Linear prediction of speech with a least absolute error criterion

Most methods for linear prediction of speech are based on a least square error criterion (L
2
criterion). In this paper, it is shown that the robustness of a least absolute error criterion (L
1
criterion) is particularly well adapted to the analysis of voiced sounds. The well-known linear programming L
1
algorithms have two important drawbacks: the stability of the synthesis filter is not guaranteed, and the computation load is too heavy for real time analysis. All these problems are alleviated by the new L
1
algorithm proposed in this paper. Experiments show comparable efficiency with classical L
2
methods.