Levinson-Durbin recursion


a = levinson(r)
a = levinson(r,n)
[a,e] levinson(r,n)
[a,e,k] levinson(r,n)


The Levinson-Durbin recursion is an algorithm for finding an all-pole IIR filter with a prescribed deterministic autocorrelation sequence. It has applications in filter design, coding, and spectral estimation. The filter that levinson produces is minimum phase.

a = levinson(r) finds the coefficients of a length(r)-1 order autoregressive linear process which has r as its autocorrelation sequence. r is a real or complex deterministic autocorrelation sequence. If r is a matrix, levinson finds the coefficients for each column of r and returns them in the rows of a. n=length(r)-1 is the default order of the denominator polynomial A(z); that is, a = [1 a(2) ... a(n+1)]. The filter coefficients are ordered in descending powers of z–1.


a = levinson(r,n) returns the coefficients for an autoregressive model of order n.

[a,e] levinson(r,n) returns the prediction error, e, of order n.

[a,e,k] levinson(r,n) returns the reflection coefficients k as a column vector of length n.

    Note   k is computed internally while computing the a coefficients, so returning k simultaneously is more efficient than converting a to k with tf2latc.

More About

expand all


levinson solves the symmetric Toeplitz system of linear equations

[r(1)r(2)*r(n)*r(2)r(1)r(n1)*r(n)r(2)r(1)] [a(2)a(3)a(n+1)]=[r(2)r(3)r(n+1)]

where r = [r(1)  ... r(n+1)] is the input autocorrelation vector, and r(i)* denotes the complex conjugate of r(i). The input r is typically a vector of autocorrelation coefficients where lag 0 is the first element r(1). The algorithm requires O(n2) flops and is thus much more efficient than the MATLAB® \ command for large n. However, the levinson function uses \ for low orders to provide the fastest possible execution.


[1] Ljung, L., System Identification: Theory for the User, Prentice-Hall, 1987, pp. 278-280.

See Also

| | | |

Introduced before R2006a

Was this topic helpful?