Yan Michalevsky


Academic, professional and personal information.

Projects

Backscatter Channel Characterization using Geometric Methods

Can we turn our smartphones into multi-purpose scanning devices, that can detect tumors, locate objects and ’see’ beyond walls? Radio waves get reflected by the environment arrive back at the transmitter, where the nature of each reflected wave depends on the reflecting surface. This suggests the possibility of learning the physical environment by detecting properties of backscatters.

Project for EE359, a Stanford University course on wireless communication.

Towards (even more) practical Faust: Polyphony, Portamento and Pitch-bend for Faust VSTi-s

Faust is a musical signal processing language. The Faust compiler and the toolset provided along with it enable generating standalone syn- thesizers and plug-ins for various architectures. We noticed that while being a very useful tool for sound synthesis its VSTi plug-ins lack several critical features for practical usage in combination with music production software and digital audio workstations (DAW). We focus on the VST architecture as one that has been used traditionally and is supported by many tools and add several important features: polyphony, note history and pitch-bend support. The contribution of polyphony is a step towards making Faust a practical tool for real-wold music production.
Taking this project a step further we presented an extended paper at the Linux Audio Conference 2014.

Final project for MUS420B Stanford University course.

Tight Schedule: Deadline Constrained Scheduling of OpenRadio on a Multi-Core Platform

One of the challenges of real-time, performance critical multi-core systems is the efficient scheduling of executed tasks. The scheduling problem consists of assigning the tasks to the different cores and deciding upon the order of execution. A special case is heterogeneous multi-core platforms where the cost of execution varies among the different processors. In this paper we present static and dynamic scheduling approaches, discuss their pros and cons and demonstrate a dynamic deadline-oriented scheduling algorithm with a low processing footprint. We apply the two approaches to the OpenRadio real-time Wi-Fi platform operating at high rates and demonstrate obtaining of feasible schedules. Using our scheduling algorithm we examine implications of hardware parameters on wi-fi processing feasibility. We also propose several possible improvements to the dynamic scheduling algorithm.

Project presentation

Final project for CS316 Stanford University course.

Speech-Singing Discrimination

For a while I was working on the problem of speech-singing discrimination using DSP and Geometrical Methods. The results of this work served later on as a basis for the work on Speaker Identification using Diffusion Maps.

  • Speech-Singing Discrimination using Geometric Methods - project summary
  • Application of Localized Diffusion Folders (LDF) to Speech-Singing Discrimination
  • Speech and Singing Discrimination using Diffusion Maps
  • The Diffusion Maps implementation written for this project and later on used in the Speaker Identification paper can be checked out from https://bitbucket.org/ymcrcat/diffusion-maps.git.

    Project done as part of Special Topics for Excellent Students in the Technion.

    Classical Analysis of Cavity Optomechanics

    Kazabooboo!