Speech recognition system that selects a probable recognition resulting candidate
View Patent ↗A system has a speech recognition unit 2 including a first recognition engine capable of recognizing predetermined word strings and a second recognition engine capable of recognizing parts of the word strings; and a deciding unit 4 that selects, in accordance with a difference between the durations of first and second recognition resulting candidates, which are recognition results of the speech recognition by the first and second recognition engines 2 a and 2 b , one or more probable recognition resulting candidates from the first and second recognition resulting candidates.
1. A speech recognition system comprising:
an interface that acquires speech inputted via a microphone;
a speech recognizer comprising at least a first recognition engine and a second recognition engine that perform recognition of the acquired speech, wherein the first recognition engine is configured to recognize at least one word string partially comprised of a word string recognizable by the second recognition engine; and
a processor programmed to execute a process for determining a recognized speech that includes
obtaining a list of recognition candidates that are recognized by the first and second recognition engines as a result of performing recognition on the acquired speech, said recognition candidates including a first recognition result recognized by the first recognition engine and a second recognition result recognized by the second recognition engine, recognize,
calculating a difference in duration between the first and second recognition results by subtracting the duration of the second recognition result from the duration of the first recognition result,
comparing the difference in duration to a first threshold set in advance,
when the difference in duration meets or exceeds the first threshold, updating the list of recognition candidates by discarding the second recognition result, and selecting the recognized speech from the updated list of first recognition candidates.
2. The speech recognition system according to claim 1 , wherein
if the difference in duration is less than the first threshold, the process reduces, in accordance with the difference in duration, a reliability score of the second recognition result.
3. The speech recognition system according to claim 1 , wherein the decider processor alters the first threshold set in advance in accordance with a category of a recognition target word string.
4. The speech recognition system according to claim 1 , in which an acquired speech duration from beginning to end of the acquired speech is detected, wherein
the process further comprises discarding, from the list of recognition candidates, a recognition resulting candidate whose duration is not within a second threshold amount of the acquired speech duration, the second threshold being set in advance.
5. The speech recognition system according to claim 4 , wherein
the process reduces, in accordance with the respective differences between the durations of the non-discarded recognition candidates in the list and the acquired speech duration, reliability scores of the respective non-discarded recognition candidates in the list.
6. The speech recognition system according to claim 4 , wherein the acquired speech duration is determined in accordance with the category of a recognition target word string and a threshold of end detection of the acquired speech.