How to normalize a discrete signal?

Hello!!

I want to calculate the signal energy in rolling windows, after first normalizing to interval [-1,1].

Firstly, I wrote the below code to read and sound the signal.

[y,Fs] = audioread('viola_series.wav');
plot(y);
title('Audio viola series.wav');
sound(y,Fs);

How to do the normalization?

Thanks in advance

Answer 1

[Adam Danz]

Audio data are typically already normalized between [-1,1]. Are you getting values outside of that range?

min(y)
max(y)

Anyway, this is how to normalize any vector to [-1,1];

ynorm = (y-min(y))/range(y)*2-1;

[RoBoTBoY]

So that is the initial signal

And that is the normalized signal

How could I convert the x axis from 0 to 10 sec for both signals?

[Adam Danz]

Data from the first plot could also be normalized but without values larger than +/- ~0.025.

Data from the second plot are also normalized and consume the full normalized range.

> How could I convert the x axis from 0 to 10 sec for both signals?

You can use the same approach in my answer except multiply by 10 instead of 2 and no need to subtract 1.

[RoBoTBoY]

I could not do it. But it does not matter.

Now I want to find the energy of this signal in rolling windows.

Note that the energy of a signal x[n] overlapping by a window w[n] is given by the following formula:

where as signal w[n] I will use the Hamming window given by the equation:

I will use window length N = 1000 samples.

I run the below code but without result.

syms m
w = hamming(1001);
energy_signal = symsum(conv(y_norm.^2,w),m,0,10);
plot(energy_signal);

How to do that?

[Adam Danz]

The normalization to [0,10] would have been

n = (x-min(x))/range(x)*10;

[Adam Danz]

Since you're new question differs greatly from the original one, it would be better to post it as a new question.