Speech compression algorithm based on non-equidistant sampling

Abstract

The method presented here is one of the methods of time domain compression. The technique uses non-equidistant sampling. Human voice and voice conversation are characterized by transitions. During the speech frequency charactericstics change so that viariable sampling can be applied. The nonuniform sampling method which is implemented is described in this paper; its usage in speech synthesis software which is being developed at the Department of Info-Com Networks is shown.