ABSTRACT
We then presented the mathematical formulation of ASR, and in particular, the fundamental equation of ASR based on Bayes' decision theory and the MAP decision rule. This fundamental equation breaks down the ASR problem into three sub-problems: acoustic modeling, language modeling, and sequential optimization or decoding. In the remaining sections of this chapter, acoustic modeling was the main focus of coverage, with the final section dedicated to language modeling. No special attempts were made to cover decoding.
