A subtitle voice precise synchronization system and method, information data processing terminal

Abstract

The invention belongs to the technical field of computer software and discloses a subtitle voice precise synchronizing system and method, and an information data processing terminal. A machine identifying module applies various technologies to improve the anti-noise property, and a Wiener filter technology is adopted for two times to eliminate background noise; waste voice is removed through a Gaussian modeling method, and the correct rate of identifying 36 kinds of sounds in the natural environment through a GMM reaches 95.83%; and the voice starting point is effectively detected through a harmonic detection technology, compared with a traditional voice identification method, the identification speed is increased by 100% on the basis of the prior art, and meanwhile the identification accuracy rate is effectively increased by two times or above. According to the subtitle voice precise synchronizing system and method, and the information data processing terminal, source text analysis, source text and target text conversion and target text generation are separated, an independent analysis and independent generation system is built; and in the independent analysis and independent generation system, the features of a target language are not considered when a source language is analyzed, the features of the source language are not considered either when the target language is generated, and the difference of the source language and the target language is solved through source text and target text conversion.

Description

technical field [0001] The invention belongs to the technical field of computer software, and in particular relates to a subtitle voice precise synchronization system and method, and an information data processing terminal. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: In many fields of social life, such as voice dialing system, bank inquiry system, telephone booking system, information retrieval and translation system, teaching activities, etc., there are applications of speech recognition technology, which are based on isolated words or small vocabulary. The speech recognition system for a specific person has a recognition accuracy greater than 98%, and is widely recognized by people. However, with the development of Internet technology, video has become a large amount of traffic on the Internet. In recent years, live video broadcasts have "ravaged" the global network. More and more people pay attention t...

AI Technical Summary

Problems solved by technology

The current large-vocabulary continuous speech recognition system still does not meet the practical and extensive requirements, especially in areas with large video traffic such as TV, movies, and live reports.

The main reason for this phenomenon is the technical bottleneck of speech recognition

Speech recognition mainly has the following problems: (1) Speech segmentation, the first step of speech recognition must decompose continuous speech into units such as phonemes or consonants, and then need to establish a rule to understand semantics

(4) Noise influence, when the environment is heavy with noise and interference, the correct rate of speech recognition will drop

[0003] In summary, the problems in the prior art are: Human translation is inefficient and expensive

[0005] For Chinese speech recognition, words, syllables, and consonants can be selected as speech recognition units. The smaller the selected primitive, the higher the flexibility of recognition, but the stability is reduced, and vice versa.

Images