the sound that bounces back. The detector analyzes the characteristic response of the landmine-dirt system and then determines whether a landmine is buried.
Two frequencies are transmitted into the ground. The detection algorithm detects the non-linear combination of these two frequencies. These non-linear (non-harmonic) peaks in the spectrum of the response is caused by the landmine bouncing against its environment.

Music Video: Youtube

PillowPal (HackMIT Finalist, 2012):

pillowpal Ever had trouble waking up in the morning? Ever unconsciously turned off your own alarms? Never have this problem again with PillowPal! Your friends now have complete control over when you wake up. Isn’t that great? PillowPal allows your facebook friends to access your pillow online and turn on lights and music to wake you up in the morning. PillowPal is an attachment for your phone. It communicates with your phone through the audio port, so it works with almost every phone. The brains of the pillow are a picaxe microcontroller and a set of 32 RGB addressable LED’s which will blink different colors when its time to wake up. Work was done with Jesika Haria, Prabhav Jain, and Geza Kovacs.

Presentation: Youtube

There’s no app for that: Sudan sanctions are holding back innovation:

sudan In Sudan there is no app store. No Google play store. No credit cards. No Adobe Flash. You will be lucky to get your hands on Google Chrome web-browser, but even if you do, you will have to go to great efforts to circumvent blocks that prevent Google hangouts (video chat) from working. The content just “isn’t available in your country”. Due to US sanctions, the export of technologies from a US- firm or person to Sudan is prohibited, and the app-store is just the tip of the ice-berg…

Full article:

No Comments »

No comments yet.

RSS feed for comments on this post. TrackBack URL

Leave a comment

Powered by WordPress