Partial traceback and dynamic programming

摘要: Dynamic programming is used in speech recognition to search efficiently for word sequences whose templates best match acoustic data. The constrained by finite-state networks embodying grammatical rules. Typically, dynamic implemented two steps: the first calculates, each state a network and time, way of arriving at that time; second traces back from final time initial determine path through network. This step cannot be initiated before determination (usually detection silence) has been reached. Such difficult truly continuous speech; there are often no reliable anchor points. Further, it desirable able recognize least part an utterance speaker stopped talking. In this paper we introduce technique discovering section optimal traversal complete. It can report system's interpretation data not-too-distant past without relying on or making any decisions which may degrade accuracy.