CONVERGENCE ANALYSIS OF A BACKWARD ADAPTATION MODEL
Jey-Hsin Yao, John J. Shynk · 2005
Backward adaptation algorithms, often used in speech coding applications, are difficult to analyze because of their inherent IIR filter structure. In this paper, we study the convergence proper- ties of the linear predictor coefficients using a backward-adaptation scheme based on a recursive form of the LMS algorithm. For three identification configurations and with some commonly-used assump- tions, we derive a set of detenninistic equations that model the avir- age behavior of the adaptive filter coefficients. Computer simula- tions show good agreement between the average performance of the algorithm and our analytical results.