Speech keyword recognition method, device, equipment and computer-readable storage medium

Abstract

A voice keyword identification method, apparatus and device, and a computer readable storage medium. The voice keyword identification method comprises: receiving an input voice signal (101); extracting audio features in the voice signal (102); calculating, according to the audio features, the probabilities of keywords to an acoustics model, a pronunciation dictionary and a language model by utilizing the acoustics model, the pronunciation dictionary and the language model; determining whether a probability is greater than a threshold; if a probability is greater than the threshold, counting the number of the keywords corresponding to the probability; and if the number of the keywords corresponding to the probability is one, using the keyword corresponding to the probability as a keyword identification result. After the probabilities of possible keywords are calculated, a keyword corresponding to the probability greater than the threshold is used as the keyword identification result, thus improving keyword identification rate.

Description

technical field [0001] The present invention relates to the technical field of information processing, in particular to a voice keyword recognition method, device, equipment and computer-readable storage medium. Background technique [0002] A general large vocabulary continuous speech recognition system (Large Vocabulary Continuous SpeechRecognition, LVCSR) speech keyword recognition process is: first recognize the speech, obtain one or more possible sentence texts, and then search for keywords in the text . If the keyword appears in a certain text, it is considered that there is a keyword in the sentence. The language model used by a general LVCSR system will try to capture the characteristics of natural language, such as the relationship between contexts in natural language, so as to identify sentences that are more likely to appear in the language. In the task of keyword recognition, often because there are many possible sentences with similar pronunciation, even if ke...

AI Technical Summary

Problems solved by technology

In the task of keyword recognition, often because there are many possible sentences with similar pronunciation, even if keywords appear in the sentence, the most likely one or more recognition results do not include keywords (because it takes into account The relationship between the context in the language), especially when the keyword itself has a low probability of appearing in the language, resulting in a low recognition rate

Images