Call Search
     

New to Ham Radio?
My Profile

Community
Articles
Forums
News
Reviews
Friends Remembered
Strays
Survey Question

Operating
Contesting
DX Cluster Spots
Propagation

Resources
Calendar
Classifieds
Ham Exams
Ham Links
List Archives
News Articles
Product Reviews
QSL Managers

Site Info
eHam Help (FAQ)
Support the site
The eHam Team
Advertising Info
Vision Statement
About eHam.net

   Home   Help Search  
Pages: [1]   Go Down
  Print  
Author Topic: FLDIGI: two new experimental features to improve CW detection and decoding  (Read 5297 times)
AG1LE
Member

Posts: 137


WWW

Ignore
« on: May 13, 2012, 10:29:38 PM »

I have implemented two new experimental features to improve CW detection and decoding capabilities in FLDIGI softare.

1) Matched Filter  feature improves signal-to-noise making barely audible morse code detection  to work better in FLDIGI  CW mode.

2) Self Organized Map feature is using an artificial neural network algorithm from SOM http://en.wikipedia.org/wiki/Self-organizing_map to calculate best matching codebook entry from input vector using Euclidian distance method.

I have documented these experimental features in my blog as well as test results:

http://ag1le.blogspot.com/2012/05/fldigi-adding-matched-filter-feature-to.html

I am looking for software developers who can compile this from FLDIGI tar ball, and don't need my support to get this running in their setup.

The software is alpha quality and not ready for broader distribution. If you have interest in testing/improving this software let me know. 
Logged
AG1LE
Member

Posts: 137


WWW

Ignore
« Reply #1 on: May 20, 2012, 12:16:31 PM »

I have implemented two new experimental features to improve CW detection and decoding capabilities in FLDIGI softare.

1) Matched Filter  feature improves signal-to-noise making barely audible morse code detection  to work better in FLDIGI  CW mode.

2) Self Organized Map feature is using an artificial neural network algorithm from SOM http://en.wikipedia.org/wiki/Self-organizing_map to calculate best matching codebook entry from input vector using Euclidian distance method.

I have documented these experimental features in my blog as well as test results:

http://ag1le.blogspot.com/2012/05/fldigi-adding-matched-filter-feature-to.html


Thanks to Dave W1HKJ  and the FLDIGI community this project has moved forward quite a bit over the last week.  

I posted new promising test results in here:
http://ag1le.blogspot.com/2012/05/fldigi-matched-filter-and-som-decoder.html

I am also working to improve the SOM decoder to handle the "outliers" - CW characters containing errors due to noise spikes or timing problems  -  see more details in the blog in here
http://ag1le.blogspot.com/2012/05/morse-code-decoding-with-self.html
.

73
Mauri AG1LE
« Last Edit: May 20, 2012, 08:19:03 PM by AG1LE » Logged
STAYVERTICAL
Member

Posts: 875




Ignore
« Reply #2 on: May 21, 2012, 03:47:35 PM »

You guys are all geniuses (and I don't mean that in the discounted way it is used these days).

I would like to thank you, Dave and all the other FLdigi developers, for developing and progressing FLdigi.
Although my computer skills are not up to the task of doing something like that, I can appreciate art in programming when I see it.

73 and many thanks for this great software.
- Rob
Logged
AG1LE
Member

Posts: 137


WWW

Ignore
« Reply #3 on: May 23, 2012, 08:24:43 PM »


I would like to thank you, Dave and all the other FLdigi developers, for developing and progressing FLdigi.
Although my computer skills are not up to the task of doing something like that, I can appreciate art in programming when I see it.

73 and many thanks for this great software.
- Rob

Thanks Rob  Smiley

I think we really ought to thank  Dave & the team he has assembled around FLdigi.  I have learned a lot just studying the FLdigi software - it has some really cool collection of code. I have also found the team very responsive and helping  newbies like myself getting new ideas implemented.

Decoding Morse code is such a perfect application  area  -  there are enough challenges even for human operators to copy CW perfectly, not to mention teaching computers how to do it.  Add some noise,  irregular fists,  pile-up conditions, random signal fading and other daily events on the ham RF bands - creating smart software that is able to deal with all this and software that learns from experience is quite a challenge.  We have now personal computers powerful enough to run complex algorithms in milliseconds  that took hours when  these neural network algorithms were first discovered 20 - 30+ years ago.

I wonder if there is anybody else interested in Self Organizing Maps and other artificial neural networks algorithms?

73
Mauri  AG1LE



Logged
AG1LE
Member

Posts: 137


WWW

Ignore
« Reply #4 on: May 28, 2012, 11:29:57 PM »

Quote
I am also working to improve the SOM decoder to handle the "outliers" - CW characters containing errors due to noise spikes or timing problems  -  see more details in the blog in here
http://ag1le.blogspot.com/2012/05/morse-code-decoding-with-self.html
.


New test results in here: http://ag1le.blogspot.com/2012/05/fldigi-analysing-som-decoder-errors.html


73
Mauri AG1LE
Logged
Pages: [1]   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.11 | SMF © 2006-2009, Simple Machines LLC Valid XHTML 1.0! Valid CSS!