Hand Gesture Remote Control

1
Hand Gesture Remote Control

• Justin Johnson, Joe Pommier, Mike Wang

2
Overview

• A consumer device that just works
• Consists of a single box that can be plugged in,
  easily configured for your television, and will
  work reliably indefinitely
• Works with already functional infrared remote
  control receivers built into most television sets
• Uses infrared light to track movements

3
Camera Module

• Encompasses camera, clock generation, and logic
  level translation circuitry
• Camera is a Pixart infrared sensitive digital
  camera with onboard digital signal processing

4
Camera Module

• Use a crystal oscillator and double inverters to
  create 25MHz clock
• Using an LTC4301 I2C bus buffer chip, perform
  logic level translations between 5V on the
  Arduino side and 3.3V on the camera side
• With properly written software, communication
  with the camera was successful

5
Camera Module Testing and Construction

• Laid out circuit in Eagle and ordered a PCB
• About to solder on the camera and realized the
  pins were reversed
• Had to re-layout the PCB
• Tested the clock circuit by plotting the waveform
  on an oscilloscope
• While the clock wasnt close to a perfect square
  wave, the rise and fall was fast enough (lt2ms)
• Once software was written, the I2C signals were
  able to be seen on an oscilliscope

6
Infrared Transmitter

• Simply an infrared LED modulated at 36KHz
• Turned on and off 67 times for various times,
  specified in milliseconds
• Tested by writing a simple program to loop
  through various remote control codes that were
  recorded
• Verified that TV reacted in the way expected

7
LED Array and Constant Current Source

• Infrared 20x10 LEDs for releasing Infrared light
  flood
• 1.7 V, 50mA DC voltage and current through each
  LED

8
LED Array and Constant Current Source

• Power Supply a 36V AC/DC Voltage Converter
  (Input 110V AC, Output 36V DC)
• Constant current source is made of 10 LM317
  Voltage Regulator and 10 100ohm resistors
• Each LM317 with a 100ohm resistor with it is
  connected in series with each row of LEDs
• Input Voltage for each row of LEDs 36V
• Input Current for each row of LEDs 5055mA
• System Power Consumption 20 Watts

9
Constant Current Source
10
Constant Current Source Development and Testing

• Initially we used Current Mirrors for making
  constant current, including 11 NPN BJT
  transistors successfully simulated, inconsistent
  at proto-type performance
• We upgraded Current Mirrors to Wilson Current
  Mirrors by adding extra transistors to the
  mirrors in order to stabilize the outputs
  successful at supporting a single row of LEDs,
  unsuccessful at supporting multi rows of LEDs
• We also thought of using Op-Amps, but its not
  cost-effective and it is hard to get Op-Amps with
  36V tolerance.

11
Final Design Reasoning

• Why we chose LM317 circuit plan
• Initially we chose Current Mirrors for
  lowering manufacturing cost, but neither Wilson
  Current Mirrors nor Op-Amp costs less than LM317.
• All other plans result over-heating problems
  for resistors as well as large power consumption

12
Final Development and Testing

• Current Measurements our final design shows that
  there is a current between 52.153.1 mA DC
  current through each row of LEDs with consistent
  performance
• LEDs have been on duty without a single failure
  for more than 45 minutes of testing since the
  completion of the LED Array.
• Temperature of resistors in the current source
  circuit has been significantly lowered.

Row Current (mA)
1 52.3
2 52.5
3 52.5
4 52.1
5 53.1
6 52.6
7 52.8
8 52.9
9 52.4
10 52.8




















13
Gesture Recognition Software

• Initializes and samples coordinate data from the
  Pixart camera via I2C.
• Stores and operates on the data received from the
  camera.
• Upon successful recognition of a gesture, outputs
  the corresponding signal to the television.

14
Software Development

• At first, wanted to have a sliding window of
  received coordinates with constant comparisons to
  an accepted gesture for all four of the available
  points given by the camera.
• Due to the severe limitations of our
  microcontroller, the gesture software had to be
  toned down.
• We decided to operate only on most prominent
  received point, and only detect the most
  rudimentary (straight line) gestures.

15
Software Implementation

• The recognizing algorithm stores (x,y) data into
  arrays. Upon seeing blank frames, or after one
  second data storage, it sends this data to a
  function that looks for a gesture.
• To find a gesture, we first find the distance the
  point moved in both the x and y directions. We
  then check direction of movement, and distance of
  movement to determine whether or not we can
  consider the data to contain a gesture.
• If a gesture is detected, we send a series of
  pulses to the IR transmitter to control the
  television.

16
Software Testing

• The first version of the software that we used to
  test simply relayed what the camera saw to a
  computer desktop. This was used extensively in
  debugging the later code and the hardware.
• The final software had to go through a couple
  iterations as we did not know the capacity of the
  Arduino, and so we had to compensate for a low
  memory, low processor situation.
• The software was then calibrated for minimum
  pixel movement to help filter out small movements
  as gestures.

17
Final Testing

• Once all the components were assembled, we
  addressed fine-tuning issues
• We were unable to get enough infrared light to be
  reflected off our fingers
• We tried using aluminum foil to reflect more
  light
• This provided more reflection, but the useful
  range was limited to about one foot
• Assembled a small LED pen in order to provide a
  single point source of light for the infrared
  camera

18
Final Testing

• With this point source of light, we were able to
  exceed 6 feet of range
• Having overhead lights on interfered only if the
  lights were in the line of sight

19
Future Improvement Solutions

• To find better reflection material in order to
  increase remote range and efficiency
• To use stronger and fewer infrared LEDs for
  better performance and lowering manufacturing
  cost
• To use more advanced cameras to increase
  gesture detection ability

20
Adaptability and Marketability

• Low Cost Analysis
• Proposed Manufacturing Cost
• Camera 35
• Microcontroller 1015
• LM317 x10 35
• LEDs and other minor components lt 5
• Total lt 30

Prototype Cost 144.47 Development cost for
Natal ???
Whats the price you are expecting from our
potential competitor Natal? 100? 200? or 300?
21
Adaptability and Marketability

• Fields which this low-cost gesture remote system
  can be applied at
• Basic TV Remote
• Projector
• Active Video Game Devices
• Confidential Access Devices
• Dancing and Other Entertainment Devices

22

• THANK YOU FOR COMING TO OUR PRESENTATION!