Method and apparatus for the derivation of at least one sequence of words from a speech signal

摘要

The method involves deviating test signals from a speech signals and comparing the test signals with a reference signal for determining weighted words. The words are used to form a word graph with nodes and edges (K). The words on a first path through the graph having the smallest sum of weighted values is output as a sequence of words. The sequence is stored as a base of a tree used for further outputs of sequences with the next highest sum. The tree includes tree nodes (T) for each of which the overall sum of sectional paths is calculated which run towards the corresponding nodes of the word graph. The tree also includes an end node. The sequence of words is output which has the smallest sum of weighted values on a sectional path to the tree node corresponding word graph node and from the tree node to the end node in the tree. The output sequence of word is added to the already stored tree.