what type of input is given for ANN (neural network)?

i have audio files of 3 different types and want to classify them. planning to use neural network for this which will automatically find my test data falls into which category. input to this network is in what form? my audio files are in .wav format and stored as 45879x1 matrix. does this can be given as input? or should be converted to zeros and ones (binary)? if yes how? another option is, i have extracted mel frequency cepstral coefficients(MFCCs) which is a feature of signal. if i want to use these as inputs how to use?because these are combination of positive and negative numbers.please help.

[Foresight india]

Hi,

 Please let us know. We are very new in ANN and would like to start development in NN , How to training my computer to answer in my voice.

Answer 1

[Stalin Samuel]

we can divide the neural network classification into two parts.before going to neural network we need to Extract some features(Ex mean,STD,Entropy....).That will be given to the neural network as a input and the category (1 0r 2 or 3) will be the target.

NN example

[Greg Heath]

Incorrect. See my answer.

[pranjal]

Thank you sir @Stalin Samuel. can u pls help me to find variance of audio file in .wav format?

Answer 2

[Greg Heath]

The MFCCs should be the inputs. For c classes the corresponding target columns should be columns of the unit matrix eye(c). The relationship between the target and trueclass indices are given by

 truclassindices = vec2ind(target)
 target          = ind2vec(truclassindices)
 [ O N ]         = size(target) % [ c N ] (c=3)

The relationship between the output and assigned class indices

 assignedclassindices = vec2ind(output)

yielding

 err = assignedclassindices~=truclassindices

 Nerr = sum(err)

 PctErr = 100*Nerr/N

The trn/val/tst indices for each class are in the training record tr. Therefore, you can obtain the 3 trn/val/tst error rates for each of the 3 classes.

Hope this helps.

Thank you for formally accepting my answer

Greg

[sai susmitha]

my doubt is when we extract mfcc's for a wave(audio) file then it gives a matrix of [no.of frames x 13]if each frame has 13 mfcc's now if to train a classifier using these(here Ann) then what is the input of BPNN classifier should i convert this matrix(what i got after extraction) into 1 x 13 then for each wav file we finally have only this one representation(1x13)and for training the target vector format would be [1 0 0] if i am classifying it into 3 classes each wav file i put target vector as 1 in the desired class and 0 in others if input is not 1x13 i.e (no.of framesx13) then what is target vector format