Search and learning strategies for impro
β
Renato De Mori; Michael Galler; Fabio Brugnara
π
Article
π
1995
π
Elsevier Science
π
English
β 121 KB
A speaker-independent automatic speech recognition system is developed using hidden Markov models (HMMs). Simulated annealing and randomized search are used to optimize discrete features of the system, including topologies, parameter ties, context clusters, and the sizes of mixture densities. Domain