Language distance relation obtaining method based on language identification system

Abstract

The invention discloses a language distance relation obtaining method based on a language identification system. The method comprises a language identification process and a language relation graph generation process. The language identification process refers to a process in which a computer, according to a segment of speech given by an unknown person, identifies the language of the speech. The language relation graph generation process refers to a process in which a distance between languages is determined according to an identification rate between the languages and a language relation graph is finally generated. According to the invention, a new speech feature set is applied to the language identification system, the language identification system is erected through a GentleAdaBoost algorithm, and a distance relation between the languages is studied by use of the language identification rate output by the language identification system, such that the language identification rate is improved effectively, a result consistent with reality is achieved, and a new thinking mode is provided for search on the language distance relation.

Description

technical field [0001] The invention relates to the fields of speech signal processing and pattern recognition, in particular to a method for acquiring language distance relations based on a language recognition system. Background technique [0002] The research on language recognition began in the 1970s. It is a process in which a machine recognizes the language type of a speech spoken by an unknown speaker based on the acoustic signal of a speech. With the current increase in exchanges between countries around the world, the demand for communication between various languages ​​has increased, which poses new challenges to language recognition. Before a machine can understand the meaning of speech, it must identify which language is used. Different from speech recognition and speaker recognition, language recognition uses the linguistic information in the speech signal without considering the meaning of the words in the speech or the personality of the speaker. Language rec...

AI Technical Summary

Problems solved by technology

The phoneme structure method is robust to channel and noise, but there are obvious defects: professional linguistic knowledge is required to establish a suitable phoneme set of various language characteristics; a large amount of artificially marked corpus is required for training Phoneme recognizer; huge amount of calculation is not suitable for real-time system development; poor system scalability

However, these methods require the support of more linguistic knowledge and are very complex to implement.

Images