GSoC Project Status Update 05: Speech Recognition in Openmoko
saurabhgupta1403 at gmail.com
Mon Jul 7 20:57:23 CEST 2008
Here is the project status update for Speech Recognition for Openmoko.
This week passed in testing phase only. I tested the various subroutines of
speech recognition code with various sound samples. The main part was the
vector quantization one which is showing somewhat good results. However for
good recognition, the size of Codebook
using vector quantization must be large enough. I finally decided to prepare
a codebook for 256 vectors for the speaker independent mode recognition. For
speaker dependent mode the size will be decided dynamically at run time
(depends on the number of training sequences fed by the user).
Besides this I have now started working on openmoko toolchain and
build some initial applications to get comfortable with openmoko tools. I
will have to learn now to understand the interfacing of openmoko peripherals
with the running applications. I am working on it and getting to learn a lot
of things. The next aim is to read the
adc channels from the openmoko and embed it with speech recognition
Electronics and Communication Engg.
NSIT,New Delhi, India
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the community