Microphone Array Speech Enhancement Technology For UAV Platform

In recent years,multi-rotor UAV is becoming more and more popular in various fields.In addition to military applications,UAV has a wide range of applications in agricultural,police,rescue,infrastructure,commercial and personal fields.The realization of these application scenarios mainly depends on the progress of sensor technology carried on the UAV platform.Therefore,it is a promising research direction to improve the type and performance of sensors carried on the UAV platform.Microphone array technology is a good method to solve the problem of long-distance sound acquisition in recent years.with its advantage of being sensitive to the spatial characteristics of sound field,microphone array is widely used in sound acquisition in far-field,high noise and high reverberation environment.it has become a standard hardware in intelligent devices,speech recognition and other fields.Therefore,the application of microphone array to UAV for sound signal acquisition is a new microphone array scene application.Through the research and analysis of the microphone array algorithm,it is found that although many microphone array algorithms achieve good results in the indoor low SNR environment,on the UAV platform,due to the very high self-noise of the UAV,many microphone array algorithms can not be effective,which makes it very difficult to use the microphone array for sound acquisition on the UAV platform.So first of all,this paper analyzes the self-noise generated by the multi-rotor UAV,analyzes the most suitable location of the microphone array,and constructs a prototype of the UAV equipped with the microphone array.In this paper,the problem of speech signal enhancement on UAV is deeply analyzed,and two schemes are proposed and tested on the experimental platform.The specific research work includes the following two aspects:First of all,this paper proposes an improved amplitude normalized controllable beam response algorithm PHAT SRP.The core of this algorithm is to add PHAT phase transformation to the traditional SRPPHAT algorithm,which improves the robustness of the algorithm in low signal-to-noise ratio(SNR)environment.At the same time,the algorithm focuses on speech signal enhancement,improves the quality of speech signal through post-filter,and finally obtains accurate source localization performance and high quality speech signal quality.Secondly,a spatial filter algorithm SP for time-frequency analysis is proposed in this paper.In this algorithm,a spatial likelihood function for UAV self-noise is constructed by using time-frequency analysis technology and the characteristics of UAV self-noise,and spatial filtering is carried out.This algorithm is compared with other microphone array algorithms.the result is better.