Noise Suppression Of Rotor Aircraft And Speech Enhancement In Cabin

With the rapid development of modern military theories,military ideas,combat methods,and science,the development of armament equipment is very rapid.Because the advantages of vertical lifting,hovering,flipping,slow flying and flying backward,it has become an important weapon for low-altitude,ultra-low altitude penetration and low-level bombing.However,the rotation of the rotor of the aircraft disturbs air and generates strong aerodynamic noise,which seriously affect the concealment of the aircraft.At the same time,the rotorcraft will radiate noise on the ground when flying at low altitude,which seriously interferes with the communication quality of the ground voice interactive system.Based on this starting point,this paper studies the suppression of rotorcraft noise.In this paper,the suppression of rotorcraft noise is mainly carried out in two aspects: First,active adaptive cancellation of aircraft rotor noise.Second,extract and enhance meaningful speech signals in the rotor noise environment.First,the rotorcraft noise characteristics are studied.The rotorcraft noise sources are classified,and then the rotorcraft noise is qualitatively analyzed from three perspectives:the production mechanism,the frequency spectrum characteristics,and the nonstationarity.Finally,an ARMA model was established for rotor noise by time series analysis,and the reliability of the model was verified through experiments.Then,the active adaptive elimination of rotor noise was studied.Based on the analysis of the principle of active self-adaptation elimination and the FXLMS and FXNLMS active adaptive algorithm,the feasibility of active self-adaptive elimination under ideal conditions is verified by experimental simulation.In order to introduce the nonlinearity of the secondary channel in the actual situation,this paper proposes a secondary channel identification method based on the adaptive kernel filter,and compares the LMS,KLMS and KAPA algorithms.Experiments show that the steady-state error of the KAPA algorithm The convergence speed is better than the LMS and KLMS algorithms.For different types of rotorcraft,different secondary sound sources and error sensor installation strategies were designed.Aiming at the time-varying problem of secondary channels,the on-line identification method of secondary channels based on KAPA algorithm is proposed.The superiority of this method is verified by simulation experiments.Finally,the extraction and enhancement of speech in the rotor environment are studied.In the aspect of speech extraction,the two-threshold method and logarithmic spectral distance method in the speech endpoint detection algorithm are mainly analyzed.For the problem that the traditional speech extraction algorithm has poor effect on non-stationary noise,fusion Cadzow spectral estimation and LSFM characteristics are proposed.Voice endpoint detection method.The Cadzow algorithm uses the ARMA model of rotor noise to estimate the power spectrum and obtain the characteristic LSFM.By comparing the LSFM with the adaptive threshold,the speech endpoint can be determined.Experiments show that this method has a better detection effect on non-stationary rotor noise.In the aspect of speech enhancement,spectral subtraction and Wiener filtering are mainly studied.To solve the problem of inadequate phase enhancement in traditional speech enhancement algorithms,a speech enhancement algorithm based on the ideal combination mask ICM is proposed.The ICM is estimated through the deep neural network and combined with the phase compensation method.At the same time,the amplitude spectrum and phase spectrum of the speech signal are enhanced.After simulation experiments,it is verified that the algorithm can effectively suppress the background noise,and can significantly improve the intelligibility and automatic recognition rate of speech.