FPGA Central

Hello,

I'm looking to meet an undergrad design requirement, which can be don
with independent study. I'll graduate in December, so I have a lot o
coursework under my belt. I'll also have all summer to research an
prepare. I've focused on communications and dsp, and these are th
upper-level undergraduate courses I've taken:

* Digital Signal Processing
* Digital Filter Design
* Communications Systems
* Digital Communications
* Wireless Communications
* Introduction to Feedback Systems
* Introduction to Digital Control.

I've also taken a required electronics course, so I have some solderin
experience as well as bit of experience building transistor circuits, bu
I'd rather work on a project that is software based. I'm a pretty goo
programmer and I have a lot of matlab experience. If I needed any hardwar
aside from my PC, I'd want to keep the cost under $150 dollars.

I think I'd probably prefer to do a design project focusing on a DS
topic.... like maybe something with audio? I'm already doing anothe
independent study in communications, and my school offers mor
communications courses than DSP, so I want to balance out a little and tr
something in DSP.

So, does anyone have any ideas for interesting things I might be able t
design that are DSP related? (Or even communications related, if you ca
think of something really interesting) I thought about designing some audi
filters or something like that (I had the idea of using one person's voic
as a reference in order to filter another person's voice and make it soun
like the first person.. but I'm guessing this might be too challenging?)
just wanted to know what others thought.

Thanks!

On May 16, 2:41 pm, "jfrog" <ajmcg...@gmail.com> wrote:
> If I needed any hardware
> aside from my PC, I'd want to keep the cost under $150 dollars.
>
> I think I'd probably prefer to do a design project focusing on a DSP
> topic.... like maybe something with audio? I'm already doing another
> independent study in communications, and my school offers more
> communications courses than DSP, so I want to balance out a little and try
> something in DSP.
>
> So, does anyone have any ideas for interesting things I might be able to
> design that are DSP related? (Or even communications related, if you can
> think of something really interesting) I thought about designing some audio
> filters or something like that (I had the idea of using one person's voice
> as a reference in order to filter another person's voice and make it sound
> like the first person.. but I'm guessing this might be too challenging?)

i think that's called a channel vocoder and might not be too bad.
lot'sa filters (both for analysis and for synthesis). another way to
do it uses Linear Predictive Coding and then you need to worry a
little about "Levison-Durbin recursion". LPC is much more involved
than simple channel vocoding where you have an analysis filter (BPF)
followed by an envelope follower where this envelope controls the
amplitude of another identical filter with your "second" person's
voice going it. two big banks of filters.

if you're gonna do this real time, you need to get yourself a dev kit
and maybe that Freescale one recently discussed here is the ticket.
or talk to Al Clark (Danville Signal Processing) about a SHArC dev
kit. for non-real-time, all you need is MATLAB or Octave (or some
other math programming environment). you could also do it in C.

r b-j

"jfrog" <[email protected]> writes:

> So, does anyone have any ideas for interesting things I might be able to
> design that are DSP related? (Or even communications related, if you can
> think of something really interesting) I thought about designing some audio
> filters or something like that (I had the idea of using one person's voice
> as a reference in order to filter another person's voice and make it sound
> like the first person.. but I'm guessing this might be too challenging?) I
> just wanted to know what others thought.

These are undergrad projects that have been done in the recent past,
they might give you some idea:

Automatic birdsong recognition
http://demos.itee.uq.edu.au/2004/pro...udent=s4009474

Content-Based Music Similarity
http://innovexpo.itee.uq.edu.au/2003/exhibits/s363304/

MusicFido: Song Searching
http://innovexpo.itee.uq.edu.au/2003/exhibits/s4019820/

Echo Cancellation on a DSP
http://innovexpo.itee.uq.edu.au/2003/exhibits/s365914/

Head-Related Transfer Functions
http://innovexpo.itee.uq.edu.au/2003/exhibits/s354240/

Real-Time Pitch Correction for the Human Voice
http://innovexpo.itee.uq.edu.au/2003/exhibits/s363076/

Most are software-based and were reasonably successful.

Ciao,

Peter K.

--
"And he sees the vision splendid
of the sunlit plains extended
And at night the wondrous glory of the everlasting stars."

jfrog wrote:
> Hello,
>
> I'm looking to meet an undergrad design requirement,

First to nitpick a little.

In the eyes of your school, what is the purpose of "an undergrad design
requirement"? Or to rephrase, how will it be graded?

The answer to above might/should (??? be modified by your personal
goal in satisfying "an undergrad design requirement" ;!


> which can be done
> with independent study. I'll graduate in December, so I have a lot of
> coursework under my belt. I'll also have all summer to research and
> prepare.

[snip info which does not relate to MY reply]
[smaller snips indicated by "..." ]

>
> I've also taken a required electronics course, ... , but
> I'd rather work on a project that is software based. ... If I needed any hardware
> aside from my PC, I'd want to keep the cost under $150 dollars.
>
> I think I'd probably prefer to do a design project focusing on a DSP
> topic.... like maybe something with audio? ...
>
> So, does anyone have any ideas for interesting things I might be able to
> design that are DSP related? ... I thought about designing some audio
> filters or something like that (I had the idea of using one person's voice
> as a reference ...

[ OK that last ... possibly should have been a *SNIP* ;]

This suggestion is an outgrowth of an personal interest in *SOME*
aspects of speech recognition that is decades old that I've not had
*simultaneous* occurrence of available time, expertise (personal or
otherwise) and motivation.

"A Modest Proposal"

1. Determine the formants of a particular person.
[Or preferably do it for a number of individuals]
2. For each set of formants, create a FIR filter to emphasize those
frequencies.
[personal "prejudice" suspects it might be beneficial to tweak gains
such that output at each frequency is similar.]

The above is neither difficult NOR *SIGNIFICANT*
[nor probably deserving of desired grade]

Suggested significant(???) experiment.

Record speech by individual(s)
Mix with experimenter selection of noise sources
Filter each with filter created above

Analyze results by any/all of
a. how accurately can human understand
b. how accurately can a specific speech recognition program [trained by
specific speaker in noise free environment] recognize the passages


I would love to see the results of this experiment.

I'm definitely not an expert, but I've got a gut feel this might
demonstrate that modern speech recognition is barking up ANTIQUE trees.

Thanks for all the links. Great stuff.


>Automatic birdsong recognition
>http://demos.itee.uq.edu.au/2004/pro...udent=s4009474
>
>Content-Based Music Similarity
>http://innovexpo.itee.uq.edu.au/2003/exhibits/s363304/
>
>MusicFido: Song Searching
>http://innovexpo.itee.uq.edu.au/2003/exhibits/s4019820/
>
>Echo Cancellation on a DSP
>http://innovexpo.itee.uq.edu.au/2003/exhibits/s365914/
>
>Head-Related Transfer Functions
>http://innovexpo.itee.uq.edu.au/2003/exhibits/s354240/
>
>Real-Time Pitch Correction for the Human Voice
>http://innovexpo.itee.uq.edu.au/2003/exhibits/s363076/
>
>Most are software-based and were reasonably successful.
>
>Ciao,
>
>Peter K.
>
>--
>"And he sees the vision splendid
>of the sunlit plains extended
>And at night the wondrous glory of the everlasting stars."
>
>
>