Voice Control System
This software is a piece of ancient history; I worked on that circa 2007.
A Voice Control System for controlling WinAmp under Microsoft Windows using voice commands. Works completely off-line, relying on user training instead of cloud computing.
A piece of software I wrote to be able to switch music without using the computer directly - it helped me avoid distractions during a period of heavy studying.
The software operates on a simple tree grammar:
Note the existence of
Release pair of command. The first one locks system to the
Music control subtree,
so that I didn’t have to repeat that part of the tree to issue more music commands. The second one releases the lock, thus
returning the default state of the program to the top level.
The system was trained using built-in Speech API training in control panel, with a set of all words in the entire grammar. After few training iterations (see testy.pdf for my training sheet) in various conditions - different places of the room, different volume of music playback - the system became pretty reliable, and I could use it from any place in the room even with loud music or radio playing.
Overall, this system kind of proves that useful voice recognition does not require computation in the cloud - it was entirely feasible using 2007 tech and a mid-range PC from that era.