how to reduce program execution time in matlab

[sslmadhu99]

I am doing speech -music discrimination using matlab as my main project. In that i am having rms module , zcr module and lpc module. For each individual signal i have to calculate all these. There are 4 music signals and 4 speech signals for training and 8 music signals and 8 speech signals for testing . Again I am having classification algorithm . While I am executing my program , I am getting correct output but it is taking around 30 or 45 minutes for that :roll: .
I am not able to find whether that is due to fault in my program or due to complexity of program. To reduce time I used infinite loop elimination etc. But still I am not able to reduce the execution time. So I am requesting anyone to please help me in-order to get my output in less time.

Thanking you.
CH.V.S.S.Madhuri

 

[emresel]

Hello;
No direct suggestion but these links may help to improve.
http://optlab.mcmaster.ca/~yzinchen/fast-matlab-code.pdf
http://www.mathworks.com/help/matlab/matlab_prog/profiling-for-improving-performance.html
Good luck

 

[floopz]

Try running a small data set through the MATLAB profiler. This will tell you where the offending (slow) bits of code are. Often the problem is simple (not pre-allocating matrices or appending datasets in loops). MATLAB also executes datasets faster when indexing by row rather than column.

 

[sslmadhu99]

Hello;
No direct suggestion but these links may help to improve.
http://optlab.mcmaster.ca/~yzinchen/fast-matlab-code.pdf
http://www.mathworks.com/help/matlab/matlab_prog/profiling-for-improving-performance.html
Good luck

Thanku sir

- - - Updated - - -

Try running a small data set through the MATLAB profiler. This will tell you where the offending (slow) bits of code are. Often the problem is simple (not pre-allocating matrices or appending datasets in loops). MATLAB also executes datasets faster when indexing by row rather than column.

Sir thanks for ur response . But actually i don't know how to work with profiler . Please give me some guide lines to work with that.
Thanking you sir.

 

[permute]

Matlab works best when you try to avoid C-style loops and focus on vector/matrix operations. rms should have a simple function, if not just based on mean() and var().

zero crossing as well, with:

zc = sum(((x(1:end-1) < 0) & (x(2:end) >= 0)) + ((x(1:end-1) >0) & (x(2:end) <= 0)));

 

[Ajinkya Talele]

I am doing speech -music discrimination using matlab as my main project. In that i am having rms module , zcr module and lpc module. For each individual signal i have to calculate all these. There are 4 music signals and 4 speech signals for training and 8 music signals and 8 speech signals for testing . Again I am having classification algorithm . While I am executing my program , I am getting correct output but it is taking around 30 or 45 minutes for that :roll: .
I am not able to find whether that is due to fault in my program or due to complexity of program. To reduce time I used infinite loop elimination etc. But still I am not able to reduce the execution time. So I am requesting anyone to please help me in-order to get my output in less time.

Thanking you.
CH.V.S.S.Madhuri

Hello,
I am also working on the speech recognition project. 30 to 45 minutes seems to be a pretty large amount of time. Which language are you using for the program? and can you please share ur knowledge about the speech recognition like the books you referred of papers you went through? I know its a late to reply to this post but i came across this today.