Iteration independent subroutine form of Durbin´s recursion for programmable signal processors

The method of Durbin´s recursion to transform a signal autocorrelation vector to a representation based on linear prediction coding (LPC) is a computation that is sufficiently irregular in structure to present difficulties in implementation on present single chip digital signal processors. During the computation, operands must be accessed from three one-dimensional vectors and a two-dimensional array, which may exceed the limited address arithmetic capability of a digital signal processor. As a result, signal processor code for each iteration of Durbin´s recursion is often strung out sequentially, consuming excess program memory, rather than written as a single subroutine executed repetitively for each iteration, as is common in high level programming languages. A new form of Durbin´s recursion is described that renders all addressing to be sequential within one iteration of the recursion. Using this technique, Durbin´s recursion may be cast into a single repetitively-called subroutine with sufficiently simple address arithmetic for signal processors. The technique may be used to compute any of the standard LPC parameters (LPC coefficients, reflection coefficients, prediction residual energy) and may be extended to any LPC model order.