PWM调制波的多层感知神经网络音频识别  

Multi-layer perceptron neural network for the audio recognition of PWM modulated waves

作　　者：刘晓杰[1] 章怀宇 董和磊[1] 

机构地区：[1]中北大学仪器与电子学院,山西太原030400

出　　处：《电声技术》2016年第9期43-46,共4页Audio Engineering

摘　　要：研究了在PWM调制情形下音频信号的保真度分析及音频信号的频率、相位分布建模及其识别。基于Matlab/Simulink建模环境下建立了音频识别系统,利用滤波器对调制波进行归一化操作来完成对信号的预处理。基于音频信号模型构造了3类音频信号。经过识别系统的调制、滤波调理后输入至训练好的多层感知神经网络,分析识别结果显示,音频频率越高时,载波频率较低会降低系统识别率;提高载波频率可提高系统的识别率;并与相应的经典音频识别方案进行了比较。Research on PWM modulation case fidelity analysis and audio signal of the audio signal frequency, phase distribution modeling and identification is made.Audio recognition system is established in the Matlab/Simulink modeling environment, and the use of filter modulation wave is normalized to complete pretreatment on signal. Three types of audio signal are constructed based on audio signal model. After modulation, recognition system filtering alter the conditioning input to the trained multilayer perceptron neural network, the identification results show that the audio frequency is higher, the lower carrier frequencies reduce the recognition rate of the system, higher carrier frequency can improve the recognition rate of the system. And it is compared with the corresponding classical audio recognition scheme and this scheme.

关 键 词：PWM调制技术 神经网络识别 音频信号建模 MATLAB/SIMULINK仿真 

分 类 号：TN912[电子电信—通信与信息系统]