Title | ||
---|---|---|
Time and memory efficient viterbi decoding for LVCSR using a precompiled search network |
Abstract | ||
---|---|---|
Abstract: In this paper, we present our recently developed timesynchronousspeech recognition decoder, which adopts theidea of representing the search space of Large VocabularyContinuous Speech Recognition (LVCSR) in a singleprecompiled network. In particular, we outline our approachesfor time and memory efficient Viterbi decodingin this scenario. This includes reducing the fast memoryneeds by keeping the search network on disk and only loadingthe required parts on demand. Evaluations are carried... |
Year | Venue | Keywords |
---|---|---|
2001 | INTERSPEECH | viterbi decoder,search space,speech recognition |
Field | DocType | Citations |
Sequential decoding,Pattern recognition,Computer science,Soft output Viterbi algorithm,Speech recognition,Viterbi decoder,Artificial intelligence,Iterative Viterbi decoding,Viterbi algorithm | Conference | 5 |
PageRank | References | Authors |
1.04 | 4 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Daniel Willett | 1 | 58 | 6.76 |
E. McDermott | 2 | 514 | 88.33 |
Yasuhiro Minami | 3 | 5 | 1.04 |
Shigeru Katagiri | 4 | 850 | 114.01 |