3. Coarticulation and nasalization
One of the most formidable challenges in speech recognition, whether by humans or machines, is dealing with variability in the speech code. Different instances of the “same” sound can contain enormous variation, yet listeners are for the most part unconscious of this variability. Some of this variability stems from differences in speakers, but even sounds produced by the same speaker may demonstrate significant acoustic dissimilarity due to the phonetic context in which the sound is uttered.