VHF Speech Enhancement Based On Short-Time Spectrum Estimation

With the rapid development of civil aviation,the number of aircraft is increasing and the flight flow is increasing.If aircraft routes are dense in the same airspace at the same time,it is easy to cause air traffic congestion,which may induce accidents when serious.Therefore,the requirements for civil aviation control are getting higher and higher,and it is necessary to provide reliable communication connection for close monitoring.The communication means of civil aviation are mainly Very High Frequency(VHF),which guarantees that voice communication,information transmission and real-time monitoring between aircraft and aircraft or aircraft and the ground,and has good effects over medium and short distance.However,due to the particularity of VHF frequency band,it is easy to be invaded by many unknown interferences.At the same time,the space electromagnetic environment is getting worse and worse due to the opening of part of the airspace to the outside world and the shortage of resources,leading to the increasingly serious interference problem of VHF communication.For example,noise interference and even speech distortion occur from time to time in the process of voice communication,resulting in blocked communication between pilots and controllers,which seriously affects the safety of flight.Therefore,this paper will mainly suppress the noise interference in the process of VHF voice communication and achieve speech enhancement.The main contents are as follows:Firstly,referring to the speech signal processing technology and combining with the characteristic parameters of VHF speech signal,the relevant content of noise estimation and the process of speech enhancement are expounded in detail,which provides a theoretical basis and thinking for the analysis and suppression of noise interference under the following research scenario.Secondly,an algorithm for detecting and suppressing impulse noise in air-ground communication system is proposed.The algorithm sets the minimum time threshold and suppresses the impulse noise combining the autocorrelation characteristics of speech signal and the duration of impulse noise in this scenario,which not only improves the quality of voice communication,but also improves the accuracy of speech endpoint detection.The experimental results show that the proposed method is effective on account of it does not appear the speech distortion and the impulse noise obvious residual phenomenon.It is very suitable for the real-time requirement of air traffic control system,because the algorithm is not involving frequency domain transformation so as the time complexity is low.Thirdly,a VHF speech enhancement algorithm based on improved power spectrum subtraction is proposed.The improved speech presence probability(SPP)is used to track the noise adaptively,which has a good noise estimation effect in a non-stationary environment.First,the ratio of energy to spectral entropy,which the ratio of the characteristic energy in the time domain to the characteristic spectrum entropy in the frequency domain,is used as a new feature to construct the proportional relationship with the SPP,so as to obtain the estimated SPP value of the current speech frame,and the value is smoothed with the double smoothing coefficient.In the next place,the estimated noise power spectrum is obtained by combining with the time recursive average algorithm and then the speech enhancement is realized by spectrum subtraction.The effectiveness of the algorithm is verified by simulation experiments from subjective auditory perception and objective evaluation criteria.In addition,for the speech signal with serious background noise in the air-ground communication environment,the enhanced voice quality is also significantly improved,effectively solving the problem of noise interference in the VHF voice communication in the air traffic control system.