End-to-end automatic voice recognition method for field of civil aviation air-ground calls

Abstract

The invention discloses an end-to-end automatic voice recognition method for the field of civil aviation air-ground calls. The end-to-end voice recognition method comprises the following steps of taking an end-to-end model as a voice recognition method of an acoustic model, adopting an end-to-end model training method, and applying a voice recognition method based on an end-to-end model in the field of civil aviation air-ground calls. The training method of the end-to-end model comprises an end-to-end model training method taking maximum correlation mutual information as a target function, a word lattice-free training method in the training process and a special end-to-end neural network structure. In addition, the invention discloses a special data processing and training method and a special application method of the end-to-end model in deep learning in the field of civil aviation air-ground calls. Compared with the time required by a conventional model, the time required by the training of the end-to-end model is greatly shortened, and the performance of a sample and a machine is greatly improved. And compared with a conventional model, the recognition effect is also improved.

Description

technical field [0001] The present invention relates to the fields of intelligent communication, signal processing and deep learning, in particular to an end-to-end voice recognition method and system for the field of civil aviation, land and air communication. Background technique [0002] With the rapid expansion of civil aviation passenger transport and cargo business, airport aircraft activities are very frequent, which has brought great work pressure to airport air traffic controllers. Therefore, we need to use intelligent machines to assist controllers in air traffic control, so as to reduce the work pressure of air controllers and better ensure airport safety and order. The premise of all intelligent assistance is the intelligent interaction between man and machine, and the fastest interaction is undoubtedly oral communication. Therefore, the function of an end-to-end automatic speech recognition method in the field of civil aviation, land and air calls in the presen...

AI Technical Summary

Problems solved by technology

Such a speech recognition model training process is undoubtedly cumbersome and time-consuming. Later, with the continuous development of deep learning to end-to-end models, more and more researchers have succeeded in the fields of image recognition, natural language processing and speech recognition. end-to-end model

Images