Method of converting phonetic file into text file

Abstract

The process of converting phonetic file into text file includes the following steps: to obtain the phonetic file to the be converted and to play back in player the phonetic file including at least 50characters; the user to identify and input corresponding text to obtain phonetic file for training and corresponding text file; to re-estimate phonetic parameters by means of phonetic adapting technology with the obtained training file and basic phonetic identifier with phonetic identifying library; and to obtain phonetic identifier related to the recorder and convert the identified phonetic fileinto text file.

Description

technical field [0001] The invention relates to a speech recognition method, in particular to an adaptive speech recognition method which can directly process and recognize speech files and convert them into text files. Background technique [0002] The wide application of computers has promoted the progress of speech recognition research, especially in the past two decades, with the introduction and gradual engineering of the Hidden Markov Model (HMM) theory, researchers have used the Hidden Markov Model to establish Some of the speech recognition systems in China have entered commercial applications. As a speech recognition system for commercial applications, it usually includes a basic speech recognizer that has nothing to do with the speaker. Since the pronunciation of different users is quite different, the basic speech recognizer must be trained, using speech adaptive technology Re-estimate the language parameters of a specific user to obtain a...

AI Technical Summary

Problems solved by technology

This method is usually used in the field of speech input. When used as an input method, the user can train the speech recognizer in advance. etc.), it is unrealistic to ask the speaker to train the speech recognizer. Moreover, due to the large number of dialects in Chinese, even if the speaker speaks in Mandarin, he often has a heavy local accent, which makes it difficult for those without The standard speech recognizer trained is difficult to recognize accurately; at the same time, even if the speech recognizer adopts a certain dialect, due to the great regional differences of Chinese dialects (for example, taking southern Jiangsu as an example, not only the neighboring cities of Suzhou and Wuxi The accents are different, and the accents of Suzhou City and its subordinate county-level cities are also different, even in Wuzhong District of Suzhou, there are many dialects), and untrained dialect speech recognizers cannot accurately recognize the dialects of neighboring regions

[0005] Therefore, with the existing speech recognizer training method, it is impossible to realize the recognition and conversion of speech files. Even if the standard recognizer that comes with it is barely used, the recognition rate is also very low, which cannot meet the practical requirements.

Images