PWM调制波的多层感知神经网络音频识别

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.

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

Research on PWM modulation case fidelity analysis and audio signal of the audio signal frequency, phase distribution modeling and identification. Audio recognition system is established in the Matlab/Simulink modeling environment, the use of filter modulation wave is normalized to complete pretreatment on signal. Audio signal model to construct 3 audio signal based on the. After modulation, recognition system filtering after the conditioning input to the trained multilayer perceptron neural network, analyses the identification results show that the audio frequency is higher, the lower carrier frequencies will reduce the recognition rate of the system; and to improve the carrier frequency can improve the recognition rate of the system; and was compared with the corresponding classical audio recognition scheme and this scheme.