ECE4006

From GanjaLinux

[[Category:ECE4006]]

Final Report: [[1]] This is Group 1's project development page for Dr. Barnwell's Real-Time DSP Senior Design course at Georgia Institute of Technology. Please read our Design Report and implementation report for more details.

Official Website

Next Meeting: none. Bunger-Henry 315.

Final Presentation: December 8, 3 pm.

Georgia Tech ECE 4006 Senior Design Hardware.

Contents 1 Minutes 2 Resources 2.1 Bitmap Info 2.2 Code 2.3 Hardware 2.4 Papers

Minutes

1. Meeting 1
   • introductions and backgrounds
   • potential projects
     • Professors and Research Interests:
       1. Array Processing:
          • Russell M. Mersereau - Image and Video Processing
          • Biing-Hwang (Fred) Juang - Speech processing
          • Douglas B. Williams - Statistical Signal Processing
       2. Speach Coding
          • Thomas P. Barnwell - DSP Hardware and Speech Processing
          • Mark A. Clements - Speech Recognition, Aids for Hearing-Impaired
          • Chin-Hui Lee - Speech and Language Processing
       3. GTREP
          • Dr. Joel Jackson
2. Meeting 2 (Lucas & Dr. Mersereau)
   • problem: eliminate camera man, automate classroom
   • blocks:
     1. acquisition (locate speaker, scan rate, red/green ratio)
     2. zoom (in on face, out on back of head)
     3. respond to gestures (hand movement controls slide progression)
   • algorithms:
     1. particle/condensation filter = difficult to achieve real-time
   • contacts:
     1. Yeong-Seon Lee (research student).
        • research: isolate current speaker in conference room using particle filter
   • equipment:
     1. polycom camera
     2. smart board
3. Meeting 3 (Overview Design and Planning)
   • Teams - Please read up on your area of expertise.
     1. FPGA Programming / Systems Integration  : Jaimin, Lucas, Vince
     2. Vision System  : Lucas, Vince
     3. Speech System  : Justin, Jaimin, Wailing
     4. Camera / Mic interface  : Wailing, Lucas, Vince
     5. GTREP contact and Data acquisition  : Wailing, Lucas
4. Meeting 4 (Dr. Barnwell, Lucas, Vince, Justin, Jaimin)
   • Project Proposal
     • incorporation of previous projects:
       • narrow band tracking (using DSK algorithm)
         • not applicable to our environment
       • video tracking projects were conducted using the GVU labs
     • overall: project is too difficult.
       • pick one component -> chose vision tracking
   • Meetings
     • have milestones for each week (updated weekly)
   • Problem Specification (suggested a second meeting with Dr. Mersereau)
     • still camera tracks single moving object (teacher head), puts a box around it, and directs a second camera
     • moving camera locates the head and tracks
     • potential to merge this algorithm with HP's research of using many low resolution cameras to make one high-resolution picture (from this picture we track, using downsample techniques to achieve real-time)
   • Test Environment
     • Room VL461
5. Meeting 5
   • Agenda
   1. Discuss abstract
   2. Obtain Polycom documentation
      • how to control PTZ
      • video output formats (S-video or composite)
   3. Select a two input real time capture card (expensive ~1000 USD! see Matrox, Pinnacle, Canopus and Dazzle for examples)
      • depends on Polycom output
      • physical separation of cameras will require very long video feed cable that will cause issues (degradation of video quality and ability to track)
      • firewire or PCI
      • call Videoguys - The Electronic Mailbox - 800-323-2325 We are the Digital Video Editing & DVD Production Experts! and get a price quote
   4. order card
   5. order IR RS232
      • just in case we use it to control the PTZ of the cameras because we can't figure out how to control the Polycoms over the IP network
6. Meeting 6
   • Mariam Crowder
   1. Digital Media Lab - VL 483
   2. Set up meeting to obtain possible data
   • Equipment Available
   1. Sony EVI-D30 Camera
      • Stepping motor camera
      • Need to find PCI card that interact with camera
   2. Video Capture Card
      • Winnov website http://www.winnov.de
      • Model - Videum 1000 VO Plus http://www.winnov.de/index.php?option=com_content&task=view&id=61&Itemid=93
      1. Driver download not required
   3. HP Vectra COC 311
      • Code Composer Installed
      • Digital Signal Processor
      1. TMS320C6701
         • Frequency - 167 MHz
         • 4ch DMA
         • 2 32-bit GP Timers
         • 1.9 V Core Supply, 3.3 V IO Supply
         • Floating-Point Digital Signal Processor

Resources

Bitmap Info

• 1. Intro to bitmap
  2. sample read code
  3. GD Graphics Tool for C

Code

• Image Processing Code
  1. Open Source Computer Vision Library
  2. Mathtools.net Image Processing for MATLAB
  3. Reference Introduction (Image Processing Toolbox)
  4. The MathWorks - MATLAB Central - File Exchange
  5. Some guy at MIT's spatial clustering alg
  6. Code Composer Resources
  7. Real-Time Digital Video Transfer via High-Speed RTDX

• Matlab: Current Files
  1. Main Method
  2. find_head
  3. filter_skin
  4. Bounding Box Function

• Matlab: Old Files
  1. version 1
  2. demo_01
  3. optical_flow_function
  4. Frame Grabber
  5. find_clusters
  6. centroid
  7. test

• C
  1. find_head.c
  2. bitmap24.h
  3. Working find_head (DO NOT EDIT)
  4. Host-side code
  5. target-side code

• C old
  1. main.c
  2. Bitmap.c
  3. Binary.c
  4. Header file
  5. Skin Tone Filter
  6. clusters
  7. Image Access

Hardware

• TI DSP

• EVI-D30
  • Command List
  • Operating Instructions
  • Manual
  • Control Setup

Papers

IEEE Xplore can be accessed from anywhere with proper login via http://gtel.gatech.edu:2065/Xplore/dynhome.jsp

• Speaker Tracking
  1. A low-cost real-time stereo vision system for looking at people
  2. Audio-visual speaker tracking with importance particle filters
  3. Real-time speaker tracking using particle filter sensor fusion
  4. Real-time speaker localization and speech separation by audio-visual integration
  5. Evolutive HMM for multi-speaker tracking system
  6. A self-calibrated speaker tracking system using both audio and video data
  7. The Intelligent Classroom
  8. Robust real time tracking of a vehicle by image processing
  9. 3-D facial pose and gaze point estimation using a robust real-time tracking paradigm
  10. Transductive inference for color-based particle filter tracking
  11. Robust real-time tracking on an active vision head