Wireless Signals and its Applications

What is the class about?
In short, this class will focus on radio signals and some of its practical applications. We will cover a set of tools, techniques and algorithms relevant to understanding signals and how we can gainfully apply them to solve some interesting, real-world, often future-facing problems like indoor localization, activity/gesture recognition in smart homes. Most techniques involve a linear algebra/statistical/machine learning component (e.g., Fourier analysis, Maximum Likelihoods, Naive Bayes, Expectation maximization, etc.) that will be taught from scratch in the class. However, we don't stop there: we will validate, investigate most techniques by building relevant applications through programming assignments and projects. We will primarily stick to using real-world signal data that will bridge the gap between theory and practice. The course will primarily be project-driven and you should aim at delivering a demo-able system, technique, mobile application at the end of the term.

What is the class not about?
Historically this class was geared towards networking as opposed to mobile computing applications. However, this fall we will be looking towards mobile computing applications and less of networking fundamentals. For example we will not cover wireless link layer or routing protocols or wireless TCP. We will not be looking at cellular standards like 3G or 4G-LTE.

Pre-requisites
  • Introductory probability and statistics. These are needed for modeling exercises. 
  • Mobile programming (Android). You needn't be a pro. Learn on the go.
  • General familiarity with Unix/Linux systems. Please get hold of a Linux system in the beginning of the term.
Text
  • No text is needed. Reading materials will be posted. 

Topics

(Details will be posted in the schedule/readings page)
  • Introductory concepts on wireless signals
    • Concepts of Signals, Noise and Interference
    • Signals as an information carrier
    • Basic probability tools needed to model signals
  • Measuring signals
    • Signal detection algorithms
      • Fourier Transform, FFT
      • Hardware (Software Defined Radios) and software tools 
    • Fundamentals of measurement error
    • Modeling signals 
      • Common existing models (e.g., log-normal)
      • Hard to model. Challenges?
      • Data-driven models. Crowd-sourcing signal measurements
      • Radio Environment Maps
    • Application: Finding unused spectrum
      • Futuristic communication systems will share wireless spectrum (a.k.a., whitespaces) and they need to know spectrum opportunities at a finer spatio-temporal granularity.
      • Measurement-augmented spectrum databases (Stony Brook)
  • Radio-signal based Localization
    • Ranging based techniques.
      • Technique
      • Challenges. Poorer performance.
    • Radio-fingerprinting based techniques
    • Indoor localization 
      • Unsupervised Indoor Localization ( WIGEM from Stony Brook, Unloc from Duke, ZEE from Microsoft Research India)
    • Indoor mapping
    • Brainstorm project ideas (office hours, piazza)
  • Radio-signal based Activity Recognition (mostly explored in projects)
    • Understanding WiFi CSI data
      • Cleaning and processing data
      • Bayesian Filtering, Kalman Filters
      • Dynamic Time Warping
      • Hilbert Transforms, Wavelet Transforms
    • Applications
      • Detecting presence/absence of human body using WiFi signals
      • Detecting simple hand gestures
      • Seeing through the walls! (X-Ray vision UCSB)
    • Brainstoriming project ideas (office hours, piazza)
  • Quality of Experience in mobile applications
    • How 'poor-signal' affetcs our application usage experiene.
      • Poor signal and how it relates to bandwidth and packet loss
    • Impact of network latency 
    • Quality of Experience
      • Page load time - Web browsing
      • Streaming video. Buffering, Stalling, Frame-drops
      • VoIP, Jitter
    • Relating Mobile Quality of Experience to network performance (QoS)
    • Brainstorm project ideas (office hours, piazza)

Grading

(Subject to change)
  • Programming assignments/Reviews (25%),
  • Midterm exam (25%),
  • Term project – in groups of 2 students (50%). 
If significant work is done in the project – specifically developing a useful tool or measurement analysis, you can choose to shift more weight to the project.



Subpages (1): Untitled