Group 1 Fire Tracker

1
Group 1Fire Tracker

• Group Members
• Kevin Moser Algorithm Development
• Aric Repstien Peripheral Devices Data
  Analysis
• Daniel Theodoseau DSP Development
• Sponsored by EMTEL
• Project Advisor Dr. L. Jones

2
Project Description

• Fire Tracker is a gun shot detection and
  location system based upon the audio signatures
  of various firearms.
• The device will
• Detect a shot being fired
• Identify the weapon type
• Relay data to a computer system for further data
  analysis, such as event triangulation

3
Goals

• Proof of Concept
• Detect an Event
• Determine a weapon type
• Reject similar non-events (firecracker, car
  horn,..)
• Global Positioning System (GPS)
• Data Interface (host application)

4
Accomplishments

• Collect Data
• Detect an Event
• Perform Wavelet Transform
• Perform Event Analysis
• Global Positioning System (GPS)
• Data Interface (host application)

5
Design

• Phase 1
• Data Acquisition
• Data Analysis
• Simulation
• Phase 2
• Hardware/Software Prototyping
• On DSP

6
Phase 1

• Data Collection
• Data Analysis
• Wavelet Transform
• Event Detection
• Statistical Analysis
• Simulation

7
Data Collection

• Weapons Used in Test Set
• 1) ArmaLite Inc., Model AR-15, 223 (Rifle)
• PMC Target Ammo, 223 Rem, 55 GR., FMJ-BT
• 2) Beretta, Model 92FS, 9mm Luger (Handgun)
• Speer Lawman Ammunition, 9mm Luger, 155 GR., FMJ
• 3) Remington, Model 870 Express,12 Gauge
  (Shotgun)
• Estate Cartridge Inc., 12 Gauge, 2 3/4 L, 3 1/2
  Dr. Eq., 1 oz. Shot, 7 1/2 shot

8
Frequency Content

• Fourier Transform
• Used to determine the frequency content
• of the gunshot samples and to plot the
• power spectrum.

9
Frequency Comparison
Green Shotgun Blue Handgun Red Rifle
10
SamplingRate

• Based on the range of frequencies from 0 - 4 kHz,
  our sampling rate was determined from the
  following formula
• Therefore our sampling rate, FS 8 kHz.

11
Wavelet

• Wavelet Transform
• Time Variant Signals
• Analyzes a small section at a time using a window
• Varies the Window Size
• Higher Accuracy

12
WaveletTransform

• Consists of shifting and scaling (stretching) the
  analyzing window through the full domain of the
  signal.
• The results of this transformation are the
  wavelet coefficients.
• Wavelet coefficients represent how closely
  correlated the wavelet is with the windowed
  section of the signal.

13
Analyzing Window

• Daubechies Family db4
• Low-scale, which allows the measurement of
  rapidly changing details.
• Analysis will be orthogonal.
• Used as a Finite Impulse Response filter (FIR).
  The filter coefficients are real.
• Filter coefficients determined using Lagrange a
  trous filter.

14
Function Shapes
15
FilterGeneration

• Step 1
• Compute a Lagrange a trous filter, P. This
  filter is symmetric.
• Pa(N) 0 a(N-1) 00 a(1) 1 a(1) 0 a(2) 00
  a(n)

16
FilterGeneration

• Step 2
• Determine the Daubechies filter, W. This is
  known as the scaling filter.
• W is a minimum phase solution of the Lagrange a
  trous filter, P, based on the roots of P.

17
FilterGeneration

• Step 3
• Compute the filters associated with the scaling
  filter, W. These are known as the decomposition
  filters.
• LD Low-Pass Filter
• HD High-Pass Filter
• LD wrev(W)
• HD quadrature mirror filter of LD

18
DecompositionStructure
19
Detail of Decomposition
20
Event Detection

• Locates Rapid changes
• Running ambient sound level
• Current sample mean ambient level
• Standard Deviation current sample

If Current average Running average Event
Tolerance and if Standard Deviation
is contained in the sample, with an origin at the
Maximum Sample value
Note Event tolerance used is 5
21
Event Detection

• Improving Detection Rate
• Find Event Slope
• Verify Maximum

22
Statistical Analysis

• Wavelet db 4
• Decomposition level 4
• Adjacent Sub bands 1 and 2
• Compare data to Ideal
• Sum of Difference Squared

23
Statistical Analysis

• Converting data to a Readable form
• Maximum accepted sum of differences
• Ideal sum of differences
• 0 100
• 100 Maximum Acceptable value
• 100 Rejected

24
Simulation

• MatLab
• Reading Wav files
• Loop Back
• Un-buffered Capture
• Gain
• Processing live audio
• Un-buffered Capture
• Gain

25
Hardware
26
Phase I
27
Phase II
28
Phase 2

• Microphone and Low-Pass Filter
• Computer
• DSP Evaluation Board
• GPS Evaluation Board

29
Microphone

• Panasonic Series WM-61B
• 20hz to 20khz
• Omni-Directional
• 2v operating voltage
• Low Signal to Noise Ratio
• Low Cost
• Small Package Foot Print

30
Filter

• 20hz to 4khz

31
Why DSP?

• Designed for signal processing applications.
• Ability to do real-time signal processing.
• Circular Hardware Buffer
• Optimize Branching Instructions
• Single Cycle Fetch Operations

32
DSP

• Texas Instruments DSP (TMS320C6711)
• 900 mflops
• 4kb program cache
• 4kb Data Cache
• 64kb Mapped Memory
• Multi-channel Buffered serial port
• Availability of Pre-assembled Evaluation Units

33
DSP

• TMDS320006711 C6711 DSP Starter kit
• Development environment
• Interface for data and program loading.
• Interface for Monitoring Program Execution.
• Onboard A/D and D/A
• 16mb SDRAM
• Power is already supplied

34
DSP Layout
Provided by Texas Instruments
35
GPS

• Motorola M12 Evaluation Kit
• 12 Channel
• Positional accuracy1-5 meter
• Timing Accuracy 500ns
• Serial Data Communication
• 3v operating voltage

36
Software
37
Software Architecture
38
Program Flow
39
GPS Initialization
40
Event Detection
41
Time Position Acquisition
42
Event Analysis
43
Data Transmit
44
Budget
45
Schedule