Construction And Performance Of Speech Communication System Based On Piezoelectric Electret Mechanical Antenna

Low-frequency electromagnetic waves have strong penetration and low attenuation performance when propagating in solid and liquid dielectric,which makes them widely used in long-wave communication fields such as underwater communication and mineral detection.The antenna radiation unit of traditional resonant antenna long-wave communication system is usually comparable to the wavelength of electromagnetic wave,which makes the long-wave communication system built by resonant antenna usually larger in size,high power consumption and low power conversion rate.In recent years,a new type of mechanical antenna is expected to be used for long-wave communication systems with the advantages of miniaturization and low power consumption.This paper focuses on the piezoelectric electret mechanical antenna long-wave communication system,and the main work is as follows.First,a long-wave communication speech system based on piezoelectric mechanical antenna is designed and constructed.Firstly,the working mechanism of piezoelectric mechanical antenna is illustrated,and then a concrete construction of a long-wave communication speech system based on PZT-42 piezoelectric mechanical antenna is proposed,including hardware system,communication protocol and software system.Finally,the workflow and performance of the constructed long-wave communication speech system are analyzed through case studies and experimental tests.Second,a signal denoising algorithm of mechanical antenna based on denoising auto-encoder for long-wave communication is studied and implemented.Firstly,based on the time-serialized characteristics of the time-domain signal of long-wave communication,this algorithm uses long short-term memory neural network to extract the signal characteristics of mechanical antenna long-wave communication.Then,Gaussian noise is added to the communication signal,and the original signal is used as supervised information to train the denoising auto-encoder model by using self-supervised learning.Finally,the effects of parameters such as the number of training epochs and the number of hidden layer neurons on the performance of the model are experimentally investigated.