Comments
Click the 'Add' link to add a comment to this page; click the 'Read More' link to view replies to a posted comment.
Julius is an Open-Source, high-performance, two-pass Large Vocabulary continuous Speech Recognition (LVCSR) software decoder. Based on word 3-gram and context-dependent Hidden Markov Models ("HMM"), according to its web site, it can perform almost real-time decoding on most current PCs in a 20k word dictation task. It incorporates major search techniques such as tree lexicon, N-gram factoring, cross-word context dependency handling, enveloped beam search, Gaussian pruning, Gaussian selection, etc. It is also modularized to be independent from model structures.
Various HMM types are supported such as shared-state triphones and tied-mixture models, with any number of mixtures, states, or phones. It uses standard formats.
It's main platform is Linux and it also works on Windows. Julius is distributed with open license together with source code.
There are two versions of Julius called "normal" and "multi-path". The differences are:
The "normal" version realizes faster decoding using a simpler HMM handling algorithm. However, it has several limitations in state transition that it allows only one transition from start node and to end node within a model. Therefore, the "normal" version does not support a multi-path model that has several phoneme models in parallel, or a bi-pass transition that has direct transition from start node to end node.
The "multi-path" version can handle ANY arbitrary transition,
but has a decoding speed that is slightly slower than the normal version.
The amount of delay is dependent on the complexity of the acoustic
model.
Julian is a grammar-based recognition parser which is integrated into Julius. Julian is a modified version of Julius that uses hand-designed DFA grammar as a language model. It can be used to build a small vocabulary voice command system or various spoken dialogue tasks.
The Obelisk Spoken Dialog System uses Julian for Speech Recognition.
Julian uses acoustic models in HTK format. Julian is language-independent, and can be used with other languages if the grammar and an acoustic model are prepared for the target language.
For further details, see the Julius web site.