基于门控网络的军事装备控制指令语音识别研究

军事装备无感控制是军事装备智能化建设进程中的一个重要研究方向，其中语音控制技术作为无人装备无感控制手段的关键组成部分，受到了越来越多的重视。为完成军事装备语音控制任务，设计一种基于门控网络的中文语音识别网络，并构建军事装备控制指令数据集，实现基于控制指令语音识别技术的军事装备控制。在传统卷积神经网络的结构基础上引入深度残差门控卷积网络，提高识别网络的准确性，同时通过多途径构建军事装备控制指令数据集，设计一套针对军事装备无感控制的语音识别方案。实验结果表明，该语音识别网络军事语音控制指令识别率可达87%，外接语言模型后可达92%，语音识别准确率高、误差率低，可完成军事装备的语音控制任务。

Research on Voice Recognition of Military Equipment Control Commands Based on Gated Network

The sensorless control of military equipment is an important research direction in intelligent military equipment development.As a key component of the sensorless control of unmanned equipment,speech control has attracted more and more attention.To realize military equipment control based on the voice recognition of control commands,this paper describes the design of a Chinese voice recognition network based on a gated network,and a data set of military equipment control commands.Based on the structure of the traditional convolutional neural networks,a deep residual gated convolutional network is introduced to improve the accuracy of the recognition network.At the same time,a data set of military equipment control commands is constructed by using multiple methods,and a voice recognition scheme for sensorless control of military equipment is designed.Experimental results show that the recognition rate of the proposed voice recognition network reaches 87%,and 92% when connected to a language model. This network improves the accuracy of voice recognition while reducing the errors,and can complete the voice control task of military equipment.