SoundSwitch

This program is an accessibility tool, intended for those with diminished motor
control - typically those users also rely on the usage of eye-tracking software
and hardware, which this program functions in tandem with.

This program was developed as a senior project for part of the Bachelor of
Information Technology at Otago Polytechnic.

Purpose & Reason

The purpose of this program is to provide an additional interface between the user
and their device. SoundSwitch allows users to quickly execute customisable keybindings
with their microphone, reducing the reliance on mechanical inputs such as mice.

This software was primarily developed for use in tandem with eye-tracking hardware and software,
with the intent of eliminating some of the tedium of using those products, such as
awkward usage when multi-key inputs are required, or the issue of eye-strain,
particularly when performing a similar task many times (such as highlighting or saving text.)

Installation

Requirements

• A microphone. Higher quality microphones will yield more accurate results.
• A little bit of patience - finding the sweet spot for your microphones sensitivity may take some trial and error.

For Users

For people simply wanting to user the software, without wanting to modify the underlying code,
we encourage you to head to our releases page and download our installer of the most recent version.

For Developers

To build SoundSwitch on your computer for editing, you will additionally require:

• A valid installation of Visual Studio 2013 or greater.

Once Visual Studio is installed:

• Clone this repository to a local directory.
• Inside the repository folder, open the sound-switch directly, and the .sln file contained inside.

Additionally, we include a pre-compiled .exe for one of the functions provided by scape-xcorrsound in this repository.
If you wish to compile your own scape-xcorrsound library (and executables), please visit their repo and follow their provided instructions.
SoundSwitch only requires one function of their library, called overlap-analysis.
Note that their library does not natively support windows, and modifications to their source code are required to make it compile on windows.

Usage

When you begin the program, you will be greeted with the home screen. You will need to perform some initial setup before this program will work correctly,
and some trial-and-error may be required before you right the correct detection values for your needs.

• Click the settings button.

  • If your recording device is present in the source list, simply double click it to set it as your current input.
    
    • If it is not, ensure that your recording device is properly plugged in, and click refresh sources.
  • Set your detection level. Detection level is a measure of how loud you must speak for your microphone to register commands.
    (Note: Detection level will vary depending on your microphone setup, you may require some trial-and-error to reach a good value, we recommend 600.)

• Once your settings have been configured, you will need to create a binding. A binding is a sound file that contains a short (~1 second) sound bite
  that you record, a key-command and a name are attached to this sound file.

• To create a binding, follow these steps:

  • From the home screen, click the bindings button.
  • Click the new binding button.
  • Follow the prompt that appears, taking special care to record a sound that is short, and easily repeatable.
    A good example for a binding should include these properties:
    • A short, distinctive sound for its trigger. Try to avoid using full words for your bindings,
      as repeated usage of such binds can cause voice strain.
    • The command should sufficiently justify the usage of a bind, such as combination keys or F-keys.
    • A descriptive name that explains what the binding will achieve for your own recognition.

An important note on bindings: The more bindings you have, the longer the program will take to respond to commands.
Additionally, frequently used bindings should use very short verbal triggers to avoid vocal strain.

• Now that you (presumably) have one or more bindings, you can head back to the main screen,
  and click Begin Listening. Your microphone is now waiting for a verbal command that matches a binding!
  When it detects a binding, it will execute that bindings associated action on the currently focused window, including the SoundSwitch itself.

Built With

• nAudio - An open source .NET audio and MIDI library, containing dozens of useful audio related classes (for use when saving recorded sounds).
• scape-xcorrsound - Library that facilitates the generation of comparison values
  between .wav files. We are using a modified version of this library to make it compatible with Windows.