why does the aasamplebiasedautoc is giving abnormal results with a noise signal?

Greetings everyone

In order to find the eigenvalues of correlation matrix Rx from an input signal x(n), i used the Book

m-function: rx = aasamplebiasedautoc(x,M) with M is the digital filter order. the function gives good results with both sinus and random signal (randn) data, but when i applied it to a noise recording signal by using the function : audioread(....), it starts giving abnormal results. i respected all the vectors' dimensions but sill can't find the problem.

here is the of the Book m-function i used :

function[r]=aasamplebiasedautoc(x,lg)
%function[r]=aasamplebiasedautoc(x,lg);
%this function finds the biased autocorrelation function
%with lag from 0 to lg;it is recommended that lg is 20-30% of 
%N;
N=length(x);%x=data;lg=lag;
for m=1:lg
    for n=1:N+1-m
        xs(m,n)=x(n-1+m);
    end;
end;
 r1=xs*x'
sz3 = size(r1)
r=r1'./N;

the size of the data vector is 2 589251

thanks for your attention

Answer 1

[Walter Roberson]

That code expects a vector of input as the first parameter. You are sending in a 2D array. The way your data is arranged, the code is going to take samples alternating between row 1 and row 2 -- row1column1 row2column1 row1column2 row2column2 row1column3 row2column3 and so on.

If you were to send in a 589251 x 2 array, where the number of rows exceeded the number of columns, then the code would effectively only work on the first column -- but when the number of columns exceeds the number of rows, the samples are going to be taken alternately.

[Amira]

Thank you sir for the answer.

i changed the code so that the input signal would be 589251 x 2 and i got an error regarding the Inner matrix dimensions in instruction :

r1=xs*x'

the dimensions of r1 with sinus sinal were M x 1 so i think i should change the matrix product so i obtain the size of r1 = M x 2 knowing that:

size(xs) = 40 x 589251

size(x') = 2 x 589251

i changed the product to : r1=xs*x and i got these results :

Dx =

-0.0058

-0.0028

-0.0008

-0.0004

-0.0002

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

-0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001

0.0002

0.0002

0.0002

0.0004

0.0006

0.0014

0.0025

0.0038

0.0062

0.0203

0.0332

wich i believe are incorrect.

i think the main problem is that my data are received in an array with 2 clomns (589251 x 2), but it's the case with all recordings using the audioread function. is there any possibility to receive my input data in a vector !

thanks for your attention.

[Walter Roberson]

That code is not designed to handle 2D array of input.

You have three choices:

1. Process one column of the audioread() output at a time. Each column will be 589251 x 1. The overall output will not relate the channels to each other.
2. Recode the function so that it accepts N x 2 arrays but processes the columns independently. The overall output will not relate the channels to each other.
3. Recode the function so that it accepts N x 2 arrays and processes the columns together, giving an overall output that does relate the channels to each other. I am not sure what the expression to relate the two channels to each other would be.

[Amira]

thank you sir for the sugestions.

i have another question related to the function audioread(), if i use this function to read a recording

[x, fs] = audioread('recording.wav')

and then plot the signal using

plot([1:length(x)],x,'b')

the signal x is a 589251 x 2 array.

My qustion is how does MATLAB plot the signal ? by using the first column ! or the second !

here is a sample of the x signal received

-0,00894165039062500	-0,000915527343750000
-0,00827026367187500	0,000183105468750000
-0,00598144531250000	0,00180053710937500
-0,00466918945312500	0,00256347656250000
-0,00277709960937500	0,00225830078125000
-0,00109863281250000	0,00369262695312500
0,000976562500000000	0,00628662109375000

thank you

[Walter Roberson]

plot([1:length(x)],x,'b')

will use 1:length(x) as the independent coordinate of the plot.

It will notice that this is a vector, and will match the length of the vector to the size() of the 2D array that is the dependent coordinate. First it will test the length against the size of the first dimension of the dependent coordinate. As it will find a match in sizes, it will plot one line for each column in the dependent coordinate. The result will be equivalent to

plot(1:length(x), x(:,1), 'b', 1:length(x), x(:,2), 'b')

If it detected a mismatch of sizes of the first dimension of the dependent coordinate, it would check the second dimension of the dependent coordinate; if it found a match, it would plot one line for every row of the dependent coordinate.

If the length of the independent coordinate does not match the first or second dimension of the dependent coordinate, then it would give up with a size error.

[Amira]

Thank you sir.

i figured the problem. it was related to the fact that my audios are in dual channel (stereo) and not single channel.

thanks for your help