Speech Enhancement Under Aviation Noise

By analyzing the characteristics of the spectrum of aviation noise, short-time energy in frequency domain of the signals is used to detect voice activity while signal noise ratio (SNR) is very low. Transition frames are added through forward decision and backward decision and divided into different groups in detail according to the energy by employing a modified method of spectral subtraction. In addition, speech frames, transition frames, and noise frames are processed separately in order to get rid of music noise. The experimental results show that the method can offer better performances of the speech under strong aviation noise. Simulation results show that both SNR and articulation are greatly improved