Neural Network(Alphabet and Number recognition)

[leekk8]

nntool manual

I never use the nntool, so Im not sure about this. I write the code myself...

Anyway, you try to find out is the error caused of the size. I think there should not be any limit on this. Maybe you need to fulfill some requirement of nntool.

 

[vjfaisal]

neural net matlab alphabet recognize

dear leekk1

m also writing my own code, i am just comfiring my work through NNtools. so please if you could send me your code.....i can get the idea , how to use....

please send me at

djarzoo@gmail.com


best regards

 

[phongphanp]

matlab neural network alphabet

hi

leekk8!!! Could you post the sample result image,which look best for understand?


thanks

 

[leekk8]

alphabet + matlab

vjfaisal,

The code is already attached in earlier post here, refer to my earlier post by scrolling up. There is backpropagation code, how you want to change the architecture, you just need to change the size of the weight matrix. Theorically, you can use any size of input and output, but the size will affect the speed of the training.

phongphanp,
What result image you want? I never done alphabet recognition or brain state recognition before. My experience is fingerprint recognition using backpropagation network. And my result is not an image but a matrix.

 

[amitgangwar_vlsi]

send my number with alphabets

thank you leekk

with your help, suggessions and guidance i have implemented the
neural network for 0-9 numbers and A-Z alphabets with commands
and nntool both.it is not 100 % accurate but it is more than 95% accurate.

many many thanks once again

amit gangwar

 

[leekk8]

alphabet targets matlab neural network

amitgangwar,

You're welcome. So happy heard that you successfully implemented the neural network. 95% accuracy is a great job already. Last time when I did, the accuracy is only 80%

Keep it on...

 

[vjfaisal]

number recognition in matlab

hi leekee!

its god to see that u help amitgangwar , he is lucky get ur guidence,....

now please also guide me in my work also , because it is a very tough work than that........... i am trying my best to achieve my goal in classification of brain states... please help me when required,


regards

 

[leekk8]

parameter in nntool + matlab

vjfaisal,

I also hope I can help you, but I have no experience on brain state classification.

On which part you're stuck? Firstly, you need to represent your input signal in a vector, and justify how's your target output. After that, develop the neural network where the architecture must fit your input and output.

 

[snoopfan]

use nntool to create smiley patterns

Anybody wrote the code for alphabet(phoneme / vowel) recognition in MATLAB?

Cheers...

 

[vjfaisal]

dear leekke1
ya i have aproblem how ,tis complex data could be represented in vector form, to import it in NN tool for analyze. i have the two possiblities here....please try to understand and help me in these...

1) To classify the three/two states of brain, it is rightto enter their respective frequencies as an input . for e.g : to classify the brain state , name ( gamma), which has standard frequency band of , let suppose 19 to 25 hz.
For input vector can i write it like this:
p=[19 20 21 22 23 24 25]

2) second possibility relate to the concept that after we extract our desired signal from brain , then filter it ( bandpass), the frequency range for gamma (19 to 25) hz
at last when we have the last filter out results, let suppose in vector k.
Can we load/import it as a input vector in classification. but here , problem is the last filtered results has vector of 1*236, for this case ,
input vector is 1*236,
How we adjust the target in this case, because in NN matlal tool , should have 2 rows for input vector, and one row for target vector with same i/o columns.

please , understand it , and help in this , so thatai could initiate my work , properly


regards
vjfaisal

 

[actinio]

Hi members of this area..
I'm interesting on using code of PCNN in Matlab. anybody has some information about it? I'm starting to read about it and I would like to work in it on my thesis..poor information can be provided from Internet, and few people had worked on it


Pd: I´ve found codes on the Alabama University, and I'm going to read a book who has this title: Image Processing Using Pulse-Coupled Neural Networks
T. Lindblad J.M. Kinser.-

Thank you in advance.

actinio.-

 

[leekk8]

dear leekke1
ya i have aproblem how ,tis complex data could be represented in vector form, to import it in NN tool for analyze. i have the two possiblities here....please try to understand and help me in these...

1) To classify the three/two states of brain, it is rightto enter their respective frequencies as an input . for e.g : to classify the brain state , name ( gamma), which has standard frequency band of , let suppose 19 to 25 hz.
For input vector can i write it like this:
p=[19 20 21 22 23 24 25]

2) second possibility relate to the concept that after we extract our desired signal from brain , then filter it ( bandpass), the frequency range for gamma (19 to 25) hz
at last when we have the last filter out results, let suppose in vector k.
Can we load/import it as a input vector in classification. but here , problem is the last filtered results has vector of 1*236, for this case ,
input vector is 1*236,
How we adjust the target in this case, because in NN matlal tool , should have 2 rows for input vector, and one row for target vector with same i/o columns.

please , understand it , and help in this , so thatai could initiate my work , properly


regards
vjfaisal

Hi, vjfaisal

For this input format, you can test it out which one is better. Anyway, my experience tell me that the second method would be better. 1x7 input is looked like too little. 1x236 is good enough, provided after the filtering, the feature is still there.

You can try to build the filter and get the 1x236 vector as input. I don't think the nntool limit your number of input neuron. You need to search it out, as I really know nothing about the nntools. With the code I write, any number of input neurons can be implemented. So as the output layer. Figure out what format of output you want, so that you can configure your output neurons accordingly.
For example, your output would be raise up hand, put down hand, raise up leg, put down leg, etc. Then your output could be [0 0 0 1] represents raise up hand, [0 0 1 0] represents put down hand, [0 0 1 1] represents raise up leg, and [0 1 0 0] represents put down leg. In this case, your number of output neuron is 4.

 

[payate]

hi leekk8..

do u know how to convert the signal into vector and represent it into neural net..
for your information, the signal is isolated of number and i'll train it using neural net (backprop)..
do u have any sample on that..?

 

[vjfaisal]

hi

amitgangwar

please answer me these two questions:

1) what is difference between desired output vectors and desired target vectors,in back propagation neural network.
( need details)



2) and please send me your code so that i can get some guidence from that , and make to classify brain states..................

regards
anjum

 

[amitgangwar_vlsi]

output is the output of the neural network which you get.
and target is the output which you want to get, which you can recognize.

what you have to do?
try yourself, if you would face any problem then ask me freely
i will send you my code..

 

[leekk8]

hi leekk8..

do u know how to convert the signal into vector and represent it into neural net..
for your information, the signal is isolated of number and i'll train it using neural net (backprop)..
do u have any sample on that..?

What signal is this?

 

[lamhoangson2003]

i'm doing project fingerprint recognition in VC,I need some some source code neural network for my project.Everybody have,pz send to me.Thanks for your advance.

 

[leekk8]

Why there is still people just asking for code? This forum is for discussion and learning, but not spoon feed. If you got any question, you can ask here, but never ask for the code directly without asking anything. Nobody will understand what you want.

 

[jsn8610]

Re: how to use a complex nntool matlab

Hi..
i'm currently doing a project related on using MATLAB in any application using neural network. any ideas? best regards.
Thanks

 

[objmode]

hello....I'm developing a source code for hand gesture recognition of sign language using neural network...where an image of the hand gesture is presented and it is classified to the correct alphabet or number..if anybody has anything close please help with it...