Maxim Vedenev
Freelancer, Matlab programmer,
Comsol programmer, Physicist

vedenev@ngs.ru

    
skype messenger:   My status      Call me via skype:   Skype MeT!
 
Matlab projects
Java projects
C# projects
C++ projects
Comsol projects
Physics projects
My Education
Payment methods
My resume
My videos
My ideas
 
 

C# projects

 
 
   

     This is captcha recognition project. The code was made in C# as dll class to be able to run in visual basic. The captcha has 4 fixed fonts with fixed size. reached recognition ratio is 75% and time is about 0.85 sec per captcha.
 
 
   

     This is optical character recognition. It can read only 3 fixed fonts. It load .tif image as input. Recognition is made as compare symbol from database inside code with symbol on the image. It can detect strings to decrease search space. Processing time about 2 sec per page.
 
 
   

     This is Dual-tone multi-frequency (DTMF) signaling detection. The employer has problem with free software for detection. It has not sufficient correctness of detection. So the task was detect the tones good enough. Goertzel algorithm was used. Tone amplitude uniformity was used to increase correctness of detection. Batch mode is possible.
 
 
   

     This is yahoo visual captcha processor. It detect is captcha wavy or straight. Employer already had some captcha breaker and this breaker has better recognition if the captcha is captcha straight. So this C# code just detect wavy/straight. This made as bottom line points spread calculation. Normalized standard deviation separated with threshold, click here to see graph. It is C# code. It was converted to dll to be able to use in Visual Basic.
 
 
   

      There was programm in C# (GUI) that get sound of piano from microphone and then recognize and print notes it in textbox. It print each 0.1 sec. Initially it was tested in Matlab. Testing matlab codes can found here: link to matlab file exchange. The problem is that piano notes sound spectrum is complex, see animated spectrum: http://www.youtube.com/watch?v=5xjD6SRY8Pg. There was made ring buffer for real-time microphone listerning. Notes recognition processing are made in separeted thread (not UI thread) to be able to contol the work.
Reached result:
1 note: 93% of right answers
2 notes simultaneously : 61% of right answers
3 notes simultaneously : 31% of right answers
4 notes simultaneously : 13% of right answers

detaled research description can be found in doc-file: notes_recog_research.doc
source C# code can be found here: http://sourceforge.net/projects/notes-detector/