Docstoc

Slide 1 - The School of Engineering Science

Document Sample
Slide 1 - The School of Engineering Science Powered By Docstoc
					Sniper Localization System
         Marko Gasic
        Sandeep Brar
      Ehsan Dallalzadeh
         Balraj Mattu
Overview
   Introduction
   Vision
   System Description
   Test Results
   Obstacles Encountered
   Project Finances
   Production Cost
   Conclusion
   Questions?
Introduction
 Snipers are a serious threat in urban warfare
  environment.
 Civilian threat in cases such as Washington DC
  sniper.
 Snipers are very effective at harassing and
  impeding military operations.
 AcousticShield Designs system enables
  identification of direction of origin of a sniper
  shot within seconds of the event.
System Overview
Vision
 Existing Products
    • Above $15,000 US
    • Available only to elite military divisions and
      not standard equipment to regular units or
      police forces
 Acoustic Shield System
    • System cost around $2000
    • Low cost enables local police departments
      and regular military units to purchase
      system
System Description
Principle of Operation
   • Sound waves reach 4 speakers at different times
   • Using these delays we can calculate the origin of
     sound


 Functional Breakdown
  •   Signal Acquisition
  •   Gunshot Recognition
  •   Delay Detection
  •   3-D Triangulation
  •   Human-Machine Interface (H.M.I)
Sound Acquisition
PC Hardware
  •   M-Audio Delta 44 PCI audio card
  •   4/4 mono analog input/output channels
  •   24bit, 8kHz – 96kHz independent channel sampling
  •   Winsound interface drivers
Sound Acquisition
 Microphones
  •   Electret Omni-directional condenser microphones.
  •   -45dB sensitivity
  •   20Hz – 16kHz Frequency Response
  •   60 dB S/N ratio
Sound Acquisition
 Microphone Preamplifier
  •   Supplies minimum voltage required for microphone
      operation
  •   Amplifies signal to 500mV swing, compatible for PC
      soundcard input.
Sound Acquisition
 Software



         Sample at   Continuously sample
          44kHz      microphone inputs


                     When sample exceeds
                     0.2V, record next 1.0
                     seconds and place in
                     memory
Recognition Algorithm
 Understanding the characteristic of a gun shot




                 Time Domain Representation
Recognition Algorithm

          Frequency Domain Representation
Recognition Algorithm
 Algorithm is based on comparison of
  average power between two bins:

              Average Power



                                                   Bin1 average power
                                              11   Bin2 average power
                              Average Power




  228 Hz (±150 Hz)                            1 kHz – 1.5 kHz
Recognition Algorithm
 Refinement after experimentation
   Needed to consider all 4 input at the
    same time
           Back




   Microphone


                  Sound Wave




                Distortion in frequency spectrum is introduced
Recognition Algorithm
 Simple Solution

 Analyze all four microphones

 Accuracy is demonstrated in Test
  Results section
Δt Extraction
 4 similar signals, out of phase
 Use Cross Correlation to determine phase difference
                       Δt1
                       4
                          Δt1    Δt1
                         2       3
3-D Triangulation
 Extrapolate origin of sound using the 3 Δt’s and speed of
  sound as input
                        Use Gauss-Newton method to
                           solve 4 non linear equations
                          x  xa 2   y  y a 2 z  z a 2  a 2
                          x  xb 2   y  yb 2 z  z a 2  a  ct ab 2
                          x  xc 2   y  yc 2 z  z c 2  a  ct ac 2
                          x  xd 2   y  y d 2 z  z d 2  a  ct ad 2
                        Recover the X Y and Z
                         coordinates of signal origin
                        Normalize vector to give
                         azimuth and elevation angles
User Interface
   Easy to Use/Navigate
   Targeted towards Army Personnel
   Displays Azimuth and Elevation
   No installation Required
Testing
 The testing was done in 2 phases:
      •   Testing for the detection in 2-D (X,Y)
      •   Testing for detection of the elevation



 Procedure A:
      •   The system was setup
      •   The software was running
      •   Located the tripod at the center of a large circle
      •   Drew a 2-D coordinate system about the center of the
          tripod
      •   Marked the imaginary circle around the center of the tripod
          with points each about 30 degrees apart
      •   Ran the sound sample of the gunshot twice at each point
Testing
•   Recorded the Average, Trigger, X and Y values
•   Took a string from the sound source(speaker) to the
    center of the tripod
•   Chose a point on the string and recorded its X and Y
    components.
•   At the end, had pairs of vectors in 2-D
• Comparison Stage……
•   Wrote a C++ code to input each pair of vectors to
    calculate the angle between the actual vector and the
    result vector from the system in Degrees
Observations
 On average, the angle difference was about
  2.78 Degrees

 The accuracy was almost the same for all the
  points in the surrounding
Testing cntd.
 Procedure B (Elevation):
      •   The system was setup
      •   The software was running
      •   Located the tripod at the center of a large circle
      •   From points 90 Degrees apart, got samples
      •   At each point, tried 3 different elevations:
            1) above the center plane
            2) at the same plane
            3) below the center plane
      •   Recorded the elevation that the program gave for each
          trial
      •   For each point, measured the elevation angle compared
          to the center of the tripod (+ if above the center, (-) if
          below the center)
Observations
 On average, elevation difference was 3.15
  Degrees

 Functional Specifications stated maximum
  allowable error of 10 degrees
Obstacles Encountered
 Initially used Texas Instruments DSP
  •   Insufficient inputs: unable to sample both stereo
      codecs simultaneously
  •   Insufficient resolution: TMSC320 C6711 main audio
      codec samples at only 11kHz, we need a minimum of
      44kHz
  •   Extremely poor user interface and non-existent (yet
      advertised) compatibility with MATLAB
Obstacles Encountered
 Initial Algorithm               Input Signal


       Divide sampled input into smaller intervals
       Analyze smaller intervals in frequency domain




      Power Spectrum of t0 – t1          Power Spectrum of t2 – t3
Obstacles Encounterd
 Initial Algorithm
             Determine if it’s gun shot or not by comparing
              with known spectrum




                              Positive
                               Match


  Power Spectrum of t0 – t1               Known Spectrum
Obstacles Encountered
 Algorithm was implemented in Matlab
  and Simulink




              Main Recognition Block
             Recognition Subsystem 1
Obstacle Encountered
 Problems
     Unable to achieve desired speed
     Didn’t do well when tried with real input
      (instead of a wave file)
Financial Aspects
 Prototype Development Cost:

  •   TMSC320 Daughter Board       $120.00
  •   MATLAB RTW Documentation     $ 35.00
  •   Microphones and Pre-Amps     $ 60.00
  •   Miscellaneous Audio Cables   $ 30.00
  •   M-Audio Delta44              $220.00
  •   Other                        $ 25.00

  •   TOTAL                        $490.00
Budget Estimate
 Initial cost estimate $2260.00
 Actual cost $490.00
 Significantly lower cost due to change in
  platform
 Savings with no loss in performance
 We were able to borrow the tripod, saving
  ~$100
Manufacturing Costs
 Assuming 50 units/month
 Based on Digi-Key bulk pricing where available

ITEM            COST
Microphone      $   1.20
Pre-Amp         $   6.00
Tripod          $ 80.00
M-Audio Card    $ 130.00
Cables          $ 20.00
PC              $ 600.00
Total           $ 837.20
Conclusion
 Successfully Demonstrated Functional
  Concept
 Demonstrated market value and
  ability to produce at reduced cost
 Encountered problems and chose
  alternate solutions
 Stayed within budget and timeline
  considerations
Thank You




            Questions?

				
DOCUMENT INFO