在連續(xù)語音識別中，N-best 有很多應(yīng)用。本論文分析了常用的N-best 搜索算法，如
并行算法，串行算法，以及Tree-Trellis 搜索在搜索過程中的問題后，指出Tree-Trellis 搜索的有效性，并且針對Tree-Trellis 搜索后向得分計算復(fù)雜的問題，提出了一種實現(xiàn)方便的方法，從而快速有效的輸出N 個有效路徑。
關(guān)鍵詞： A*算法 N-best Tree-Trellis

A Efficient Tree-Trellis N-best Algorithm And Its Application In Chinese Speech Recognition
Abstract N-best search is of vital importance in large vocabulary continuous speech recognition. This paper first gives a deep insight of the traditional N-best search algorithm, such as parallel algorithm, sequential algorithm and Tree-Trellis algorithm, and point out the efficiency of Tree-Trellis algorithm, then gives out a more efficient method based on Tree-Trellis algorithm, finally gives out N best path efficiently.
keywords: A* algorithm N-best Tree-Trellis