Automatic voice recognizing method and system

Abstract

An automatic speech recognition method includes at a computer having one or more processors and a memory for storing one or more programs to be executed by the processors, obtaining a plurality of speech corpus categories through classifying and calculating raw speech corpus (801); obtaining a plurality of classified language models that respectively correspond to the plurality of speech corpus categories through language model training applied on each speech corpus category (802); obtaining an interpolation language model through implementing a weighted interpolation on each classified language model and merging the interpolated plurality of classified language models (803); constructing a decoding resource in accordance with an acoustic model and the interpolation language model (804); decoding input speech using the decoding resource, and outputting a character string with a highest probability as the recognition result of the input speech (805).

Description

Technical field [0001] This application relates to the technical field of Automatic Speech Recognition (ASR), and in particular to an automatic speech recognition method and system. Background technique [0002] Automatic speech recognition technology is a technology that converts vocabulary content in human speech into computer-readable input characters. Speech recognition has a complicated processing flow, which mainly includes four processes: acoustic model training, language model training, decoding resource construction, and decoding. figure 1 It is a schematic diagram of a main processing flow of an existing automatic speech recognition system. See figure 1 , The main process includes: [0003] In steps 101 and 102, it is necessary to perform acoustic model training based on acoustic raw materials to obtain an acoustic model, and perform language model training based on raw corpus to obtain a language model. [0004] The acoustic model is one of the most important parts of th...

AI Technical Summary

Problems solved by technology

[0009] However, most of the existing speech recognition technologies are based on universal speech recognition applications, that is, to build models for commonly used speech recognition. In this case, the training corpus of the language model is mainly based on data collection and actual user input. Although from To some extent, it better reflects the user's speaking habits, and it often has a better recognition effect on daily expressions; however, since there are few rare words in the training corpus of the language model, such as medical names and place names, it cannot form effective words. The probability and statistics model of the language model, the probability value of the string corresponding to the uncommon word in the language model is very low, so when it is necessary to identify the relatively uncommon words spoken by the user, the data offset problem often occurs, that is, the recognized string is not the user Spoken words, that is to say, the recognition accuracy of the speech of uncommon words is low, and it is difficult to obtain better recognition results

Images