Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Neural Network(Alphabet and Number recognition)

Not open for further replies.


Member level 5
Jul 6, 2007
Reaction score
Trophy points
Activity points
matlab number recognition

can anybody please tell me how can i implement neural network in matlab for number and alphabet recognition.

thanks in advance
amit gangwar

alphabet recognition newral network

You can represent the number/alphabet with a binary matrix, then treat this as input.

Which neural network you want to use?
alphabet recognition matlab

thanks for reply

actually i am working on a project....number plate recognition

so, i did till character segmentation.
after that how can i generate neural network...i don't know.......

plz can you tell how can i store target value,input,output...etc.....

plz guide me

amit gangwar


    Points: 2
    Helpful Answer Positive Rating
matlab alphabet

What neural network you want to use? Do you have idea about neural network?

After you do segmentation, I guess you can convert it to a matrix. Store the matrix as the image.

What software you're using to do the recognition? I only have experience in using MATLAB.

Target image, you should get a proper alphabet and number matrix...
Input will be your training sample and testing sample.
Output, this depends on how you want to do the can have the target image as your output, or use something to represent the alphabet and number.

input image and target image in nntool matlab

dear leekk8

i am new in neural network to implement it in MAtlab , so if u ahve experienced in it, so please tell, me how to create neuron in Matlab. please five some code help , which clear me how to make it??

bst regards

Added after 2 minutes:

im doing my project in brain computer interface , so i have to do neuro network for the classification of states ,in brain.


    Points: 2
    Helpful Answer Positive Rating
recognition network of the brain

Hi, mdzafar

I never use the built in nntool in Matlab, but I write the code myself in matlab. Actually, once you understand the algorithm of the neural network, you can write the code in any programming language.

Anyway, before this, you must understand about neural network and decide which network is most suitable for your purpose, classification of states.

For some beginner guide and code, you can refer to

alphabet recognition test

hello sir/mam

i am trying very hardly on neural network.
but i could not get the solution.

but i get that i can use feedforward back prop method for my project number and character recognition.

link which you have given is opened but downloaded .rar file is empty.
can you please send it on my mail

i have some doubts also , please clarify.....

my input is 200X1 vector, so what is the input range for it??????? is it [0 1] ???

i have to recognize 26 char and 10 number, so my target should be 36X36????

how should i have to decide number of neurons for different layers??????

please please guide me.
it's very urgent for me..

so please help me

many many thanks in advance

amit gangwar

how to use a complex nntool matlab


This is the zip files containing all the matlab codes. I have tested, it's fine.
Anyway, you need MATLAB to open the files, as all the codes are m files.

About the number of neuron at each layer, you need to test and identify which architecture is the best solution for your project.

The input, you can make binary if the vector you use is in binary...In fact, I feel that binary input is more efficient for backpropagation network.
  • Like
Reactions: konarkk


    Points: 2
    Helpful Answer Positive Rating
backpropagation, pattern recognition, matlab

dear leell8,

i am newqw in neural network , but i want to use feed forward backpropagation for pattern recognition in EEG data. Ho i do for it.

i want to recognize/classify the brain states during one of the four task perform by our subject ( motor imaginary)

please help me..i really want this...for my project help ......

best regards

number recognition from an image, matlab


Im not sure what is EEG and how is the brain states things...

For backpropagation network, you need to decide how you want to represent your input and output.
After you decide the input and output format, you can write the code for the neural network and train the network to recognize the pattern.

nntool matlab

hi leekk8.......

i did study on neural network.
i decided to use feedforward back-prop network.
my input is 10 numbers 0-9.
i have changed these to the 200X1 vec.
i took 20 and 10 neuron respectively for first and second layer.
and logsig function.
right now i m doing for only numbers.
i thought that if it would work efficiently then i would try for alphabet.
i trained the network.

but it does not work properly for 1 and 0.
it is giving other numbers for 1 and 0.
is it depend upon the epochs,lr etc training parameter????
how should i decide to set tese parameters????

so please clarify my doubts.

many thanks in advance

amit gangwar

number plate recognition neural network in matlab

hi lekk8

i have a case like this....
i want to classify the brain states, when our subject perform four different task one after other. i want to classify which signal has varition at the time when his (hand , foot )he imagine to move in what ever direction he want to move.

there were four states ,when our subject perform four different tasks.
for this have to use backpropagation algorithm to solve it.
Now please if u have or anyone could help me in this problem. its very urgent, becoz its my final year project......................

best regards

recognition backpropagation network source

Yes, you need to train your network with different parameters, such as learning rate, number of neurons etc...

Other numbers 2-9 are ok? Try to check is it 0 and 1 have similarities with other numbers.


I think you should have pairs of data, right? For example, this signal input, u get an output, another signal input, you get another output.

Try to represent your input with a vector or array, input to the neural network, and also represent your output with an array and put it as target output.

nntool examples



i cannot understand you, please stepwise tell me.. for input neuron ,how much neurons required in my case, and how much hidden layers is requried, then for output vector............

best regards

matlab code for number plate recognition

thamks leekk......

with your suggestions i have designed neural network for numbers.
now i am trying to design neural network for alphabets....

thanks again

amit gangwar

matlab number recognition source code

I have source code for neural network implementation in C.
I use for my final year project...

If you want to get the code please contact me

backpropagation network number plate


For the number of input neuron, output neuron and hidden layer, you need to test it out. Test by using different combination of architecture and find out the most suitable combination.

There is no fixed number of neuron or hidden layer for a particular network. Different application may need different parameter.

matrix to vector matlab neural network nntool

hi leekk1

i have the input vector as my original signal, but it forms very long vector, when i call it in NN tools in matlab, which gives EEror, so how i take the small data .

Q1. Is there any method to represent our data upto some samples in vector form.If yes, then please wirte there step how to represent.

Q2. i donot know my target vector? what should i have to give in target vector.

best regards

nntool for image

Hi, vjfaisal

How long is the vector of yours? I never try the NNtool in matlab, but I use 1x3600 matrix as input.

1. As I do not know what is your input, I have no idea how to represent it.

2. Target vector is the result you expect.
I'm not good in your brain state recognition, but just take alphabets recognition as example:
You can represent "A" with 0001, "B" with 0010, and so forth. So, in this case, your output neuron will be 4 neurons, and your target output for input "A" will be 0001.

nntool character recognition

hi leekk1

i have the input matrix of 1*2750. which is smaller than ,what u have used.I have applied this matrix os 1*2750 but the NN tool give an eeror ,in Row adjustment....

please tell how you use the 1*3600 matrix as an input , if you have done in matlab Editor.........please show me how you use it.....


Not open for further replies.

Part and Inventory Search

Welcome to