Mathematical Treatment of Uncertainty in the Speech Recognition Process

摘要

One of the main difficulties in the speech recognition process is the treatment of the imprecisions around it. They have origin in the differences between the articulatory system of each person and the physical properties of the sound propagation. Moreover, the circuits involved in the sound storing and analysis works with a degree of uncertainty and also adds some imprecision in the process. This paper discuss the applications of mathematical methods used to treat uncertainty in the computational speech recognition and the possibility of using interval analysis for this propose.