Thomas Mack, Michael Hiller, Trevor Andreas, Christian Estrada, Trinh Tran

1
Wireless Locator Voice Control
Thomas Mack, Michael Hiller, Trevor Andreas,
Christian Estrada, Trinh Tran
tam023000_at_utdallas.edu, mdh026000_at_utdallas.edu,
student3_at_utdallas.edu, cxe041000_at_utdallas.edu,
t.n.tran_at_student.utdallas.edu
Department of Electrical Engineering Erik Jonsson
School of Engineering Computer
Science University of Texas at Dallas Richardson,
Texas 75083-0688, U.S.A.
Project Goals
Project Results

• Develop a system to determine the users position
  and use voicerecognition software to control
  devices in the surrounding area.
• Develop a local position system using wireless
  access points
• Record and send a .wav file from the client to
  the server.
• Perform voice recognition on a .wav file sent to
  the server
• Use existing 802.11 networks

• A combination of both Java and C code were used
  to create the system. Different programming
  languages were used because some of the functions
  were easier to implement in one as opposed to the
  other.
• The signal strengths of all access point that the
  client sees are found using the windows network
  API through a C program. These are saved to a
  file for transmission across the network.
• The server reads the file of the signal strengths
  and compares then to a database of empirical that
  was collected for each room involved in the test.
  This involves doing a correlation between the
  transmitted signal strengths and the empirical
  data by comparing matching MAC addresses. The
  signal strengths are normalized to the mean of
  each set of values to increase the sensitivity of
  the system. This algorithm was written in Java.
• The algorithm used to determine the location of
  the transmitted signal was not as accurate as
  originally hoped, but given enough space between
  the rooms, it can fairly consistently determine
  the correct location. Testing was done in
  different environments and seems to work best
  when an access point is only seen by one room and
  not the other rooms.
• Using the Windows API we were able to record a
  voice command into a .wav format for the client.
  Then using the TCP/IP protocol the .wav file,
  containing the voice command, and the .dat file,
  containing the locationdata, are sent to the
  server.
• We were able to successfully take in a
  pre-recorded .wav file and use the Sphinx voice
  recognition library developed by Carnegie Mellon
  University to successfully decipher what the user
  said. We then were able to dictate a one of the
  predefined outputs.

Project Overview

• A proof of concept for the system was developed.
  The goal was to create a prototype with the basic
  abilities of determining location, data transfer,
  and voice recognition.
• The user will speak into a microphone located on
  the client device. The client device then
  converts the voice into a .wav file.
• The client will read in signal strengths from
  nearby access points.
• The client sends both the signal strengths and
  .wav file across the wireless network.
• The server reads in the .wav file and performs
  voice recognition to determine the command.
• The server determines the location of the client
  by comparing the transmitted signal strengths
  with a set of empirical values.
• The server then dictates a predefined output
  based upon the clients location and the voice
  command.

Project Conclusions/Outcomes

• Laptops were used for both the client and server
  within this project. A final implementation
  would include a design of a small handheld device
  that would be more practical for real world
  usage.
• The system used to determine the location of the
  user was not as accurate as hoped. The accuracy
  of the system would need to improve in order for
  this to really work within a real household.
  Alternatively, an entirely different method could
  be developed.
• We were able to complete a proof of concept.
  There is still a substantial amount of
  development that would need to be done to
  implement this system.