Design And Implementation Of Chinese Speech Recognition System Based On End-to-end

With the rapid development of artificial intelligence,intelligent scenarios take up an increasing proportion of people's daily work and life,and the requirement for barrier-free language communication between humans and machines has become more and more urgent.As one of the key technologies to realize human-machine language communication,speech recognition technology has always been a hot research field of artificial intelligence,and its technological development and innovation have always attracted attention.Compared with traditional speech recognition methods,the end-to-end speech recognition technology based on deep learning realizes the direct mapping from input acoustic features to word sequences,making the process of constructing a speech recognition system more concise.In recent years,there have been more and more voice recognition products on the market,among which end-to-end voice products are gradually increasing.According to the in-depth analysis of requirements,a design plan based on end-to-end Chinese speech recognition was determined,the system was divided into four major modules,and each module was developed and deployed.The core module in the whole system is the speech recognition module.This module uses the transfer learning method to train the CTC-based deep convolutional neural network,and obtains the encoding and decoding model of Chinese speech recognition,combined with the 4-gram language The model completes the construction of the speech recognition module.After that,using Flask application framework,front-end technology,SQLite database and other technologies,combined with system requirements and module function design,the complete development of the Chinese speech recognition system was realized.Finally,the speech recognition system was tested comprehensively to verify the integrity of the system's functions.In addition,the performance of the end-to-end speech recognition model in actual scenarios is evaluated,and the evaluation results are analyzed in detail.The Chinese speech recognition system in this paper meets two major design goals.First,the training and deployment of an end-to-end deep convolutional neural network speech recognition model,and its performance evaluation.Second,the collection of Chinese voice data uploaded by users is realized,which can be used to enrich the training data set of the voice recognition model,which is conducive to the improvement and optimization of the model in the future.