Alignment system of on-line speech text and method thereof

Abstract

The invention relates to an alignment system of an on-line speech text and a method thereof. The system comprises a text processing module, an error detection module, an error recovery module and a compulsory alignment module. The compulsory alignment module comprises a characteristics extraction module, a search space construction module and an alignment decoding module. The error recovery module comprises a language model estimation module, a language model interpolation module, a speech recognition module and a text alignment and similarity calculation module. Method for the system and the method to detect end of a sentence is an improvement of a conventional method based on Viterbi alignment, information of a search space by beam search is used for estimating activity degree A (t, somegae) of the search space at the end of the sentence, and estimating sentence end time in partial meaning. The system and the method have the function for automatically detecting and jumping over unmatched segment errors in a text and a speech; can generate online input speech current and corresponding text alignment result at real time and can process a long text with errors.

Description

technical field [0001] The invention relates to the field of television subtitle display, in particular to an online voice-text alignment system and method. Background technique [0002] The proportion of subtitled programs in a country reflects the level of humanity of a country and the degree of social concern for the disabled. At present, TV programs in many countries such as Japan, the United States, and the United Kingdom have been added with subtitles. However, there are very few domestic programs with subtitles added. Even if there is a small amount of subtitle addition, it is limited to recorded programs, and the addition of subtitles is done manually by professionals, which takes a lot of time and energy. The output speed is added one by one. [0003] The core module of the system of the online speech-to-text method in the prior art is an alignment module based on a hidden Markov model. Its main function is to generate the corresponding real-time time correspond...

AI Technical Summary

Problems solved by technology

The traditional alignment method is to obtain all the voices at the same time, because there may be inaccuracies in the aligned text, in the news subtitles, it mainly shows that there is no corresponding text for news interviews (simultaneous sound) of some on-site news

Traditional alignment methods cannot handle these errors online

Traditional voice-to-text alignment methods, in order to deal with erroneous segments in the text stream, are usually completed offline after all the voices are acquired, so it is impossible to take into account the processing of erroneous text and the real-time acquisition of real-time voice input alignment results

Images