An Event-Space Feedforward Network Using Maximum Entropy Partitioning with Application to Low Level Speech Data
Autor: | David Chiu, D. Bockus, J. Bradford |
---|---|
Rok vydání: | 1997 |
Předmět: | |
Zdroj: | Mathematics of Neural Networks ISBN: 9781461377948 |
DOI: | 10.1007/978-1-4615-6099-9_22 |
Popis: | This paper describes an event-space feedforward network based on partitioning of the input space using maximum entropy criterion. It shows how primitives defined as partitioned hypercells (event space) can be selected for the purpose of class discrimination. Class discrimination of a hypercell is evaluated statistically. Observed primitives corresponding to observed characteristics in selected hypercells are used as inputs to a feedforward network in classification. Preliminary experimental results using simulated data and as it pertains to speaker discrimination using low-level speech data have shown very good classification rates. |
Databáze: | OpenAIRE |
Externí odkaz: |