节点文献
基于神经网络的武器声控系统的实现
Realization of Artificial Neural Network Based Speech-control Weapon System
【摘要】 文中介绍了利用神经网络进行实际系统设计的要点和实际中遇到的问题,由于语音信号的时域时变和频域漂移干扰神经网络,系统采用了降采样、频域归一化技术、倒谱域信号移位技术进行补偿。武器声控控制系统采用了VxWorks嵌入式实时操作系统,保证了对命令的及时响应。基于神经网络的武器声控控制系统能够大大地减轻武器操作的复杂程度,降低误操作率,并极大地提高作战效率。
【Abstract】 Some basic and practical tips for realization of the system were introduced.For the network works well some Endpoint detection technology must be adopted.To solve the invariant in time and cepstrum domain of speech signal,frequency shift and normalized frequency domain technology were chosen for compensation.Our operating system was VxWorks,it was embedded real-time operate system,and could respond the order in time.Simulation indicates that speech-controlled weapon system could decrease the complexity and mistake of operation,and improve efficiency.
【Key words】 speech-control; endpoint detection; backpropagation neural network; embedded real-time operating system;
- 【文献出处】 弹箭与制导学报 ,Journal of Projectiles,Rockets,Missiles and Guidance , 编辑部邮箱 ,2008年02期
- 【分类号】E920
- 【下载频次】88