<div dir="ltr">Oh my! Thank you Mr. Nico! :) I will pick up a smaller project for google summer of code. However, I am too much obsessed with voice recognition systems. But now I understand that it is not a small issue. Someday, I will surely work on speech synthesis. As you mentioned above, it is certainly not doable in small span of SoC. Thanks once again! :)<br></div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Feb 17, 2015 at 6:01 PM, Nico Kadel-Garcia <span dir="ltr"><<a href="mailto:nkadel@gmail.com" target="_blank">nkadel@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">On Tue, Feb 17, 2015 at 5:49 AM, Abhilash Mhaisne<br>
<<a href="mailto:abhilashmhaisne@gmail.com">abhilashmhaisne@gmail.com</a>> wrote:<br>
> Sir / Madam,<br>
> I am an engineering student, and gsoc aspirant. I want to<br>
> develop a easy-to-use voice recognition system, which would be capable of<br>
> judging even minor changes in accents, for fedora systems.<br>
<br>
</span>And I'd like a pony, with pretty wings that can fly to Jupiter and<br>
makes cookies.<br>
<br>
More seriously, voice recognition is one of the great challenges of<br>
the last 50 years of speech analysis, approached by large and small<br>
companies and research labs around the world. "Judging even minor<br>
changes in accents" is a nightmare in real work: every commercially<br>
available computer microphone system for the last few decades has made<br>
the same basic mistake for collecting speech. They completely screw up<br>
"plosives", sounds like "b" and "p" that have a lot of high frequency<br>
information that gets completely munged by the digitization and<br>
undersampling. And do not get me *started* on the ongoing fallacy that<br>
"if we just collect more speech samples, we'll somehow be able to<br>
analyze them". I'm aware of companies with software patents spending<br>
many millions of investment money in the work, and they're still<br>
getting their ass handed to them in the marketplace by smaller, local<br>
apps that handle basic speech locally and *don't try to get fancy*.<br>
<br>
And in case it's not clear: I did artificial hearing research for a<br>
dozen years, designing electronics for cochlear implants. Every modern<br>
microphone=>speech analysis system screws up the plosives, for a stack<br>
of reasons we could discuss elsewhere.<br>
<span class=""><br>
<br>
> I wish to implement this system in python, making use of<br>
> the open source CMU Sphinx system. I would love to do this under a mentor<br>
> for gsoc 2015.<br>
><br>
> Regards!<br>
><br>
> Abhilash Mhaisne<br>
<br>
<br>
</span>See above. "voice recognition" and accent immunity is not a "python"<br>
sort of problem, it's a massive ongoing computer and speech research<br>
issue. Pick something *smaller*.<br>
<span class="HOEnZb"><font color="#888888">--<br>
devel mailing list<br>
<a href="mailto:devel@lists.fedoraproject.org">devel@lists.fedoraproject.org</a><br>
<a href="https://admin.fedoraproject.org/mailman/listinfo/devel" target="_blank">https://admin.fedoraproject.org/mailman/listinfo/devel</a><br>
Fedora Code of Conduct: <a href="http://fedoraproject.org/code-of-conduct" target="_blank">http://fedoraproject.org/code-of-conduct</a></font></span></blockquote></div><br></div>